Protocol Buffers - Google's data interchange format (grpc依赖) https://developers.google.com/protocol-buffers/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

859 lines
35 KiB

// Protocol Buffers - Google's data interchange format
// Copyright 2023 Google LLC. All rights reserved.
//
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file or at
// https://developers.google.com/open-source/licenses/bsd
/* Test of mini table accessors.
*
* Messages are created and mutated using generated code, and then
* accessed through reflective APIs exposed through mini table accessors.
*/
#include "upb/message/promote.h"
#include <string.h>
#include <cstddef>
#include <cstdint>
#include <string>
#include <gtest/gtest.h>
#include "upb/base/descriptor_constants.h"
#include "upb/base/status.h"
#include "upb/base/string_view.h"
#include "upb/base/upcast.h"
#include "upb/mem/arena.h"
#include "upb/mem/arena.hpp"
#include "upb/message/accessors.h"
#include "upb/message/array.h"
#include "upb/message/copy.h"
#include "upb/message/internal/extension.h"
#include "upb/message/internal/message.h"
#include "upb/message/map.h"
#include "upb/message/message.h"
#include "upb/message/tagged_ptr.h"
#include "upb/mini_descriptor/decode.h"
#include "upb/mini_descriptor/internal/encode.hpp"
#include "upb/mini_descriptor/internal/modifiers.h"
#include "upb/mini_descriptor/link.h"
#include "upb/mini_table/extension.h"
#include "upb/mini_table/field.h"
#include "upb/mini_table/message.h"
#include "upb/test/test.upb.h"
#include "upb/test/test.upb_minitable.h"
#include "upb/wire/decode.h"
#include "upb/wire/encode.h"
namespace {
TEST(GeneratedCode, FindUnknown) {
upb_Arena* arena = upb_Arena_New();
upb_test_ModelWithExtensions* msg = upb_test_ModelWithExtensions_new(arena);
upb_test_ModelWithExtensions_set_random_int32(msg, 10);
upb_test_ModelWithExtensions_set_random_name(
msg, upb_StringView_FromString("Hello"));
upb_test_ModelExtension1* extension1 = upb_test_ModelExtension1_new(arena);
upb_test_ModelExtension1_set_str(extension1,
upb_StringView_FromString("World"));
upb_test_ModelExtension1_set_model_ext(msg, extension1, arena);
size_t serialized_size;
char* serialized =
upb_test_ModelWithExtensions_serialize(msg, arena, &serialized_size);
upb_test_EmptyMessageWithExtensions* base_msg =
upb_test_EmptyMessageWithExtensions_parse(serialized, serialized_size,
arena);
upb_FindUnknownRet result = upb_Message_FindUnknown(
UPB_UPCAST(base_msg),
upb_MiniTableExtension_Number(&upb_test_ModelExtension1_model_ext_ext),
0);
EXPECT_EQ(kUpb_FindUnknown_Ok, result.status);
result = upb_Message_FindUnknown(
UPB_UPCAST(base_msg),
upb_MiniTableExtension_Number(&upb_test_ModelExtension2_model_ext_ext),
0);
EXPECT_EQ(kUpb_FindUnknown_NotPresent, result.status);
upb_Arena_Free(arena);
}
TEST(GeneratedCode, Extensions) {
upb_Arena* arena = upb_Arena_New();
upb_test_ModelWithExtensions* msg = upb_test_ModelWithExtensions_new(arena);
upb_test_ModelWithExtensions_set_random_int32(msg, 10);
upb_test_ModelWithExtensions_set_random_name(
msg, upb_StringView_FromString("Hello"));
upb_test_ModelExtension1* extension1 = upb_test_ModelExtension1_new(arena);
upb_test_ModelExtension1_set_str(extension1,
upb_StringView_FromString("World"));
upb_test_ModelExtension2* extension2 = upb_test_ModelExtension2_new(arena);
upb_test_ModelExtension2_set_i(extension2, 5);
upb_test_ModelExtension2* extension3 = upb_test_ModelExtension2_new(arena);
upb_test_ModelExtension2_set_i(extension3, 6);
upb_test_ModelExtension2* extension4 = upb_test_ModelExtension2_new(arena);
upb_test_ModelExtension2_set_i(extension4, 7);
upb_test_ModelExtension2* extension5 = upb_test_ModelExtension2_new(arena);
upb_test_ModelExtension2_set_i(extension5, 8);
upb_test_ModelExtension2* extension6 = upb_test_ModelExtension2_new(arena);
upb_test_ModelExtension2_set_i(extension6, 9);
// Set many extensions, to exercise code paths that involve reallocating the
// extensions and unknown fields array.
upb_test_ModelExtension1_set_model_ext(msg, extension1, arena);
upb_test_ModelExtension2_set_model_ext(msg, extension2, arena);
upb_test_ModelExtension2_set_model_ext_2(msg, extension3, arena);
upb_test_ModelExtension2_set_model_ext_3(msg, extension4, arena);
upb_test_ModelExtension2_set_model_ext_4(msg, extension5, arena);
upb_test_ModelExtension2_set_model_ext_5(msg, extension6, arena);
size_t serialized_size;
char* serialized =
upb_test_ModelWithExtensions_serialize(msg, arena, &serialized_size);
upb_test_ModelExtension1* ext1;
upb_test_ModelExtension2* ext2;
upb_GetExtension_Status promote_status;
upb_MessageValue value;
// Test known GetExtension 1
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension1_model_ext_ext, 0, arena,
&value);
ext1 = (upb_test_ModelExtension1*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_TRUE(upb_StringView_IsEqual(upb_StringView_FromString("World"),
upb_test_ModelExtension1_str(ext1)));
// Test known GetExtension 2
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension2_model_ext_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(5, upb_test_ModelExtension2_i(ext2));
// Test known GetExtension 3
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension2_model_ext_2_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(6, upb_test_ModelExtension2_i(ext2));
// Test known GetExtension 4
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension2_model_ext_3_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(7, upb_test_ModelExtension2_i(ext2));
// Test known GetExtension 5
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension2_model_ext_4_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(8, upb_test_ModelExtension2_i(ext2));
// Test known GetExtension 6
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(msg), &upb_test_ModelExtension2_model_ext_5_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(9, upb_test_ModelExtension2_i(ext2));
upb_test_EmptyMessageWithExtensions* base_msg =
upb_test_EmptyMessageWithExtensions_parse(serialized, serialized_size,
arena);
// Get unknown extension bytes before promotion.
size_t start_len;
upb_Message_GetUnknown(UPB_UPCAST(base_msg), &start_len);
EXPECT_GT(start_len, 0);
EXPECT_EQ(0, upb_Message_ExtensionCount(UPB_UPCAST(base_msg)));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension1_model_ext_ext, 0, arena,
&value);
ext1 = (upb_test_ModelExtension1*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_TRUE(upb_StringView_IsEqual(upb_StringView_FromString("World"),
upb_test_ModelExtension1_str(ext1)));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension2_model_ext_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(5, upb_test_ModelExtension2_i(ext2));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension2_model_ext_2_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(6, upb_test_ModelExtension2_i(ext2));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension2_model_ext_3_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(7, upb_test_ModelExtension2_i(ext2));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension2_model_ext_4_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(8, upb_test_ModelExtension2_i(ext2));
// Test unknown GetExtension.
promote_status = upb_Message_GetOrPromoteExtension(
UPB_UPCAST(base_msg), &upb_test_ModelExtension2_model_ext_5_ext, 0, arena,
&value);
ext2 = (upb_test_ModelExtension2*)value.msg_val;
EXPECT_EQ(kUpb_GetExtension_Ok, promote_status);
EXPECT_EQ(9, upb_test_ModelExtension2_i(ext2));
size_t end_len;
upb_Message_GetUnknown(UPB_UPCAST(base_msg), &end_len);
EXPECT_LT(end_len, start_len);
EXPECT_EQ(6, upb_Message_ExtensionCount(UPB_UPCAST(base_msg)));
upb_Arena_Free(arena);
}
// Create a minitable to mimic ModelWithSubMessages with unlinked subs
// to lazily promote unknowns after parsing.
upb_MiniTable* CreateMiniTableWithEmptySubTables(upb_Arena* arena) {
upb::MtDataEncoder e;
e.StartMessage(0);
e.PutField(kUpb_FieldType_Int32, 4, 0);
e.PutField(kUpb_FieldType_Message, 5, 0);
e.PutField(kUpb_FieldType_Message, 6, kUpb_FieldModifier_IsRepeated);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
EXPECT_EQ(status.ok, true);
return table;
}
upb_MiniTable* CreateMapEntryMiniTable(upb_Arena* arena) {
upb::MtDataEncoder e;
e.EncodeMap(kUpb_FieldType_Int32, kUpb_FieldType_Message, 0, 0);
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
EXPECT_EQ(status.ok, true);
return table;
}
// Create a minitable to mimic ModelWithMaps with unlinked subs
// to lazily promote unknowns after parsing.
upb_MiniTable* CreateMiniTableWithEmptySubTablesForMaps(upb_Arena* arena) {
upb::MtDataEncoder e;
e.StartMessage(0);
e.PutField(kUpb_FieldType_Int32, 1, 0);
e.PutField(kUpb_FieldType_Message, 3, kUpb_FieldModifier_IsRepeated);
e.PutField(kUpb_FieldType_Message, 5, kUpb_FieldModifier_IsRepeated);
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
// Field 5 corresponds to ModelWithMaps.map_sm.
upb_MiniTableField* map_field = const_cast<upb_MiniTableField*>(
upb_MiniTable_FindFieldByNumber(table, 5));
EXPECT_NE(map_field, nullptr);
upb_MiniTable* sub_table = CreateMapEntryMiniTable(arena);
upb_MiniTable_SetSubMessage(table, map_field, sub_table);
EXPECT_EQ(status.ok, true);
return table;
}
void CheckReserialize(const upb_Message* msg, const upb_MiniTable* mini_table,
upb_Arena* arena, char* serialized,
size_t serialized_size) {
// We can safely encode the "empty" message. We expect to get the same bytes
// out as were parsed.
size_t reserialized_size;
char* reserialized;
upb_EncodeStatus encode_status =
upb_Encode(msg, mini_table, kUpb_EncodeOption_Deterministic, arena,
&reserialized, &reserialized_size);
EXPECT_EQ(encode_status, kUpb_EncodeStatus_Ok);
EXPECT_EQ(reserialized_size, serialized_size);
EXPECT_EQ(0, memcmp(reserialized, serialized, serialized_size));
// We should get the same result if we copy+reserialize.
upb_Message* clone = upb_Message_DeepClone(msg, mini_table, arena);
encode_status = upb_Encode(clone, mini_table, kUpb_EncodeOption_Deterministic,
arena, &reserialized, &reserialized_size);
EXPECT_EQ(encode_status, kUpb_EncodeStatus_Ok);
EXPECT_EQ(reserialized_size, serialized_size);
EXPECT_EQ(0, memcmp(reserialized, serialized, serialized_size));
}
TEST(GeneratedCode, PromoteUnknownMessage) {
upb::Arena arena;
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena.ptr());
upb_test_ModelWithExtensions* sub_message =
upb_test_ModelWithExtensions_new(arena.ptr());
upb_test_ModelWithSubMessages_set_id(input_msg, 11);
upb_test_ModelWithExtensions_set_random_int32(sub_message, 12);
upb_test_ModelWithSubMessages_set_optional_child(input_msg, sub_message);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(
input_msg, arena.ptr(), &serialized_size);
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTables(arena.ptr());
upb_DecodeStatus decode_status;
// If we parse without allowing unlinked objects, the parse will fail.
// TODO: re-enable this test once the old method of tree shaking is
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// removed
// upb_Message* fail_msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// decode_status =
// upb_Decode(serialized, serialized_size, fail_msg, mini_table, nullptr,
// 0,
// arena.ptr());
// EXPECT_EQ(decode_status, kUpb_DecodeStatus_UnlinkedSubMessage);
// if we parse while allowing unlinked objects, the parse will succeed.
upb_Message* msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
CheckReserialize(msg, mini_table, arena.ptr(), serialized, serialized_size);
// We can encode the "empty" message and get the same output bytes.
size_t reserialized_size;
char* reserialized;
upb_EncodeStatus encode_status = upb_Encode(
msg, mini_table, 0, arena.ptr(), &reserialized, &reserialized_size);
EXPECT_EQ(encode_status, kUpb_EncodeStatus_Ok);
EXPECT_EQ(reserialized_size, serialized_size);
EXPECT_EQ(0, memcmp(reserialized, serialized, serialized_size));
// Int32 field is present, as normal.
int32_t val = upb_Message_GetInt32(
msg, upb_MiniTable_FindFieldByNumber(mini_table, 4), 0);
EXPECT_EQ(val, 11);
// Unlinked sub-message is present, but getting the value returns NULL.
const upb_MiniTableField* submsg_field =
upb_MiniTable_FindFieldByNumber(mini_table, 5);
ASSERT_TRUE(submsg_field != nullptr);
EXPECT_TRUE(upb_Message_HasBaseField(msg, submsg_field));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_TaggedMessagePtr tagged =
upb_Message_GetTaggedMessagePtr(msg, submsg_field, nullptr);
EXPECT_TRUE(upb_TaggedMessagePtr_IsEmpty(tagged));
// Update mini table and promote unknown to a message.
EXPECT_TRUE(
upb_MiniTable_SetSubMessage(mini_table, (upb_MiniTableField*)submsg_field,
&upb_0test__ModelWithExtensions_msg_init));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
const int decode_options =
upb_DecodeOptions_MaxDepth(0); // UPB_DECODE_ALIAS disabled.
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_test_ModelWithExtensions* promoted;
upb_DecodeStatus promote_result =
upb_Message_PromoteMessage(msg, mini_table, submsg_field, decode_options,
arena.ptr(), (upb_Message**)&promoted);
EXPECT_EQ(promote_result, kUpb_DecodeStatus_Ok);
EXPECT_NE(nullptr, promoted);
EXPECT_EQ(UPB_UPCAST(promoted), upb_Message_GetMessage(msg, submsg_field));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(promoted), 12);
}
// Tests a second parse that reuses an empty/unlinked message while the message
// is still unlinked.
TEST(GeneratedCode, ReparseUnlinked) {
upb::Arena arena;
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena.ptr());
upb_test_ModelWithExtensions* sub_message =
upb_test_ModelWithExtensions_new(arena.ptr());
upb_test_ModelWithSubMessages_set_id(input_msg, 11);
upb_test_ModelWithExtensions_add_repeated_int32(sub_message, 12, arena.ptr());
upb_test_ModelWithSubMessages_set_optional_child(input_msg, sub_message);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(
input_msg, arena.ptr(), &serialized_size);
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTables(arena.ptr());
// Parse twice without linking the MiniTable.
upb_Message* msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_DecodeStatus decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
// Update mini table and promote unknown to a message.
const upb_MiniTableField* submsg_field =
upb_MiniTable_FindFieldByNumber(mini_table, 5);
EXPECT_TRUE(
upb_MiniTable_SetSubMessage(mini_table, (upb_MiniTableField*)submsg_field,
&upb_0test__ModelWithExtensions_msg_init));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
const int decode_options =
upb_DecodeOptions_MaxDepth(0); // UPB_DECODE_ALIAS disabled.
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_test_ModelWithExtensions* promoted;
upb_DecodeStatus promote_result =
upb_Message_PromoteMessage(msg, mini_table, submsg_field, decode_options,
arena.ptr(), (upb_Message**)&promoted);
EXPECT_EQ(promote_result, kUpb_DecodeStatus_Ok);
EXPECT_NE(nullptr, promoted);
EXPECT_EQ(UPB_UPCAST(promoted), upb_Message_GetMessage(msg, submsg_field));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// The repeated field should have two entries for the two parses.
size_t repeated_size;
const int32_t* entries =
upb_test_ModelWithExtensions_repeated_int32(promoted, &repeated_size);
EXPECT_EQ(repeated_size, 2);
EXPECT_EQ(entries[0], 12);
EXPECT_EQ(entries[1], 12);
}
// Tests a second parse that promotes a message within the parser because we are
// merging into an empty/unlinked message after the message has been linked.
TEST(GeneratedCode, PromoteInParser) {
upb::Arena arena;
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena.ptr());
upb_test_ModelWithExtensions* sub_message =
upb_test_ModelWithExtensions_new(arena.ptr());
upb_test_ModelWithSubMessages_set_id(input_msg, 11);
upb_test_ModelWithExtensions_add_repeated_int32(sub_message, 12, arena.ptr());
upb_test_ModelWithSubMessages_set_optional_child(input_msg, sub_message);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(
input_msg, arena.ptr(), &serialized_size);
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTables(arena.ptr());
// Parse once without linking the MiniTable.
upb_Message* msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_DecodeStatus decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
// Link the MiniTable.
const upb_MiniTableField* submsg_field =
upb_MiniTable_FindFieldByNumber(mini_table, 5);
EXPECT_TRUE(
upb_MiniTable_SetSubMessage(mini_table, (upb_MiniTableField*)submsg_field,
&upb_0test__ModelWithExtensions_msg_init));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// Parse again. This will promote the message. An explicit promote will not
// be required.
decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
upb_test_ModelWithExtensions* promoted =
(upb_test_ModelWithExtensions*)upb_Message_GetMessage(msg, submsg_field);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
EXPECT_NE(nullptr, promoted);
EXPECT_EQ(UPB_UPCAST(promoted), upb_Message_GetMessage(msg, submsg_field));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// The repeated field should have two entries for the two parses.
size_t repeated_size;
const int32_t* entries =
upb_test_ModelWithExtensions_repeated_int32(promoted, &repeated_size);
EXPECT_EQ(repeated_size, 2);
EXPECT_EQ(entries[0], 12);
EXPECT_EQ(entries[1], 12);
}
TEST(GeneratedCode, PromoteUnknownRepeatedMessage) {
upb::Arena arena;
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena.ptr());
upb_test_ModelWithSubMessages_set_id(input_msg, 123);
// Add 2 repeated messages to input_msg.
upb_test_ModelWithExtensions* item =
upb_test_ModelWithSubMessages_add_items(input_msg, arena.ptr());
upb_test_ModelWithExtensions_set_random_int32(item, 5);
item = upb_test_ModelWithSubMessages_add_items(input_msg, arena.ptr());
upb_test_ModelWithExtensions_set_random_int32(item, 6);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(
input_msg, arena.ptr(), &serialized_size);
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTables(arena.ptr());
upb_DecodeStatus decode_status;
// If we parse without allowing unlinked objects, the parse will fail.
// TODO: re-enable this test once the old method of tree shaking is
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// removed
// upb_Message* fail_msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
// decode_status =
// upb_Decode(serialized, serialized_size, fail_msg, mini_table, nullptr,
// 0,
// arena.ptr());
// EXPECT_EQ(decode_status, kUpb_DecodeStatus_UnlinkedSubMessage);
// if we parse while allowing unlinked objects, the parse will succeed.
upb_Message* msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
CheckReserialize(msg, mini_table, arena.ptr(), serialized, serialized_size);
// Int32 field is present, as normal.
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
int32_t val = upb_Message_GetInt32(
msg, upb_MiniTable_FindFieldByNumber(mini_table, 4), 0);
EXPECT_EQ(val, 123);
const upb_MiniTableField* repeated_field =
upb_MiniTable_FindFieldByNumber(mini_table, 6);
upb_Array* array = upb_Message_GetMutableArray(msg, repeated_field);
// Array length is 2 even though the messages are empty.
EXPECT_EQ(2, upb_Array_Size(array));
// Update mini table and promote unknown to a message.
EXPECT_TRUE(upb_MiniTable_SetSubMessage(
mini_table, (upb_MiniTableField*)repeated_field,
&upb_0test__ModelWithExtensions_msg_init));
const int decode_options =
upb_DecodeOptions_MaxDepth(0); // UPB_DECODE_ALIAS disabled.
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_DecodeStatus promote_result =
upb_Array_PromoteMessages(array, &upb_0test__ModelWithExtensions_msg_init,
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
decode_options, arena.ptr());
EXPECT_EQ(promote_result, kUpb_DecodeStatus_Ok);
const upb_Message* promoted_message = upb_Array_Get(array, 0).msg_val;
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(
(upb_test_ModelWithExtensions*)promoted_message),
5);
promoted_message = upb_Array_Get(array, 1).msg_val;
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(
(upb_test_ModelWithExtensions*)promoted_message),
6);
}
TEST(GeneratedCode, PromoteUnknownToMap) {
upb::Arena arena;
upb_test_ModelWithMaps* input_msg = upb_test_ModelWithMaps_new(arena.ptr());
upb_test_ModelWithMaps_set_id(input_msg, 123);
upb_test_ModelWithExtensions* submsg1 =
upb_test_ModelWithExtensions_new(arena.ptr());
upb_test_ModelWithExtensions_set_random_int32(submsg1, 123);
upb_test_ModelWithExtensions* submsg2 =
upb_test_ModelWithExtensions_new(arena.ptr());
upb_test_ModelWithExtensions_set_random_int32(submsg2, 456);
// Add 2 map entries.
upb_test_ModelWithMaps_map_im_set(input_msg, 111, submsg1, arena.ptr());
upb_test_ModelWithMaps_map_im_set(input_msg, 222, submsg2, arena.ptr());
size_t serialized_size;
char* serialized = upb_test_ModelWithMaps_serialize_ex(
input_msg, kUpb_EncodeOption_Deterministic, arena.ptr(),
&serialized_size);
upb_MiniTable* mini_table =
CreateMiniTableWithEmptySubTablesForMaps(arena.ptr());
// If we parse without allowing unlinked objects, the parse will fail.
upb_Message* fail_msg1 = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_DecodeStatus decode_status =
upb_Decode(serialized, serialized_size, fail_msg1, mini_table, nullptr, 0,
arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_UnlinkedSubMessage);
// if we parse while allowing unlinked objects, the parse will succeed.
upb_Message* msg = _upb_Message_New(mini_table, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
kUpb_DecodeOption_ExperimentalAllowUnlinked, arena.ptr());
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
CheckReserialize(msg, mini_table, arena.ptr(), serialized, serialized_size);
upb_MiniTableField* map_field = const_cast<upb_MiniTableField*>(
upb_MiniTable_FindFieldByNumber(mini_table, 5));
upb_Map* map = upb_Message_GetMutableMap(msg, map_field);
// Map size is 2 even though messages are unlinked.
EXPECT_EQ(2, upb_Map_Size(map));
// Update mini table and promote unknown to a message.
upb_MiniTable* entry = const_cast<upb_MiniTable*>(
upb_MiniTable_GetSubMessageTable(mini_table, map_field));
upb_MiniTableField* entry_value = const_cast<upb_MiniTableField*>(
upb_MiniTable_FindFieldByNumber(entry, 2));
upb_MiniTable_SetSubMessage(entry, entry_value,
&upb_0test__ModelWithExtensions_msg_init);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_DecodeStatus promote_result = upb_Map_PromoteMessages(
map, &upb_0test__ModelWithExtensions_msg_init, 0, arena.ptr());
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
EXPECT_EQ(promote_result, kUpb_DecodeStatus_Ok);
upb_MessageValue key;
upb_MessageValue val;
key.int32_val = 111;
EXPECT_TRUE(upb_Map_Get(map, key, &val));
EXPECT_EQ(123, upb_test_ModelWithExtensions_random_int32(
static_cast<const upb_test_ModelWithExtensions*>(
(void*)(val.msg_val))));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
key.int32_val = 222;
EXPECT_TRUE(upb_Map_Get(map, key, &val));
EXPECT_EQ(456, upb_test_ModelWithExtensions_random_int32(
static_cast<const upb_test_ModelWithExtensions*>(
(void*)(val.msg_val))));
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
}
} // namespace
// OLD tests, to be removed!
namespace {
// Create a minitable to mimic ModelWithSubMessages with unlinked subs
// to lazily promote unknowns after parsing.
upb_MiniTable* CreateMiniTableWithEmptySubTablesOld(upb_Arena* arena) {
upb::MtDataEncoder e;
e.StartMessage(0);
e.PutField(kUpb_FieldType_Int32, 4, 0);
e.PutField(kUpb_FieldType_Message, 5, 0);
e.PutField(kUpb_FieldType_Message, 6, kUpb_FieldModifier_IsRepeated);
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
EXPECT_EQ(status.ok, true);
return table;
}
// Create a minitable to mimic ModelWithMaps with unlinked subs
// to lazily promote unknowns after parsing.
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_MiniTable* CreateMiniTableWithEmptySubTablesForMapsOld(upb_Arena* arena) {
upb::MtDataEncoder e;
e.StartMessage(0);
e.PutField(kUpb_FieldType_Int32, 1, 0);
e.PutField(kUpb_FieldType_Message, 3, kUpb_FieldModifier_IsRepeated);
e.PutField(kUpb_FieldType_Message, 4, kUpb_FieldModifier_IsRepeated);
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
EXPECT_EQ(status.ok, true);
return table;
}
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_MiniTable* CreateMapEntryMiniTableOld(upb_Arena* arena) {
upb::MtDataEncoder e;
e.EncodeMap(kUpb_FieldType_String, kUpb_FieldType_String, 0, 0);
upb_Status status;
upb_Status_Clear(&status);
upb_MiniTable* table =
upb_MiniTable_Build(e.data().data(), e.data().size(), arena, &status);
EXPECT_EQ(status.ok, true);
return table;
}
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
TEST(GeneratedCode, PromoteUnknownMessageOld) {
upb_Arena* arena = upb_Arena_New();
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena);
upb_test_ModelWithExtensions* sub_message =
upb_test_ModelWithExtensions_new(arena);
upb_test_ModelWithSubMessages_set_id(input_msg, 11);
upb_test_ModelWithExtensions_set_random_int32(sub_message, 12);
upb_test_ModelWithSubMessages_set_optional_child(input_msg, sub_message);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(input_msg, arena,
&serialized_size);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTablesOld(arena);
upb_Message* msg = _upb_Message_New(mini_table, arena);
upb_DecodeStatus decode_status = upb_Decode(serialized, serialized_size, msg,
mini_table, nullptr, 0, arena);
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
int32_t val = upb_Message_GetInt32(
msg, upb_MiniTable_FindFieldByNumber(mini_table, 4), 0);
EXPECT_EQ(val, 11);
upb_FindUnknownRet unknown = upb_Message_FindUnknown(msg, 5, 0);
EXPECT_EQ(unknown.status, kUpb_FindUnknown_Ok);
// Update mini table and promote unknown to a message.
EXPECT_TRUE(upb_MiniTable_SetSubMessage(
mini_table,
(upb_MiniTableField*)upb_MiniTable_GetFieldByIndex(mini_table, 1),
&upb_0test__ModelWithExtensions_msg_init));
const int decode_options =
upb_DecodeOptions_MaxDepth(0); // UPB_DECODE_ALIAS disabled.
upb_UnknownToMessageRet promote_result =
upb_MiniTable_PromoteUnknownToMessage(
msg, mini_table, upb_MiniTable_GetFieldByIndex(mini_table, 1),
&upb_0test__ModelWithExtensions_msg_init, decode_options, arena);
EXPECT_EQ(promote_result.status, kUpb_UnknownToMessage_Ok);
const upb_Message* promoted_message =
upb_Message_GetMessage(msg, upb_MiniTable_GetFieldByIndex(mini_table, 1));
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(
(upb_test_ModelWithExtensions*)promoted_message),
12);
upb_Arena_Free(arena);
}
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
TEST(GeneratedCode, PromoteUnknownRepeatedMessageOld) {
upb_Arena* arena = upb_Arena_New();
upb_test_ModelWithSubMessages* input_msg =
upb_test_ModelWithSubMessages_new(arena);
upb_test_ModelWithSubMessages_set_id(input_msg, 123);
// Add 2 repeated messages to input_msg.
upb_test_ModelWithExtensions* item =
upb_test_ModelWithSubMessages_add_items(input_msg, arena);
upb_test_ModelWithExtensions_set_random_int32(item, 5);
item = upb_test_ModelWithSubMessages_add_items(input_msg, arena);
upb_test_ModelWithExtensions_set_random_int32(item, 6);
size_t serialized_size;
char* serialized = upb_test_ModelWithSubMessages_serialize(input_msg, arena,
&serialized_size);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_MiniTable* mini_table = CreateMiniTableWithEmptySubTablesOld(arena);
upb_Message* msg = _upb_Message_New(mini_table, arena);
upb_DecodeStatus decode_status = upb_Decode(serialized, serialized_size, msg,
mini_table, nullptr, 0, arena);
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
int32_t val = upb_Message_GetInt32(
msg, upb_MiniTable_FindFieldByNumber(mini_table, 4), 0);
EXPECT_EQ(val, 123);
// Check that we have repeated field data in an unknown.
upb_FindUnknownRet unknown = upb_Message_FindUnknown(msg, 6, 0);
EXPECT_EQ(unknown.status, kUpb_FindUnknown_Ok);
// Update mini table and promote unknown to a message.
EXPECT_TRUE(upb_MiniTable_SetSubMessage(
mini_table,
(upb_MiniTableField*)upb_MiniTable_GetFieldByIndex(mini_table, 2),
&upb_0test__ModelWithExtensions_msg_init));
const int decode_options =
upb_DecodeOptions_MaxDepth(0); // UPB_DECODE_ALIAS disabled.
upb_UnknownToMessage_Status promote_result =
upb_MiniTable_PromoteUnknownToMessageArray(
msg, upb_MiniTable_GetFieldByIndex(mini_table, 2),
&upb_0test__ModelWithExtensions_msg_init, decode_options, arena);
EXPECT_EQ(promote_result, kUpb_UnknownToMessage_Ok);
upb_Array* array = upb_Message_GetMutableArray(
msg, upb_MiniTable_GetFieldByIndex(mini_table, 2));
const upb_Message* promoted_message = upb_Array_Get(array, 0).msg_val;
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(
(upb_test_ModelWithExtensions*)promoted_message),
5);
promoted_message = upb_Array_Get(array, 1).msg_val;
EXPECT_EQ(upb_test_ModelWithExtensions_random_int32(
(upb_test_ModelWithExtensions*)promoted_message),
6);
upb_Arena_Free(arena);
}
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
TEST(GeneratedCode, PromoteUnknownToMapOld) {
upb_Arena* arena = upb_Arena_New();
upb_test_ModelWithMaps* input_msg = upb_test_ModelWithMaps_new(arena);
upb_test_ModelWithMaps_set_id(input_msg, 123);
// Add 2 map entries.
upb_test_ModelWithMaps_map_ss_set(input_msg,
upb_StringView_FromString("key1"),
upb_StringView_FromString("value1"), arena);
upb_test_ModelWithMaps_map_ss_set(input_msg,
upb_StringView_FromString("key2"),
upb_StringView_FromString("value2"), arena);
size_t serialized_size;
char* serialized =
upb_test_ModelWithMaps_serialize(input_msg, arena, &serialized_size);
Added a new dynamic tree shaking model to upb, with the intention of removing the old model once YouTube has migrated. The `kUpb_DecodeOption_ExperimentalAllowUnlinked` flag to the decoder will enable the new behavior. When that flag is not passed, tree shaking with the old model will still be possible. "Dynamic tree shaking" in upb is a feature that allows messages to be parsed even if the MiniTables have not been fully linked. Unlinked sub-message fields can be parsed by preserving their data in the unknown fields. If the application later discovers that the message field is actually needed, the MiniTable can be patched to properly link that field, and existing message instances can "promote" the data from the unknown fields to an actual message of the correct type. Before this change, dynamic tree shaking stored unparsed message data in the unknown fields of the *parent*. In effect, we were treating the field as if it did not exist at all. This meant that parsing an unlinked field did not affect the hasbits or oneof cases of the parent, nor did it create a `upb_Array` or `upb_Map` for array/map fields. Only when a message was linked and promoted did any of these things occur. While this model had some amount of conceptual simplicity, it caused significant problems with oneofs. When multiple fields inside a single oneof are parsed from the wire, order matters, because later oneof fields must overwrite earlier ones. Dynamic tree shaking can mean that some fields in a oneof are linked while others are not. It is essential that we preserve this ordering semantic even when dynamic tree shaking is being used, but it is difficult to do if the oneof's data can be split between linked fields (which have been reified into parsed field data) and unlinked fields (whose data lives in the unknown fields of the parent). To solve this problem, this CL changes the representation for unlinked fields. Instead of being placed in the parent's unknown fields, we create an actual message instance for each unlinked message we parse, but we use a placeholder "empty message" MiniTable as the message's type. All of the message's data will therefore be placed into the "empty message's" unknown fields. But unlike before, this "empty message" is actually present according to the hasbits, oneof case, and `upb_Array`/`upb_Map` of the parent. This means that all of the oneof presence logic works as normal. Since the MiniTable can be patched at any time, we need a bit in the message instance itself to signal whether a pointer to a sub-message is an "empty message" or not. When dynamic tree shaking is in use, all users must be capable of recognizing an empty message and acting accordingly (promoting, etc) even if the MiniTable itself says that the field is linked. Because dynamic tree shaking imposes this extra requirement on users, we require that users pass an extra option to the decoder to allow parsing of unlinked sub-messages. Many existing users of upb (Ruby, PHP, Python, etc) will always have fully-linked MiniTables, so there is no reason for them to add extra logic to handle empty messages. By omitting the `kUpb_DecodeOption_ExperimentalAllowUnlinked` option, they will be relieved of the duty to check the tagged pointer that would indicate an empty, unlinked message. For existing users of dynamic tree shaking, there are three main changes: 1. The APIs in message/promote.h have changed, and users will need to update to the new interfaces. 2. The model for maps has changed slightly. Before, we required that map entries always had their values linked; for dynamic tree shaking to apply to maps, we required that the *entry* was left unlinked, not the entry's value. In the new model, that is reversed: map entries must always be linked, but a map entry's value can be unlinked. 3. The presence model for unlinked fields has changed. Unlinked fields will now register as "present" from the perspective of hasbits, oneof cases, and array/map entries. Users must test the tagged pointer to know if a message is of the correct, linked type or whether it is a placeholder "empty" message. There is a new function `upb_Message_GetTaggedMessagePtr()`, as well as a new accessor `upb_MessageValue.tagged_msg_val` that can be used to read and test the tagged pointer directly. PiperOrigin-RevId: 535288031
2 years ago
upb_MiniTable* mini_table =
CreateMiniTableWithEmptySubTablesForMapsOld(arena);
upb_MiniTable* map_entry_mini_table = CreateMapEntryMiniTableOld(arena);
upb_Message* msg = _upb_Message_New(mini_table, arena);
const int decode_options = upb_DecodeOptions_MaxDepth(0);
upb_DecodeStatus decode_status =
upb_Decode(serialized, serialized_size, msg, mini_table, nullptr,
decode_options, arena);
EXPECT_EQ(decode_status, kUpb_DecodeStatus_Ok);
int32_t val = upb_Message_GetInt32(
msg, upb_MiniTable_FindFieldByNumber(mini_table, 1), 0);
EXPECT_EQ(val, 123);
// Check that we have map data in an unknown.
upb_FindUnknownRet unknown = upb_Message_FindUnknown(msg, 3, 0);
EXPECT_EQ(unknown.status, kUpb_FindUnknown_Ok);
// Update mini table and promote unknown to a message.
EXPECT_TRUE(upb_MiniTable_SetSubMessage(
mini_table,
(upb_MiniTableField*)upb_MiniTable_GetFieldByIndex(mini_table, 1),
map_entry_mini_table));
upb_UnknownToMessage_Status promote_result =
upb_MiniTable_PromoteUnknownToMap(
msg, mini_table, upb_MiniTable_GetFieldByIndex(mini_table, 1),
decode_options, arena);
EXPECT_EQ(promote_result, kUpb_UnknownToMessage_Ok);
upb_Map* map = upb_Message_GetOrCreateMutableMap(
msg, map_entry_mini_table, upb_MiniTable_GetFieldByIndex(mini_table, 1),
arena);
EXPECT_NE(map, nullptr);
// Lookup in map.
upb_MessageValue key;
key.str_val = upb_StringView_FromString("key2");
upb_MessageValue value;
EXPECT_TRUE(upb_Map_Get(map, key, &value));
EXPECT_EQ(0, strncmp(value.str_val.data, "value2", 5));
upb_Arena_Free(arena);
}
} // namespace