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.

785 lines
22 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
#include "upb/json/encode.h"
#include <ctype.h>
#include <float.h>
#include <inttypes.h>
#include <math.h>
#include <stdarg.h>
#include <string.h>
#include "upb/lex/round_trip.h"
#include "upb/message/map.h"
#include "upb/port/vsnprintf_compat.h"
#include "upb/reflection/message.h"
#include "upb/wire/decode.h"
// Must be last.
#include "upb/port/def.inc"
typedef struct {
char *buf, *ptr, *end;
size_t overflow;
int indent_depth;
int options;
const upb_DefPool* ext_pool;
jmp_buf err;
upb_Status* status;
upb_Arena* arena;
} jsonenc;
static void jsonenc_msg(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m);
static void jsonenc_scalar(jsonenc* e, upb_MessageValue val,
const upb_FieldDef* f);
static void jsonenc_msgfield(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m);
static void jsonenc_msgfields(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m, bool first);
static void jsonenc_value(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m);
UPB_NORETURN static void jsonenc_err(jsonenc* e, const char* msg) {
upb_Status_SetErrorMessage(e->status, msg);
UPB_LONGJMP(e->err, 1);
}
UPB_PRINTF(2, 3)
UPB_NORETURN static void jsonenc_errf(jsonenc* e, const char* fmt, ...) {
va_list argp;
va_start(argp, fmt);
upb_Status_VSetErrorFormat(e->status, fmt, argp);
va_end(argp);
UPB_LONGJMP(e->err, 1);
}
static upb_Arena* jsonenc_arena(jsonenc* e) {
/* Create lazily, since it's only needed for Any */
if (!e->arena) {
e->arena = upb_Arena_New();
}
return e->arena;
}
static void jsonenc_putbytes(jsonenc* e, const void* data, size_t len) {
size_t have = e->end - e->ptr;
if (UPB_LIKELY(have >= len)) {
memcpy(e->ptr, data, len);
e->ptr += len;
} else {
if (have) {
memcpy(e->ptr, data, have);
e->ptr += have;
}
e->overflow += (len - have);
}
}
static void jsonenc_putstr(jsonenc* e, const char* str) {
jsonenc_putbytes(e, str, strlen(str));
}
UPB_PRINTF(2, 3)
static void jsonenc_printf(jsonenc* e, const char* fmt, ...) {
size_t n;
size_t have = e->end - e->ptr;
va_list args;
va_start(args, fmt);
n = _upb_vsnprintf(e->ptr, have, fmt, args);
va_end(args);
if (UPB_LIKELY(have > n)) {
e->ptr += n;
} else {
e->ptr = UPB_PTRADD(e->ptr, have);
e->overflow += (n - have);
}
}
static void jsonenc_nanos(jsonenc* e, int32_t nanos) {
int digits = 9;
if (nanos == 0) return;
if (nanos < 0 || nanos >= 1000000000) {
jsonenc_err(e, "error formatting timestamp as JSON: invalid nanos");
}
while (nanos % 1000 == 0) {
nanos /= 1000;
digits -= 3;
}
jsonenc_printf(e, ".%.*" PRId32, digits, nanos);
}
static void jsonenc_timestamp(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* seconds_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_FieldDef* nanos_f = upb_MessageDef_FindFieldByNumber(m, 2);
int64_t seconds = upb_Message_GetFieldByDef(msg, seconds_f).int64_val;
int32_t nanos = upb_Message_GetFieldByDef(msg, nanos_f).int32_val;
int L, N, I, J, K, hour, min, sec;
if (seconds < -62135596800) {
jsonenc_err(e,
"error formatting timestamp as JSON: minimum acceptable value "
"is 0001-01-01T00:00:00Z");
} else if (seconds > 253402300799) {
jsonenc_err(e,
"error formatting timestamp as JSON: maximum acceptable value "
"is 9999-12-31T23:59:59Z");
}
/* Julian Day -> Y/M/D, Algorithm from:
* Fliegel, H. F., and Van Flandern, T. C., "A Machine Algorithm for
* Processing Calendar Dates," Communications of the Association of
* Computing Machines, vol. 11 (1968), p. 657. */
seconds += 62135596800; // Ensure seconds is positive.
L = (int)(seconds / 86400) - 719162 + 68569 + 2440588;
N = 4 * L / 146097;
L = L - (146097 * N + 3) / 4;
I = 4000 * (L + 1) / 1461001;
L = L - 1461 * I / 4 + 31;
J = 80 * L / 2447;
K = L - 2447 * J / 80;
L = J / 11;
J = J + 2 - 12 * L;
I = 100 * (N - 49) + I + L;
sec = seconds % 60;
min = (seconds / 60) % 60;
hour = (seconds / 3600) % 24;
jsonenc_printf(e, "\"%04d-%02d-%02dT%02d:%02d:%02d", I, J, K, hour, min, sec);
jsonenc_nanos(e, nanos);
jsonenc_putstr(e, "Z\"");
}
static void jsonenc_duration(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* seconds_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_FieldDef* nanos_f = upb_MessageDef_FindFieldByNumber(m, 2);
int64_t seconds = upb_Message_GetFieldByDef(msg, seconds_f).int64_val;
int32_t nanos = upb_Message_GetFieldByDef(msg, nanos_f).int32_val;
bool negative = false;
if (seconds > 315576000000 || seconds < -315576000000 ||
(seconds != 0 && nanos != 0 && (seconds < 0) != (nanos < 0))) {
jsonenc_err(e, "bad duration");
}
if (seconds < 0) {
negative = true;
seconds = -seconds;
}
if (nanos < 0) {
negative = true;
nanos = -nanos;
}
jsonenc_putstr(e, "\"");
if (negative) {
jsonenc_putstr(e, "-");
}
jsonenc_printf(e, "%" PRId64, seconds);
jsonenc_nanos(e, nanos);
jsonenc_putstr(e, "s\"");
}
static void jsonenc_enum(int32_t val, const upb_FieldDef* f, jsonenc* e) {
const upb_EnumDef* e_def = upb_FieldDef_EnumSubDef(f);
if (strcmp(upb_EnumDef_FullName(e_def), "google.protobuf.NullValue") == 0) {
jsonenc_putstr(e, "null");
} else {
const upb_EnumValueDef* ev =
(e->options & upb_JsonEncode_FormatEnumsAsIntegers)
? NULL
: upb_EnumDef_FindValueByNumber(e_def, val);
if (ev) {
jsonenc_printf(e, "\"%s\"", upb_EnumValueDef_Name(ev));
} else {
jsonenc_printf(e, "%" PRId32, val);
}
}
}
static void jsonenc_bytes(jsonenc* e, upb_StringView str) {
/* This is the regular base64, not the "web-safe" version. */
static const char base64[] =
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
const unsigned char* ptr = (unsigned char*)str.data;
const unsigned char* end = UPB_PTRADD(ptr, str.size);
char buf[4];
jsonenc_putstr(e, "\"");
while (end - ptr >= 3) {
buf[0] = base64[ptr[0] >> 2];
buf[1] = base64[((ptr[0] & 0x3) << 4) | (ptr[1] >> 4)];
buf[2] = base64[((ptr[1] & 0xf) << 2) | (ptr[2] >> 6)];
buf[3] = base64[ptr[2] & 0x3f];
jsonenc_putbytes(e, buf, 4);
ptr += 3;
}
switch (end - ptr) {
case 2:
buf[0] = base64[ptr[0] >> 2];
buf[1] = base64[((ptr[0] & 0x3) << 4) | (ptr[1] >> 4)];
buf[2] = base64[(ptr[1] & 0xf) << 2];
buf[3] = '=';
jsonenc_putbytes(e, buf, 4);
break;
case 1:
buf[0] = base64[ptr[0] >> 2];
buf[1] = base64[((ptr[0] & 0x3) << 4)];
buf[2] = '=';
buf[3] = '=';
jsonenc_putbytes(e, buf, 4);
break;
}
jsonenc_putstr(e, "\"");
}
static void jsonenc_stringbody(jsonenc* e, upb_StringView str) {
const char* ptr = str.data;
const char* end = UPB_PTRADD(ptr, str.size);
while (ptr < end) {
switch (*ptr) {
case '\n':
jsonenc_putstr(e, "\\n");
break;
case '\r':
jsonenc_putstr(e, "\\r");
break;
case '\t':
jsonenc_putstr(e, "\\t");
break;
case '\"':
jsonenc_putstr(e, "\\\"");
break;
case '\f':
jsonenc_putstr(e, "\\f");
break;
case '\b':
jsonenc_putstr(e, "\\b");
break;
case '\\':
jsonenc_putstr(e, "\\\\");
break;
default:
if ((uint8_t)*ptr < 0x20) {
jsonenc_printf(e, "\\u%04x", (int)(uint8_t)*ptr);
} else {
/* This could be a non-ASCII byte. We rely on the string being valid
* UTF-8. */
jsonenc_putbytes(e, ptr, 1);
}
break;
}
ptr++;
}
}
static void jsonenc_string(jsonenc* e, upb_StringView str) {
jsonenc_putstr(e, "\"");
jsonenc_stringbody(e, str);
jsonenc_putstr(e, "\"");
}
static bool upb_JsonEncode_HandleSpecialDoubles(jsonenc* e, double val) {
if (val == INFINITY) {
jsonenc_putstr(e, "\"Infinity\"");
} else if (val == -INFINITY) {
jsonenc_putstr(e, "\"-Infinity\"");
} else if (val != val) {
jsonenc_putstr(e, "\"NaN\"");
} else {
return false;
}
return true;
}
static void upb_JsonEncode_Double(jsonenc* e, double val) {
if (upb_JsonEncode_HandleSpecialDoubles(e, val)) return;
char buf[32];
_upb_EncodeRoundTripDouble(val, buf, sizeof(buf));
jsonenc_putstr(e, buf);
}
static void upb_JsonEncode_Float(jsonenc* e, float val) {
if (upb_JsonEncode_HandleSpecialDoubles(e, val)) return;
char buf[32];
_upb_EncodeRoundTripFloat(val, buf, sizeof(buf));
jsonenc_putstr(e, buf);
}
static void jsonenc_wrapper(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* val_f = upb_MessageDef_FindFieldByNumber(m, 1);
upb_MessageValue val = upb_Message_GetFieldByDef(msg, val_f);
jsonenc_scalar(e, val, val_f);
}
static const upb_MessageDef* jsonenc_getanymsg(jsonenc* e,
upb_StringView type_url) {
/* Find last '/', if any. */
const char* end = type_url.data + type_url.size;
const char* ptr = end;
const upb_MessageDef* ret;
if (!e->ext_pool) {
jsonenc_err(e, "Tried to encode Any, but no symtab was provided");
}
if (type_url.size == 0) goto badurl;
while (true) {
if (--ptr == type_url.data) {
/* Type URL must contain at least one '/', with host before. */
goto badurl;
}
if (*ptr == '/') {
ptr++;
break;
}
}
ret = upb_DefPool_FindMessageByNameWithSize(e->ext_pool, ptr, end - ptr);
if (!ret) {
jsonenc_errf(e, "Couldn't find Any type: %.*s", (int)(end - ptr), ptr);
}
return ret;
badurl:
jsonenc_errf(e, "Bad type URL: " UPB_STRINGVIEW_FORMAT,
UPB_STRINGVIEW_ARGS(type_url));
}
static void jsonenc_any(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* type_url_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_FieldDef* value_f = upb_MessageDef_FindFieldByNumber(m, 2);
upb_StringView type_url = upb_Message_GetFieldByDef(msg, type_url_f).str_val;
upb_StringView value = upb_Message_GetFieldByDef(msg, value_f).str_val;
const upb_MessageDef* any_m = jsonenc_getanymsg(e, type_url);
const upb_MiniTable* any_layout = upb_MessageDef_MiniTable(any_m);
upb_Arena* arena = jsonenc_arena(e);
upb_Message* any = upb_Message_New(any_layout, arena);
if (upb_Decode(value.data, value.size, any, any_layout, NULL, 0, arena) !=
kUpb_DecodeStatus_Ok) {
jsonenc_err(e, "Error decoding message in Any");
}
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
jsonenc_putstr(e, "{\"@type\":");
jsonenc_string(e, type_url);
if (upb_MessageDef_WellKnownType(any_m) == kUpb_WellKnown_Unspecified) {
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
/* Regular messages: {"@type": "...","foo": 1, "bar": 2} */
jsonenc_msgfields(e, any, any_m, false);
} else {
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
/* Well-known type: {"@type": "...","value": <well-known encoding>} */
jsonenc_putstr(e, ",\"value\":");
jsonenc_msgfield(e, any, any_m);
}
jsonenc_putstr(e, "}");
}
static void jsonenc_putsep(jsonenc* e, const char* str, bool* first) {
if (*first) {
*first = false;
} else {
jsonenc_putstr(e, str);
}
}
static void jsonenc_fieldpath(jsonenc* e, upb_StringView path) {
const char* ptr = path.data;
const char* end = ptr + path.size;
while (ptr < end) {
char ch = *ptr;
if (ch >= 'A' && ch <= 'Z') {
jsonenc_err(e, "Field mask element may not have upper-case letter.");
} else if (ch == '_') {
if (ptr == end - 1 || *(ptr + 1) < 'a' || *(ptr + 1) > 'z') {
jsonenc_err(e, "Underscore must be followed by a lowercase letter.");
}
ch = *++ptr - 32;
}
jsonenc_putbytes(e, &ch, 1);
ptr++;
}
}
static void jsonenc_fieldmask(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* paths_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_Array* paths = upb_Message_GetFieldByDef(msg, paths_f).array_val;
bool first = true;
size_t i, n = 0;
if (paths) n = upb_Array_Size(paths);
jsonenc_putstr(e, "\"");
for (i = 0; i < n; i++) {
jsonenc_putsep(e, ",", &first);
jsonenc_fieldpath(e, upb_Array_Get(paths, i).str_val);
}
jsonenc_putstr(e, "\"");
}
static void jsonenc_struct(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
jsonenc_putstr(e, "{");
const upb_FieldDef* fields_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_Map* fields = upb_Message_GetFieldByDef(msg, fields_f).map_val;
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
if (fields) {
const upb_MessageDef* entry_m = upb_FieldDef_MessageSubDef(fields_f);
const upb_FieldDef* value_f = upb_MessageDef_FindFieldByNumber(entry_m, 2);
size_t iter = kUpb_Map_Begin;
bool first = true;
upb_MessageValue key, val;
while (upb_Map_Next(fields, &key, &val, &iter)) {
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
jsonenc_putsep(e, ",", &first);
jsonenc_string(e, key.str_val);
jsonenc_putstr(e, ":");
jsonenc_value(e, val.msg_val, upb_FieldDef_MessageSubDef(value_f));
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
}
}
jsonenc_putstr(e, "}");
}
static void jsonenc_listvalue(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
const upb_FieldDef* values_f = upb_MessageDef_FindFieldByNumber(m, 1);
const upb_MessageDef* values_m = upb_FieldDef_MessageSubDef(values_f);
const upb_Array* values = upb_Message_GetFieldByDef(msg, values_f).array_val;
size_t i;
bool first = true;
jsonenc_putstr(e, "[");
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
if (values) {
const size_t size = upb_Array_Size(values);
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
for (i = 0; i < size; i++) {
upb_MessageValue elem = upb_Array_Get(values, i);
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
jsonenc_putsep(e, ",", &first);
jsonenc_value(e, elem.msg_val, values_m);
}
}
jsonenc_putstr(e, "]");
}
static void jsonenc_value(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
/* TODO: do we want a reflection method to get oneof case? */
size_t iter = kUpb_Message_Begin;
const upb_FieldDef* f;
upb_MessageValue val;
if (!upb_Message_Next(msg, m, NULL, &f, &val, &iter)) {
jsonenc_err(e, "No value set in Value proto");
}
switch (upb_FieldDef_Number(f)) {
case 1:
jsonenc_putstr(e, "null");
break;
case 2:
if (upb_JsonEncode_HandleSpecialDoubles(e, val.double_val)) {
jsonenc_err(
e,
"google.protobuf.Value cannot encode double values for "
"infinity or nan, because they would be parsed as a string");
}
upb_JsonEncode_Double(e, val.double_val);
break;
case 3:
jsonenc_string(e, val.str_val);
break;
case 4:
jsonenc_putstr(e, val.bool_val ? "true" : "false");
break;
case 5:
jsonenc_struct(e, val.msg_val, upb_FieldDef_MessageSubDef(f));
break;
case 6:
jsonenc_listvalue(e, val.msg_val, upb_FieldDef_MessageSubDef(f));
break;
}
}
static void jsonenc_msgfield(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
switch (upb_MessageDef_WellKnownType(m)) {
case kUpb_WellKnown_Unspecified:
jsonenc_msg(e, msg, m);
break;
case kUpb_WellKnown_Any:
jsonenc_any(e, msg, m);
break;
case kUpb_WellKnown_FieldMask:
jsonenc_fieldmask(e, msg, m);
break;
case kUpb_WellKnown_Duration:
jsonenc_duration(e, msg, m);
break;
case kUpb_WellKnown_Timestamp:
jsonenc_timestamp(e, msg, m);
break;
case kUpb_WellKnown_DoubleValue:
case kUpb_WellKnown_FloatValue:
case kUpb_WellKnown_Int64Value:
case kUpb_WellKnown_UInt64Value:
case kUpb_WellKnown_Int32Value:
case kUpb_WellKnown_UInt32Value:
case kUpb_WellKnown_StringValue:
case kUpb_WellKnown_BytesValue:
case kUpb_WellKnown_BoolValue:
jsonenc_wrapper(e, msg, m);
break;
case kUpb_WellKnown_Value:
jsonenc_value(e, msg, m);
break;
case kUpb_WellKnown_ListValue:
jsonenc_listvalue(e, msg, m);
break;
case kUpb_WellKnown_Struct:
jsonenc_struct(e, msg, m);
break;
}
}
static void jsonenc_scalar(jsonenc* e, upb_MessageValue val,
const upb_FieldDef* f) {
switch (upb_FieldDef_CType(f)) {
case kUpb_CType_Bool:
jsonenc_putstr(e, val.bool_val ? "true" : "false");
break;
case kUpb_CType_Float:
upb_JsonEncode_Float(e, val.float_val);
break;
case kUpb_CType_Double:
upb_JsonEncode_Double(e, val.double_val);
break;
case kUpb_CType_Int32:
jsonenc_printf(e, "%" PRId32, val.int32_val);
break;
case kUpb_CType_UInt32:
jsonenc_printf(e, "%" PRIu32, val.uint32_val);
break;
case kUpb_CType_Int64:
jsonenc_printf(e, "\"%" PRId64 "\"", val.int64_val);
break;
case kUpb_CType_UInt64:
jsonenc_printf(e, "\"%" PRIu64 "\"", val.uint64_val);
break;
case kUpb_CType_String:
jsonenc_string(e, val.str_val);
break;
case kUpb_CType_Bytes:
jsonenc_bytes(e, val.str_val);
break;
case kUpb_CType_Enum:
jsonenc_enum(val.int32_val, f, e);
break;
case kUpb_CType_Message:
jsonenc_msgfield(e, val.msg_val, upb_FieldDef_MessageSubDef(f));
break;
}
}
static void jsonenc_mapkey(jsonenc* e, upb_MessageValue val,
const upb_FieldDef* f) {
jsonenc_putstr(e, "\"");
switch (upb_FieldDef_CType(f)) {
case kUpb_CType_Bool:
jsonenc_putstr(e, val.bool_val ? "true" : "false");
break;
case kUpb_CType_Int32:
jsonenc_printf(e, "%" PRId32, val.int32_val);
break;
case kUpb_CType_UInt32:
jsonenc_printf(e, "%" PRIu32, val.uint32_val);
break;
case kUpb_CType_Int64:
jsonenc_printf(e, "%" PRId64, val.int64_val);
break;
case kUpb_CType_UInt64:
jsonenc_printf(e, "%" PRIu64, val.uint64_val);
break;
case kUpb_CType_String:
jsonenc_stringbody(e, val.str_val);
break;
default:
UPB_UNREACHABLE();
}
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
jsonenc_putstr(e, "\":");
}
static void jsonenc_array(jsonenc* e, const upb_Array* arr,
const upb_FieldDef* f) {
size_t i;
size_t size = arr ? upb_Array_Size(arr) : 0;
bool first = true;
jsonenc_putstr(e, "[");
for (i = 0; i < size; i++) {
Fixes for PHP. (#286) - A new PHP-specific upb amalgamation. It contains everything related to upb_msg, but leaves out all of the old handlers-related interfaces and encoders/decoders. # Schema/Defs Changes - Changed `upb_fielddef_msgsubdef()` and `upb_fielddef_enumsubdef()` to return `NULL` instead of assert-failing if the field is not a message or enum. - Added `upb_msgdef_iswrapper()`, to test whether this is a wrapper well-known type. # Decoder - Decoder bugfix: when we parse a submessage inside a oneof, we need to clear out any previous data, so we don't misinterpret it as a pointer to an existing submessage. # JSON Decoder - Allowed well-known types at the top level to have their special processing. - Fixed a bug that could occur when parsing nested empty lists/objects, eg `[[]]`. - Made the "ignore unknown" option also be permissive about unknown enumerators by setting them to 0. # JSON Encoder - Allowed well-known types at the top level to have their special processing. - Removed all spaces after `:` and `,` characters, to match the old encoder and pass goldenfile tests. # Message / Reflection - Changed `upb_msg_hasoneof()` -> `upb_msg_whichoneof()`. The new function returns the `upb_fielddef*` of whichever oneof is set. - Implemented `upb_msg_clearfield()` and added/implemented `upb_msg_clear()`. - Added `upb_msg_discardunknown()`. Part of me thinks this should go in a util library instead of core reflection since it is a recursive algorithm. # Compiler - Always emit descriptors as an array instead of as a string, to avoid exceeding maximum string lengths. If this becomes a speed issue later we can go back to two separate paths.
5 years ago
jsonenc_putsep(e, ",", &first);
jsonenc_scalar(e, upb_Array_Get(arr, i), f);
}
jsonenc_putstr(e, "]");
}
static void jsonenc_map(jsonenc* e, const upb_Map* map, const upb_FieldDef* f) {
jsonenc_putstr(e, "{");
const upb_MessageDef* entry = upb_FieldDef_MessageSubDef(f);
const upb_FieldDef* key_f = upb_MessageDef_FindFieldByNumber(entry, 1);
const upb_FieldDef* val_f = upb_MessageDef_FindFieldByNumber(entry, 2);
if (map) {
size_t iter = kUpb_Map_Begin;
bool first = true;
upb_MessageValue key, val;
while (upb_Map_Next(map, &key, &val, &iter)) {
jsonenc_putsep(e, ",", &first);
jsonenc_mapkey(e, key, key_f);
jsonenc_scalar(e, val, val_f);
}
}
jsonenc_putstr(e, "}");
}
static void jsonenc_fieldval(jsonenc* e, const upb_FieldDef* f,
upb_MessageValue val, bool* first) {
const char* name;
jsonenc_putsep(e, ",", first);
if (upb_FieldDef_IsExtension(f)) {
// TODO: For MessageSet, I would have expected this to print the message
// name here, but Python doesn't appear to do this. We should do more
// research here about what various implementations do.
jsonenc_printf(e, "\"[%s]\":", upb_FieldDef_FullName(f));
} else {
if (e->options & upb_JsonEncode_UseProtoNames) {
name = upb_FieldDef_Name(f);
} else {
name = upb_FieldDef_JsonName(f);
}
jsonenc_printf(e, "\"%s\":", name);
}
if (upb_FieldDef_IsMap(f)) {
jsonenc_map(e, val.map_val, f);
} else if (upb_FieldDef_IsRepeated(f)) {
jsonenc_array(e, val.array_val, f);
} else {
jsonenc_scalar(e, val, f);
}
}
static void jsonenc_msgfields(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m, bool first) {
upb_MessageValue val;
const upb_FieldDef* f;
if (e->options & upb_JsonEncode_EmitDefaults) {
/* Iterate over all fields. */
int i = 0;
int n = upb_MessageDef_FieldCount(m);
for (i = 0; i < n; i++) {
f = upb_MessageDef_Field(m, i);
if (!upb_FieldDef_HasPresence(f) || upb_Message_HasFieldByDef(msg, f)) {
jsonenc_fieldval(e, f, upb_Message_GetFieldByDef(msg, f), &first);
}
}
} else {
/* Iterate over non-empty fields. */
size_t iter = kUpb_Message_Begin;
while (upb_Message_Next(msg, m, e->ext_pool, &f, &val, &iter)) {
jsonenc_fieldval(e, f, val, &first);
}
}
}
static void jsonenc_msg(jsonenc* e, const upb_Message* msg,
const upb_MessageDef* m) {
jsonenc_putstr(e, "{");
jsonenc_msgfields(e, msg, m, true);
jsonenc_putstr(e, "}");
}
static size_t jsonenc_nullz(jsonenc* e, size_t size) {
size_t ret = e->ptr - e->buf + e->overflow;
if (size > 0) {
if (e->ptr == e->end) e->ptr--;
*e->ptr = '\0';
}
return ret;
}
static size_t upb_JsonEncoder_Encode(jsonenc* const e,
const upb_Message* const msg,
const upb_MessageDef* const m,
const size_t size) {
if (UPB_SETJMP(e->err) != 0) return -1;
jsonenc_msgfield(e, msg, m);
if (e->arena) upb_Arena_Free(e->arena);
return jsonenc_nullz(e, size);
}
size_t upb_JsonEncode(const upb_Message* msg, const upb_MessageDef* m,
const upb_DefPool* ext_pool, int options, char* buf,
size_t size, upb_Status* status) {
jsonenc e;
e.buf = buf;
e.ptr = buf;
e.end = UPB_PTRADD(buf, size);
e.overflow = 0;
e.options = options;
e.ext_pool = ext_pool;
e.status = status;
e.arena = NULL;
return upb_JsonEncoder_Encode(&e, msg, m, size);
}