Squashed 'third_party/upb/' changes from 97bcd5276c..4e2505edaa

4e2505edaa Merge pull request #259 from haberman/rmsubmodule
91fd76cc9c Merge pull request #258 from haberman/bazel_version
d0d5339620 Removed .gitmodules, we don't use submodules any more.
6c5d5afc43 Remove compatibility code for old Bazel versions.
c1357afb2e Merge pull request #230 from moroten/maybe-deps
a3d693544b Merge pull request #257 from haberman/warnings2
63e673383b Fixed narrowing warnings in text_encode.c.
39bc93a527 Merge pull request #252 from haberman/jsondecode
0fdd65f223 Merge branch 'master' into jsondecode
82af3d661c Merge pull request #255 from protocolbuffers/const-array
90b3a20af0 Making sure _upb_fieldtype_to_sizelg2 is readonly
5667a7a806 Removed stray assert, causing C90 build errors.
23a5af3513 [json] fixed all remaining conformance bugs.
6ec4df82c1 Updated other parts of upb for new JSON name API.
a292261aeb Added JSON decoder to conformance tests, and fixed tons of bugs.
d49c1db6c2 New JSON decoder, string->msg using reflection.
f593289087 Merge branch 'master' into maybe-deps
0c2046f732 Make workspace_deps.bzl overridable using maybe()

git-subtree-dir: third_party/upb
git-subtree-split: 4e2505edaa325bcf0ffd69eefa7cb810d6dde861

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "third_party/protobuf"]
-	path = third_party/protobuf
-	url = https://github.com/google/protobuf.git

BUILD

@@ -162,9 +162,11 @@ cc_library(
 cc_library(
     name = "json",
     srcs = [
+        "upb/json_decode.c",
         "upb/json_encode.c",
     ],
     hdrs = [
+        "upb/json_decode.h",
         "upb/json_encode.h",
     ],
     deps = [

CMakeLists.txt

@@ -97,7 +97,9 @@ target_link_libraries(textformat
   port
   reflection)
 add_library(json
+  upb/json_decode.c
   upb/json_encode.c
+  upb/json_decode.h
   upb/json_encode.h)
 target_link_libraries(json
   port

bazel/repository_defs.bzl

@@ -1,15 +0,0 @@
-# A hacky way to work around the fact that native.bazel_version is only
-# available from WORKSPACE macros, not BUILD macros or rules.
-#
-# Hopefully we can remove this if/when this is fixed:
-#   https://github.com/bazelbuild/bazel/issues/8305
-
-def _impl(repository_ctx):
-    s = "bazel_version = \"" + native.bazel_version + "\""
-    repository_ctx.file("bazel_version.bzl", s)
-    repository_ctx.file("BUILD", "")
-
-bazel_version_repository = repository_rule(
-    implementation = _impl,
-    local = True,
-)

bazel/upb_proto_library.bzl

@@ -6,11 +6,6 @@
 load("@bazel_skylib//lib:paths.bzl", "paths")
 load("@bazel_tools//tools/cpp:toolchain_utils.bzl", "find_cpp_toolchain")
 
-# copybara:strip_for_google3_begin
-load("@bazel_skylib//lib:versions.bzl", "versions")
-load("@upb_bazel_version//:bazel_version.bzl", "bazel_version")
-# copybara:strip_end
-
 # Generic support code #########################################################
 
 _is_bazel = not hasattr(native, "genmpm")
@@ -79,34 +74,6 @@ def _cc_library_func(ctx, name, hdrs, srcs, dep_ccinfos):
         unsupported_features = ctx.disabled_features,
     )
 
-    # copybara:strip_for_google3_begin
-    if bazel_version == "0.24.1":
-        # Compatibility code until gRPC is on 0.25.2 or later.
-        compilation_info = cc_common.compile(
-            ctx = ctx,
-            feature_configuration = feature_configuration,
-            cc_toolchain = toolchain,
-            srcs = srcs,
-            hdrs = hdrs,
-            compilation_contexts = compilation_contexts,
-        )
-        linking_info = cc_common.link(
-            ctx = ctx,
-            feature_configuration = feature_configuration,
-            cc_toolchain = toolchain,
-            cc_compilation_outputs = compilation_info.cc_compilation_outputs,
-            linking_contexts = linking_contexts,
-        )
-        return CcInfo(
-            compilation_context = compilation_info.compilation_context,
-            linking_context = linking_info.linking_context,
-        )
-
-    if not versions.is_at_least("0.25.2", bazel_version):
-        fail("upb requires Bazel >=0.25.2 or 0.24.1")
-
-    # copybara:strip_end
-
     blaze_only_args = {}
 
     if not _is_bazel:

bazel/workspace_deps.bzl

@@ -1,33 +1,33 @@
 
 load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
 load("@bazel_tools//tools/build_defs/repo:git.bzl", "git_repository")
-load("//bazel:repository_defs.bzl", "bazel_version_repository")
+load("@bazel_tools//tools/build_defs/repo:utils.bzl", "maybe")
 
 def upb_deps():
-    bazel_version_repository(
-        name = "upb_bazel_version",
-    )
-
-    git_repository(
+    maybe(
+        git_repository,
         name = "com_google_absl",
         commit = "070f6e47b33a2909d039e620c873204f78809492",
         remote = "https://github.com/abseil/abseil-cpp.git",
         shallow_since = "1541627663 -0500",
     )
 
-    git_repository(
+    maybe(
+        git_repository,
         name = "com_google_protobuf",
         remote = "https://github.com/protocolbuffers/protobuf.git",
         commit = "d41002663fd04325ead28439dfd5ce2822b0d6fb",
     )
 
-    http_archive(
+    maybe(
+        http_archive,
         name = "bazel_skylib",
         strip_prefix = "bazel-skylib-master",
         urls = ["https://github.com/bazelbuild/bazel-skylib/archive/master.tar.gz"],
     )
 
-    http_archive(
+    maybe(
+        http_archive,
         name = "zlib",
         build_file = "@com_google_protobuf//:third_party/zlib.BUILD",
         sha256 = "629380c90a77b964d896ed37163f5c3a34f6e6d897311f1df2a7016355c45eff",

generated_for_cmake/upb/json/parser.c

@@ -3311,15 +3311,13 @@ static upb_json_parsermethod *parsermethod_new(upb_json_codecache *c,
       upb_msg_field_next(&i)) {
     const upb_fielddef *f = upb_msg_iter_field(&i);
     upb_value v = upb_value_constptr(f);
-    char *buf;
+    const char *name;
 
     /* Add an entry for the JSON name. */
-    size_t len = upb_fielddef_getjsonname(f, NULL, 0);
-    buf = upb_malloc(alloc, len);
-    upb_fielddef_getjsonname(f, buf, len);
-    upb_strtable_insert3(&m->name_table, buf, strlen(buf), v, alloc);
+    name = upb_fielddef_jsonname(f);
+    upb_strtable_insert3(&m->name_table, name, strlen(name), v, alloc);
 
-    if (strcmp(buf, upb_fielddef_name(f)) != 0) {
+    if (strcmp(name, upb_fielddef_name(f)) != 0) {
       /* Since the JSON name is different from the regular field name, add an
        * entry for the raw name (compliant proto3 JSON parsers must accept
        * both). */

tests/conformance_upb.c

@@ -15,6 +15,7 @@
 #include "upb/decode.h"
 #include "upb/encode.h"
 #include "upb/reflection.h"
+#include "upb/json_decode.h"
 #include "upb/json_encode.h"
 #include "upb/text_encode.h"
 
@@ -85,9 +86,11 @@ void serialize_text(const upb_msg *msg, const upb_msgdef *m, const ctx *c) {
   size_t len2;
   int opts = 0;
   char *data;
+
   if (!conformance_ConformanceRequest_print_unknown_fields(c->request)) {
     opts |= UPB_TXTENC_SKIPUNKNOWN;
   }
+
   len = upb_text_encode(msg, m, c->symtab, opts, NULL, 0);
   data = upb_arena_malloc(c->arena, len + 1);
   len2 = upb_text_encode(msg, m, c->symtab, opts, data, len + 1);
@@ -96,6 +99,33 @@ void serialize_text(const upb_msg *msg, const upb_msgdef *m, const ctx *c) {
       c->response, upb_strview_make(data, len));
 }
 
+bool parse_json(upb_msg *msg, const upb_msgdef *m, const ctx* c) {
+  upb_strview json =
+      conformance_ConformanceRequest_json_payload(c->request);
+  upb_status status;
+  int opts = 0;
+
+  if (conformance_ConformanceRequest_test_category(c->request) ==
+      conformance_JSON_IGNORE_UNKNOWN_PARSING_TEST) {
+    opts |= UPB_JSONDEC_IGNOREUNKNOWN;
+  }
+
+  upb_status_clear(&status);
+  if (upb_json_decode(json.data, json.size, msg, m, c->symtab, opts, c->arena,
+                      &status)) {
+    return true;
+  } else {
+    const char *inerr = upb_status_errmsg(&status);
+    size_t len = strlen(inerr);
+    char *err = upb_arena_malloc(c->arena, len + 1);
+    memcpy(err, inerr, strlen(inerr));
+    err[len] = '\0';
+    conformance_ConformanceResponse_set_parse_error(c->response,
+                                                    upb_strview_makez(err));
+    return false;
+  }
+}
+
 void serialize_json(const upb_msg *msg, const upb_msgdef *m, const ctx *c) {
   size_t len;
   size_t len2;
@@ -104,16 +134,16 @@ void serialize_json(const upb_msg *msg, const upb_msgdef *m, const ctx *c) {
   upb_status status;
 
   upb_status_clear(&status);
-  if (!conformance_ConformanceRequest_print_unknown_fields(c->request)) {
-    opts |= UPB_TXTENC_SKIPUNKNOWN;
-  }
-
   len = upb_json_encode(msg, m, c->symtab, opts, NULL, 0, &status);
 
   if (len == -1) {
-    static const char msg[] = "Error serializing.";
-    conformance_ConformanceResponse_set_serialize_error(
-        c->response, upb_strview_make(msg, strlen(msg)));
+    const char *inerr = upb_status_errmsg(&status);
+    size_t len = strlen(inerr);
+    char *err = upb_arena_malloc(c->arena, len + 1);
+    memcpy(err, inerr, strlen(inerr));
+    err[len] = '\0';
+    conformance_ConformanceResponse_set_serialize_error(c->response,
+                                                        upb_strview_makez(err));
     return;
   }
 
@@ -128,6 +158,8 @@ bool parse_input(upb_msg *msg, const upb_msgdef *m, const ctx* c) {
   switch (conformance_ConformanceRequest_payload_case(c->request)) {
     case conformance_ConformanceRequest_payload_protobuf_payload:
       return parse_proto(msg, m, c);
+    case conformance_ConformanceRequest_payload_json_payload:
+      return parse_json(msg, m, c);
     case conformance_ConformanceRequest_payload_NOT_SET:
       fprintf(stderr, "conformance_upb: Request didn't have payload.\n");
       return false;

upb/def.c

@@ -27,6 +27,7 @@ struct upb_fielddef {
   const upb_filedef *file;
   const upb_msgdef *msgdef;
   const char *full_name;
+  const char *json_name;
   union {
     int64_t sint;
     uint64_t uint;
@@ -117,10 +118,15 @@ struct upb_symtab {
 
 /* Inside a symtab we store tagged pointers to specific def types. */
 typedef enum {
-  UPB_DEFTYPE_MSG = 0,
-  UPB_DEFTYPE_ENUM = 1,
-  UPB_DEFTYPE_FIELD = 2,
-  UPB_DEFTYPE_ONEOF = 3
+  UPB_DEFTYPE_FIELD = 0,
+
+  /* Only inside symtab table. */
+  UPB_DEFTYPE_MSG = 1,
+  UPB_DEFTYPE_ENUM = 2,
+
+  /* Only inside message table. */
+  UPB_DEFTYPE_ONEOF = 1,
+  UPB_DEFTYPE_FIELD_JSONNAME = 2
 } upb_deftype_t;
 
 static const void *unpack_def(upb_value v, upb_deftype_t type) {
@@ -462,47 +468,12 @@ const char *upb_fielddef_name(const upb_fielddef *f) {
   return shortdefname(f->full_name);
 }
 
-uint32_t upb_fielddef_selectorbase(const upb_fielddef *f) {
-  return f->selector_base;
+const char *upb_fielddef_jsonname(const upb_fielddef *f) {
+  return f->json_name;
 }
 
-size_t upb_fielddef_getjsonname(const upb_fielddef *f, char *buf, size_t len) {
-  const char *name = upb_fielddef_name(f);
-  size_t src, dst = 0;
-  bool ucase_next = false;
-
-#define WRITE(byte) \
-  ++dst; \
-  if (dst < len) buf[dst - 1] = byte; \
-  else if (dst == len) buf[dst - 1] = '\0'
-
-  if (!name) {
-    WRITE('\0');
-    return 0;
-  }
-
-  /* Implement the transformation as described in the spec:
-   *   1. upper case all letters after an underscore.
-   *   2. remove all underscores.
-   */
-  for (src = 0; name[src]; src++) {
-    if (name[src] == '_') {
-      ucase_next = true;
-      continue;
-    }
-
-    if (ucase_next) {
-      WRITE(toupper(name[src]));
-      ucase_next = false;
-    } else {
-      WRITE(name[src]);
-    }
-  }
-
-  WRITE('\0');
-  return dst;
-
-#undef WRITE
+uint32_t upb_fielddef_selectorbase(const upb_fielddef *f) {
+  return f->selector_base;
 }
 
 const upb_msgdef *upb_fielddef_containingtype(const upb_fielddef *f) {
@@ -690,18 +661,30 @@ bool upb_msgdef_lookupname(const upb_msgdef *m, const char *name, size_t len,
 
   *o = unpack_def(val, UPB_DEFTYPE_ONEOF);
   *f = unpack_def(val, UPB_DEFTYPE_FIELD);
-  UPB_ASSERT((*o != NULL) ^ (*f != NULL));  /* Exactly one of the two should be set. */
-  return true;
+  return *o || *f;  /* False if this was a JSON name. */
+}
+
+const upb_fielddef *upb_msgdef_lookupjsonname(const upb_msgdef *m,
+                                              const char *name, size_t len) {
+  upb_value val;
+  const upb_fielddef* f;
+
+  if (!upb_strtable_lookup2(&m->ntof, name, len, &val)) {
+    return NULL;
+  }
+
+  f = unpack_def(val, UPB_DEFTYPE_FIELD);
+  if (!f) f = unpack_def(val, UPB_DEFTYPE_FIELD_JSONNAME);
+
+  return f;
 }
 
 int upb_msgdef_numfields(const upb_msgdef *m) {
-  /* The number table contains only fields. */
-  return (int)upb_inttable_count(&m->itof);
+  return m->field_count;
 }
 
 int upb_msgdef_numoneofs(const upb_msgdef *m) {
-  /* The name table includes oneofs, and the number table does not. */
-  return (int)(upb_strtable_count(&m->ntof) - upb_inttable_count(&m->itof));
+  return m->oneof_count;
 }
 
 const upb_msglayout *upb_msgdef_layout(const upb_msgdef *m) {
@@ -1098,6 +1081,51 @@ static const char *makefullname(const symtab_addctx *ctx, const char *prefix,
   }
 }
 
+size_t getjsonname(const char *name, char *buf, size_t len) {
+  size_t src, dst = 0;
+  bool ucase_next = false;
+
+#define WRITE(byte) \
+  ++dst; \
+  if (dst < len) buf[dst - 1] = byte; \
+  else if (dst == len) buf[dst - 1] = '\0'
+
+  if (!name) {
+    WRITE('\0');
+    return 0;
+  }
+
+  /* Implement the transformation as described in the spec:
+   *   1. upper case all letters after an underscore.
+   *   2. remove all underscores.
+   */
+  for (src = 0; name[src]; src++) {
+    if (name[src] == '_') {
+      ucase_next = true;
+      continue;
+    }
+
+    if (ucase_next) {
+      WRITE(toupper(name[src]));
+      ucase_next = false;
+    } else {
+      WRITE(name[src]);
+    }
+  }
+
+  WRITE('\0');
+  return dst;
+
+#undef WRITE
+}
+
+static char* makejsonname(const char* name, upb_alloc *alloc) {
+  size_t size = getjsonname(name, NULL, 0);
+  char* json_name = upb_malloc(alloc, size);
+  getjsonname(name, json_name, size);
+  return json_name;
+}
+
 static bool symtab_add(const symtab_addctx *ctx, const char *name,
                        upb_value v) {
   upb_value tmp;
@@ -1311,6 +1339,7 @@ static bool create_fielddef(
   const google_protobuf_FieldOptions *options;
   upb_strview name;
   const char *full_name;
+  const char *json_name;
   const char *shortname;
   uint32_t field_number;
 
@@ -1324,6 +1353,13 @@ static bool create_fielddef(
   full_name = makefullname(ctx, prefix, name);
   shortname = shortdefname(full_name);
 
+  if (google_protobuf_FieldDescriptorProto_has_json_name(field_proto)) {
+    json_name = strviewdup(
+        ctx, google_protobuf_FieldDescriptorProto_json_name(field_proto));
+  } else {
+    json_name = makejsonname(shortname, ctx->alloc);
+  }
+
   field_number = google_protobuf_FieldDescriptorProto_number(field_proto);
 
   if (field_number == 0 || field_number > UPB_MAX_FIELDNUMBER) {
@@ -1333,26 +1369,42 @@ static bool create_fielddef(
 
   if (m) {
     /* direct message field. */
-    upb_value v, packed_v;
+    upb_value v, field_v, json_v;
+    size_t json_size;
 
     f = (upb_fielddef*)&m->fields[m->field_count++];
     f->msgdef = m;
     f->is_extension_ = false;
 
-    packed_v = pack_def(f, UPB_DEFTYPE_FIELD);
-    v = upb_value_constptr(f);
-
-    if (!upb_strtable_insert3(&m->ntof, name.data, name.size, packed_v, alloc)) {
+    if (upb_strtable_lookup(&m->ntof, shortname, NULL)) {
       upb_status_seterrf(ctx->status, "duplicate field name (%s)", shortname);
       return false;
     }
 
-    if (!upb_inttable_insert2(&m->itof, field_number, v, alloc)) {
+    if (upb_strtable_lookup(&m->ntof, json_name, NULL)) {
+      upb_status_seterrf(ctx->status, "duplicate json_name (%s)", json_name);
+      return false;
+    }
+
+    if (upb_inttable_lookup(&m->itof, field_number, NULL)) {
       upb_status_seterrf(ctx->status, "duplicate field number (%u)",
                          field_number);
       return false;
     }
 
+    field_v = pack_def(f, UPB_DEFTYPE_FIELD);
+    json_v = pack_def(f, UPB_DEFTYPE_FIELD_JSONNAME);
+    v = upb_value_constptr(f);
+    json_size = strlen(json_name);
+
+    CHK_OOM(
+        upb_strtable_insert3(&m->ntof, name.data, name.size, field_v, alloc));
+    CHK_OOM(upb_inttable_insert2(&m->itof, field_number, v, alloc));
+
+    if (strcmp(shortname, json_name) != 0) {
+      upb_strtable_insert3(&m->ntof, json_name, json_size, json_v, alloc);
+    }
+
     if (ctx->layouts) {
       const upb_msglayout_field *fields = m->layout->fields;
       int count = m->layout->field_count;
@@ -1369,12 +1421,13 @@ static bool create_fielddef(
     }
   } else {
     /* extension field. */
-    f = (upb_fielddef*)&ctx->file->exts[ctx->file->ext_count];
+    f = (upb_fielddef*)&ctx->file->exts[ctx->file->ext_count++];
     f->is_extension_ = true;
     CHK_OOM(symtab_add(ctx, full_name, pack_def(f, UPB_DEFTYPE_FIELD)));
   }
 
   f->full_name = full_name;
+  f->json_name = json_name;
   f->file = ctx->file;
   f->type_ = (int)google_protobuf_FieldDescriptorProto_type(field_proto);
   f->label_ = (int)google_protobuf_FieldDescriptorProto_label(field_proto);
@@ -1741,7 +1794,8 @@ static bool build_filedef(
     } else if (streql_view(syntax, "proto3")) {
       file->syntax = UPB_SYNTAX_PROTO3;
     } else {
-      upb_status_seterrf(ctx->status, "Invalid syntax '%s'", syntax);
+      upb_status_seterrf(ctx->status, "Invalid syntax '" UPB_STRVIEW_FORMAT "'",
+                         UPB_STRVIEW_ARGS(syntax));
       return false;
     }
   } else {

upb/def.h

@@ -99,10 +99,10 @@ upb_descriptortype_t upb_fielddef_descriptortype(const upb_fielddef *f);
 upb_label_t upb_fielddef_label(const upb_fielddef *f);
 uint32_t upb_fielddef_number(const upb_fielddef *f);
 const char *upb_fielddef_name(const upb_fielddef *f);
+const char *upb_fielddef_jsonname(const upb_fielddef *f);
 bool upb_fielddef_isextension(const upb_fielddef *f);
 bool upb_fielddef_lazy(const upb_fielddef *f);
 bool upb_fielddef_packed(const upb_fielddef *f);
-size_t upb_fielddef_getjsonname(const upb_fielddef *f, char *buf, size_t len);
 const upb_msgdef *upb_fielddef_containingtype(const upb_fielddef *f);
 const upb_oneofdef *upb_fielddef_containingoneof(const upb_fielddef *f);
 uint32_t upb_fielddef_index(const upb_fielddef *f);
@@ -151,32 +151,10 @@ class upb::FieldDefPtr {
   Type type() const { return upb_fielddef_type(ptr_); }
   Label label() const { return upb_fielddef_label(ptr_); }
   const char* name() const { return upb_fielddef_name(ptr_); }
+  const char* json_name() const { return upb_fielddef_jsonname(ptr_); }
   uint32_t number() const { return upb_fielddef_number(ptr_); }
   bool is_extension() const { return upb_fielddef_isextension(ptr_); }
 
-  /* Copies the JSON name for this field into the given buffer.  Returns the
-   * actual size of the JSON name, including the NULL terminator.  If the
-   * return value is 0, the JSON name is unset.  If the return value is
-   * greater than len, the JSON name was truncated.  The buffer is always
-   * NULL-terminated if len > 0.
-   *
-   * The JSON name always defaults to a camelCased version of the regular
-   * name.  However if the regular name is unset, the JSON name will be unset
-   * also.
-   */
-  size_t GetJsonName(char *buf, size_t len) const {
-    return upb_fielddef_getjsonname(ptr_, buf, len);
-  }
-
-  /* Convenience version of the above function which copies the JSON name
-   * into the given string, returning false if the name is not set. */
-  template <class T>
-  bool GetJsonName(T* str) {
-    str->resize(GetJsonName(NULL, 0));
-    GetJsonName(&(*str)[0], str->size());
-    return str->size() > 0;
-  }
-
   /* For UPB_TYPE_MESSAGE fields only where is_tag_delimited() == false,
    * indicates whether this field should have lazy parsing handlers that yield
    * the unparsed string for the submessage.
@@ -455,6 +433,10 @@ UPB_INLINE bool upb_msgdef_lookupnamez(const upb_msgdef *m, const char *name,
   return upb_msgdef_lookupname(m, name, strlen(name), f, o);
 }
 
+/* Returns a field by either JSON name or regular proto name. */
+const upb_fielddef *upb_msgdef_lookupjsonname(const upb_msgdef *m,
+                                              const char *name, size_t len);
+
 /* Iteration over fields and oneofs.  For example:
  *
  * upb_msg_field_iter i;

upb/json/parser.rl

@@ -2874,15 +2874,13 @@ static upb_json_parsermethod *parsermethod_new(upb_json_codecache *c,
       upb_msg_field_next(&i)) {
     const upb_fielddef *f = upb_msg_iter_field(&i);
     upb_value v = upb_value_constptr(f);
-    char *buf;
+    const char *name;
 
     /* Add an entry for the JSON name. */
-    size_t len = upb_fielddef_getjsonname(f, NULL, 0);
-    buf = upb_malloc(alloc, len);
-    upb_fielddef_getjsonname(f, buf, len);
-    upb_strtable_insert3(&m->name_table, buf, strlen(buf), v, alloc);
+    name = upb_fielddef_jsonname(f);
+    upb_strtable_insert3(&m->name_table, name, strlen(name), v, alloc);
 
-    if (strcmp(buf, upb_fielddef_name(f)) != 0) {
+    if (strcmp(name, upb_fielddef_name(f)) != 0) {
       /* Since the JSON name is different from the regular field name, add an
        * entry for the raw name (compliant proto3 JSON parsers must accept
        * both). */

upb/json/printer.c

@@ -65,12 +65,8 @@ strpc *newstrpc(upb_handlers *h, const upb_fielddef *f,
     ret->ptr = upb_gstrdup(upb_fielddef_name(f));
     ret->len = strlen(ret->ptr);
   } else {
-    size_t len;
-    ret->len = upb_fielddef_getjsonname(f, NULL, 0);
-    ret->ptr = upb_gmalloc(ret->len);
-    len = upb_fielddef_getjsonname(f, ret->ptr, ret->len);
-    UPB_ASSERT(len == ret->len);
-    ret->len--;  /* NULL */
+    ret->ptr = upb_gstrdup(upb_fielddef_jsonname(f));
+    ret->len = strlen(ret->ptr);
   }
 
   upb_handlers_addcleanup(h, ret, freestrpc);

upb/json_decode.h

@@ -0,0 +1,24 @@
+
+#ifndef UPB_JSONDECODE_H_
+#define UPB_JSONDECODE_H_
+
+#include "upb/def.h"
+#include "upb/msg.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+enum {
+  UPB_JSONDEC_IGNOREUNKNOWN = 1
+};
+
+bool upb_json_decode(const char *buf, size_t size, upb_msg *msg,
+                     const upb_msgdef *m, const upb_symtab *any_pool,
+                     int options, upb_arena *arena, upb_status *status);
+
+#ifdef __cplusplus
+}  /* extern "C" */
+#endif
+
+#endif  /* UPB_JSONDECODE_H_ */

upb/json_encode.c

@@ -29,13 +29,23 @@ static void jsonenc_msg(jsonenc *e, const upb_msg *msg, const upb_msgdef *m);
 static void jsonenc_scalar(jsonenc *e, upb_msgval val, const upb_fielddef *f);
 static void jsonenc_msgfield(jsonenc *e, const upb_msg *msg,
                              const upb_msgdef *m);
+static void jsonenc_msgfields(jsonenc *e, const upb_msg *msg,
+                              const upb_msgdef *m);
 static void jsonenc_value(jsonenc *e, const upb_msg *msg, const upb_msgdef *m);
 
-static void jsonenc_err(jsonenc *e, const char *msg) {
+UPB_NORETURN static void jsonenc_err(jsonenc *e, const char *msg) {
   upb_status_seterrmsg(e->status, msg);
   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)) {
@@ -70,19 +80,19 @@ static void jsonenc_printf(jsonenc *e, const char *fmt, ...) {
 }
 
 static void jsonenc_nanos(jsonenc *e, int32_t nanos) {
-  const char zeros[3] = "000";
+  int digits = 9;
 
   if (nanos == 0) return;
   if (nanos < 0 || nanos >= 1000000000) {
     jsonenc_err(e, "error formatting timestamp as JSON: invalid nanos");
   }
 
-  jsonenc_printf(e, "%09" PRId32, nanos);
-
-  /* Remove trailing zeros, 3 at a time. */
-  while ((e->ptr - e->buf) >= 3 && memcmp(e->ptr, zeros, 3) == 0) {
-    e->ptr -= 3;
+  while (nanos % 1000 == 0) {
+    nanos /= 1000;
+    digits -= 3;
   }
+
+  jsonenc_printf(e, ".%0.*" PRId32, digits, nanos);
 }
 
 static void jsonenc_timestamp(jsonenc *e, const upb_msg *msg,
@@ -107,7 +117,7 @@ static void jsonenc_timestamp(jsonenc *e, const upb_msg *msg,
    * 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.  */
-  L = (seconds / 86400) + 2440588;
+  L = (seconds / 86400) + 68569 + 2440588;
   N = 4 * L / 146097;
   L = L - (146097 * N + 3) / 4;
   I = 4000 * (L + 1) / 1461001;
@@ -138,6 +148,10 @@ static void jsonenc_duration(jsonenc *e, const upb_msg *msg, const upb_msgdef *m
     jsonenc_err(e, "bad duration");
   }
 
+  if (nanos < 0) {
+    nanos = -nanos;
+  }
+
   jsonenc_printf(e, "\"%" PRId64, seconds);
   jsonenc_nanos(e, nanos);
   jsonenc_putstr(e, "s\"");
@@ -158,8 +172,8 @@ static void jsonenc_bytes(jsonenc *e, upb_strview str) {
   /* This is the regular base64, not the "web-safe" version. */
   static const char base64[] =
       "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
-  const char *ptr = str.data;
-  const char *end = ptr + str.size;
+  const unsigned char *ptr = (unsigned char*)str.data;
+  const unsigned char *end = ptr + str.size;
   char buf[4];
 
   jsonenc_putstr(e, "\"");
@@ -212,10 +226,10 @@ static void jsonenc_stringbody(jsonenc *e, upb_strview str) {
         jsonenc_putstr(e, "\\\"");
         break;
       case '\f':
-        jsonenc_putstr(e, "\f'");
+        jsonenc_putstr(e, "\\f");
         break;
       case '\b':
-        jsonenc_putstr(e, "\b'");
+        jsonenc_putstr(e, "\\b");
         break;
       case '\\':
         jsonenc_putstr(e, "\\\\");
@@ -255,21 +269,22 @@ static void jsonenc_double(jsonenc *e, const char *fmt, double val) {
 static void jsonenc_wrapper(jsonenc *e, const upb_msg *msg,
                             const upb_msgdef *m) {
   const upb_fielddef *val_f = upb_msgdef_itof(m, 1);
-  upb_msgval val = upb_msg_get(m, val_f);
+  upb_msgval val = upb_msg_get(msg, val_f);
   jsonenc_scalar(e, val, val_f);
 }
 
-const upb_msgdef *jsonenc_getanymsg(jsonenc *e, upb_strview type_url) {
+static const upb_msgdef *jsonenc_getanymsg(jsonenc *e, upb_strview type_url) {
   /* Find last '/', if any. */
   const char *end = type_url.data + type_url.size;
   const char *ptr = end;
+  const upb_msgdef *ret;
 
-  if (!e->ext_pool || type_url.size == 0) return NULL;
+  if (!e->ext_pool || type_url.size == 0) goto badurl;
 
   while (true) {
     if (--ptr == type_url.data) {
       /* Type URL must contain at least one '/', with host before. */
-      return NULL;
+      goto badurl;
     }
     if (*ptr == '/') {
       ptr++;
@@ -277,19 +292,29 @@ const upb_msgdef *jsonenc_getanymsg(jsonenc *e, upb_strview type_url) {
     }
   }
 
-  return upb_symtab_lookupmsg2(e->ext_pool, ptr, end - ptr);
+  ret = upb_symtab_lookupmsg2(e->ext_pool, ptr, end - ptr);
+
+  if (!ret) {
+    jsonenc_err(e, "Couldn't find Any type");
+  }
+
+  return ret;
+
+badurl:
+  jsonenc_err(e, "Bad type URL");
 }
 
 static void jsonenc_any(jsonenc *e, const upb_msg *msg, const upb_msgdef *m) {
   const upb_fielddef *type_url_f = upb_msgdef_itof(m, 1);
-  const upb_fielddef *value_f = upb_msgdef_itof(m, 1);
+  const upb_fielddef *value_f = upb_msgdef_itof(m, 2);
   upb_strview type_url = upb_msg_get(msg, type_url_f).str_val;
   upb_strview value = upb_msg_get(msg, value_f).str_val;
   const upb_msgdef *any_m = jsonenc_getanymsg(e, type_url);
   const upb_msglayout *any_layout = upb_msgdef_layout(any_m);
-  upb_msg *any = upb_msg_new(any_m, e->arena);
+  upb_arena *arena = jsonenc_arena(e);
+  upb_msg *any = upb_msg_new(any_m, arena);
 
-  if (!upb_decode(value.data, value.size, any, any_layout, e->arena)) {
+  if (!upb_decode(value.data, value.size, any, any_layout, arena)) {
     jsonenc_err(e, "Error decoding message in Any");
   }
 
@@ -297,9 +322,9 @@ static void jsonenc_any(jsonenc *e, const upb_msg *msg, const upb_msgdef *m) {
   jsonenc_string(e, type_url);
   jsonenc_putstr(e, ", ");
 
-  if (upb_msgdef_wellknowntype(m) == UPB_WELLKNOWN_UNSPECIFIED) {
+  if (upb_msgdef_wellknowntype(any_m) == UPB_WELLKNOWN_UNSPECIFIED) {
     /* Regular messages: {"@type": "...", "foo": 1, "bar": 2} */
-    jsonenc_msg(e, any, any_m);
+    jsonenc_msgfields(e, any, any_m);
   } else {
     /* Well-known type: {"@type": "...", "value": <well-known encoding>} */
     jsonenc_putstr(e, "value: ");
@@ -323,15 +348,17 @@ static void jsonenc_fieldpath(jsonenc *e, upb_strview path) {
 
   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.");
       }
-    } else {
-      jsonenc_putbytes(e, &ch, 1);
+      ch = *++ptr - 32;
     }
+
+    jsonenc_putbytes(e, &ch, 1);
     ptr++;
   }
 }
@@ -468,7 +495,7 @@ static void jsonenc_msgfield(jsonenc *e, const upb_msg *msg,
       jsonenc_listvalue(e, msg, m);
       break;
     case UPB_WELLKNOWN_STRUCT:
-      jsonenc_listvalue(e, msg, m);
+      jsonenc_struct(e, msg, m);
       break;
   }
 }
@@ -532,6 +559,7 @@ static void jsonenc_mapkey(jsonenc *e, upb_msgval val, const upb_fielddef *f) {
       break;
     case UPB_TYPE_STRING:
       jsonenc_stringbody(e, val.str_val);
+      break;
     default:
       UPB_UNREACHABLE();
   }
@@ -575,15 +603,12 @@ static void jsonenc_map(jsonenc *e, const upb_map *map, const upb_fielddef *f) {
 
 static void jsonenc_fieldval(jsonenc *e, const upb_fielddef *f,
                              upb_msgval val, bool *first) {
-  char buf[128];
   const char *name;
 
   if (e->options & UPB_JSONENC_PROTONAMES) {
     name = upb_fielddef_name(f);
   } else {
-    /* TODO(haberman): we need a better JSON name API. */
-    upb_fielddef_getjsonname(f, buf, sizeof(buf));
-    name = buf;
+    name = upb_fielddef_jsonname(f);
   }
 
   jsonenc_putsep(e, ", ", first);
@@ -598,13 +623,12 @@ static void jsonenc_fieldval(jsonenc *e, const upb_fielddef *f,
   }
 }
 
-static void jsonenc_msg(jsonenc *e, const upb_msg *msg, const upb_msgdef *m) {
+static void jsonenc_msgfields(jsonenc *e, const upb_msg *msg,
+                              const upb_msgdef *m) {
   upb_msgval val;
   const upb_fielddef *f;
   bool first = true;
 
-  jsonenc_putstr(e, "{");
-
   if (e->options & UPB_JSONENC_EMITDEFAULTS) {
     /* Iterate over all fields. */
     upb_msg_field_iter i;
@@ -620,11 +644,15 @@ static void jsonenc_msg(jsonenc *e, const upb_msg *msg, const upb_msgdef *m) {
       jsonenc_fieldval(e, f, val, &first);
     }
   }
+}
 
+static void jsonenc_msg(jsonenc *e, const upb_msg *msg, const upb_msgdef *m) {
+  jsonenc_putstr(e, "{");
+  jsonenc_msgfields(e, msg, m);
   jsonenc_putstr(e, "}");
 }
 
-size_t jsonenc_nullz(jsonenc *e, size_t size) {
+static size_t jsonenc_nullz(jsonenc *e, size_t size) {
   size_t ret = e->ptr - e->buf + e->overflow;
 
   if (size > 0) {
@@ -647,9 +675,11 @@ size_t upb_json_encode(const upb_msg *msg, const upb_msgdef *m,
   e.options = options;
   e.ext_pool = ext_pool;
   e.status = status;
+  e.arena = NULL;
 
   if (setjmp(e.err)) return -1;
 
   jsonenc_msg(&e, msg, m);
+  if (e.arena) upb_arena_free(e.arena);
   return jsonenc_nullz(&e, size);
 }

upb/msg.c

@@ -9,7 +9,7 @@
 
 /** upb_msg *******************************************************************/
 
-static char _upb_fieldtype_to_sizelg2[12] = {
+static const char _upb_fieldtype_to_sizelg2[12] = {
   0,
   0,  /* UPB_TYPE_BOOL */
   2,  /* UPB_TYPE_FLOAT */

upb/reflection.c

@@ -90,6 +90,18 @@ bool upb_msg_has(const upb_msg *msg, const upb_fielddef *f) {
   }
 }
 
+bool upb_msg_hasoneof(const upb_msg *msg, const upb_oneofdef *o) {
+  upb_oneof_iter i;
+  const upb_fielddef *f;
+  const upb_msglayout_field *field;
+
+  upb_oneof_begin(&i, o);
+  if (upb_oneof_done(&i)) return false;
+  f = upb_oneof_iter_field(&i);
+  field = upb_fielddef_layout(f);
+  return *oneofcase(msg, field) != 0;
+}
+
 upb_msgval upb_msg_get(const upb_msg *msg, const upb_fielddef *f) {
   if (!upb_fielddef_haspresence(f) || upb_msg_has(msg, f)) {
     return _upb_msg_getraw(msg, f);
@@ -136,8 +148,11 @@ upb_mutmsgval upb_msg_mutable(upb_msg *msg, const upb_fielddef *f,
   const upb_msglayout_field *field = upb_fielddef_layout(f);
   upb_mutmsgval ret;
   char *mem = PTR_AT(msg, field->offset, char);
+  bool wrong_oneof = in_oneof(field) && *oneofcase(msg, field) != field->number;
+
   memcpy(&ret, mem, sizeof(void*));
-  if (a && !ret.msg) {
+
+  if (a && (!ret.msg || wrong_oneof)) {
     if (upb_fielddef_ismap(f)) {
       const upb_msgdef *entry = upb_fielddef_msgsubdef(f);
       const upb_fielddef *key = upb_msgdef_itof(entry, UPB_MAPENTRY_KEY);
@@ -149,7 +164,12 @@ upb_mutmsgval upb_msg_mutable(upb_msg *msg, const upb_fielddef *f,
       UPB_ASSERT(upb_fielddef_issubmsg(f));
       ret.msg = upb_msg_new(upb_fielddef_msgsubdef(f), a);
     }
+
     memcpy(mem, &ret, sizeof(void*));
+
+    if (wrong_oneof) {
+      *oneofcase(msg, field) = field->number;
+    }
   }
   return ret;
 }

upb/reflection.h

@@ -44,6 +44,9 @@ upb_mutmsgval upb_msg_mutable(upb_msg *msg, const upb_fielddef *f, upb_arena *a)
 /* May only be called for fields where upb_fielddef_haspresence(f) == true. */
 bool upb_msg_has(const upb_msg *msg, const upb_fielddef *f);
 
+/* Returns whether any field is set in the oneof. */
+bool upb_msg_hasoneof(const upb_msg *msg, const upb_oneofdef *o);
+
 /* Sets the given field to the given value.  For a msg/array/map/string, the
  * value must be in the same arena.  */
 void upb_msg_set(upb_msg *msg, const upb_fielddef *f, upb_msgval val,

upb/text_encode.c

@@ -260,7 +260,7 @@ static const char *txtenc_unknown(txtenc *e, const char *ptr, const char *end,
     uint32_t tag;
     CHK(ptr = txtenc_parsevarint(ptr, end, &tag_64));
     CHK(tag_64 < UINT32_MAX);
-    tag = tag_64;
+    tag = (uint32_t)tag_64;
 
     if ((tag & 7) == UPB_WIRE_TYPE_END_GROUP) {
       CHK((tag >> 3) == groupnum);