Implement json decoding for Any message. (protocolbuffers#130)

* Implement json decoding for Any message.

type url may not appear as the first value in json. As a result,
other data cannot be resolved before resolving type url. To solve that,
this change caches the start and end position of unparsed values and
resolve them in end_any_object when type url has been resolved.

* Handle Any in switch

* Update json parser size

* Fix comments

* Sync upstream

* Add dependency on upb_pb for upb_json

* Debug failed test

* Fix cmake

* Update test generated files

* Remove debug tests

.gitignore

@@ -1,3 +1,4 @@
 *.s??
 obj/
 lib/
+bazel-*

BUILD

@@ -115,7 +115,10 @@ cc_library(
         "-pedantic",
         "-Wno-long-long",
     ],
-    deps = [":upb"],
+    deps = [
+        ":upb",
+        ":upb_pb",
+    ],
 )
 
 cc_library(

CMakeLists.txt

@@ -116,7 +116,8 @@ add_library(upb_json
   upb/json/parser.h
   upb/json/printer.h)
 target_link_libraries(upb_json
-  upb)
+  upb
+  upb_pb)
 add_library(upb_cc_bindings
   upb/bindings/stdc++/string.h)
 target_link_libraries(upb_cc_bindings

Makefile

@@ -191,6 +191,9 @@ endif
 upb_json_SRCS = \
   upb/json/parser.c \
   upb/json/printer.c \
+  upb/pb/encoder.c \
+  upb/pb/varint.c \
+  upb/sink.c \
 
 # Ideally we could keep this uncommented, but Git apparently sometimes skews
 # timestamps slightly at "clone" time, which makes "Make" think that it needs

tests/json/test_json.cc

@@ -177,7 +177,7 @@ void test_json_roundtrip_message(const char* json_src,
       env.env(), serialize_handlers, data_sink.Sink());
   upb::json::Parser* parser =
       upb::json::Parser::Create(
-          env.env(), parser_method, printer->input(), false);
+          env.env(), parser_method, NULL, printer->input(), false);
   env.ResetBytesSink(parser->input());
   env.Reset(json_src, strlen(json_src), false, false);
 

upb/def.c

@@ -371,7 +371,9 @@ static void assign_msg_wellknowntype(upb_msgdef *m) {
     m->well_known_type = UPB_WELLKNOWN_UNSPECIFIED;
     return;
   }
-  if (!strcmp(name, "google.protobuf.Duration")) {
+  if (!strcmp(name, "google.protobuf.Any")) {
+    m->well_known_type = UPB_WELLKNOWN_ANY;
+  } else if (!strcmp(name, "google.protobuf.Duration")) {
     m->well_known_type = UPB_WELLKNOWN_DURATION;
   } else if (!strcmp(name, "google.protobuf.Timestamp")) {
     m->well_known_type = UPB_WELLKNOWN_TIMESTAMP;
@@ -2146,6 +2148,14 @@ const upb_msgdef *upb_symtab_lookupmsg(const upb_symtab *s, const char *sym) {
   return def ? upb_dyncast_msgdef(def) : NULL;
 }
 
+const upb_msgdef *upb_symtab_lookupmsg2(const upb_symtab *s, const char *sym,
+                                        size_t len) {
+  upb_value v;
+  upb_def *def = upb_strtable_lookup2(&s->symtab, sym, len, &v) ?
+      upb_value_getptr(v) : NULL;
+  return def ? upb_dyncast_msgdef(def) : NULL;
+}
+
 const upb_enumdef *upb_symtab_lookupenum(const upb_symtab *s, const char *sym) {
   upb_value v;
   upb_def *def = upb_strtable_lookup(&s->symtab, sym, &v) ?

upb/def.h

@@ -285,6 +285,7 @@ typedef enum {
  */
 typedef enum {
   UPB_WELLKNOWN_UNSPECIFIED,
+  UPB_WELLKNOWN_ANY,
   UPB_WELLKNOWN_DURATION,
   UPB_WELLKNOWN_TIMESTAMP,
   /* number wrappers */
@@ -1543,6 +1544,8 @@ const upb_def *upb_symtab_resolve(const upb_symtab *s, const char *base,
                                   const char *sym);
 const upb_def *upb_symtab_lookup(const upb_symtab *s, const char *sym);
 const upb_msgdef *upb_symtab_lookupmsg(const upb_symtab *s, const char *sym);
+const upb_msgdef *upb_symtab_lookupmsg2(
+    const upb_symtab *s, const char *sym, size_t len);
 const upb_enumdef *upb_symtab_lookupenum(const upb_symtab *s, const char *sym);
 bool upb_symtab_add(upb_symtab *s, upb_def *const*defs, size_t n,
                     void *ref_donor, upb_status *status);