google-protobuf 3.0.0.alpha.2.0 → 3.0.0.alpha.3

data/ext/google/protobuf_c/defs.c

@@ -34,8 +34,6 @@
 // Common utilities.
 // -----------------------------------------------------------------------------
 
-const char* kDescriptorInstanceVar = "descriptor";
-
 static const char* get_str(VALUE str) {
   Check_Type(str, T_STRING);
   return RSTRING_PTR(str);
@@ -250,7 +248,7 @@ void Descriptor_free(void* _self) {
                        &self->pb_serialize_handlers);
   }
   if (self->json_serialize_handlers) {
-    upb_handlers_unref(self->pb_serialize_handlers,
+    upb_handlers_unref(self->json_serialize_handlers,
                        &self->json_serialize_handlers);
   }
   xfree(self);
@@ -1590,9 +1588,9 @@ VALUE Builder_add_message(VALUE _self, VALUE name) {
  * call-seq:
  *     Builder.add_enum(name, &block)
  *
- * Creates a new, empty enum descriptor with the given name, and invokes the block in
- * the context of an EnumBuilderContext on that descriptor. The block can then
- * call EnumBuilderContext#add_value to define the enum values.
+ * Creates a new, empty enum descriptor with the given name, and invokes the
+ * block in the context of an EnumBuilderContext on that descriptor. The block
+ * can then call EnumBuilderContext#add_value to define the enum values.
  *
  * This is the recommended, idiomatic way to build enum definitions.
  */

data/ext/google/protobuf_c/encode_decode.c

@@ -622,6 +622,49 @@ static const upb_pbdecodermethod *msgdef_decodermethod(Descriptor* desc) {
   return desc->fill_method;
 }
 
+
+// Stack-allocated context during an encode/decode operation. Contains the upb
+// environment and its stack-based allocator, an initial buffer for allocations
+// to avoid malloc() when possible, and a template for Ruby exception messages
+// if any error occurs.
+#define STACK_ENV_STACKBYTES 4096
+typedef struct {
+  upb_env env;
+  upb_seededalloc alloc;
+  const char* ruby_error_template;
+  char allocbuf[STACK_ENV_STACKBYTES];
+} stackenv;
+
+static void stackenv_init(stackenv* se, const char* errmsg);
+static void stackenv_uninit(stackenv* se);
+
+// Callback invoked by upb if any error occurs during parsing or serialization.
+static bool env_error_func(void* ud, const upb_status* status) {
+  stackenv* se = ud;
+  // Free the env -- rb_raise will longjmp up the stack past the encode/decode
+  // function so it would not otherwise have been freed.
+  stackenv_uninit(se);
+  rb_raise(rb_eRuntimeError, se->ruby_error_template,
+           upb_status_errmsg(status));
+  // Never reached: rb_raise() always longjmp()s up the stack, past all of our
+  // code, back to Ruby.
+  return false;
+}
+
+static void stackenv_init(stackenv* se, const char* errmsg) {
+  se->ruby_error_template = errmsg;
+  upb_env_init(&se->env);
+  upb_seededalloc_init(&se->alloc, &se->allocbuf, STACK_ENV_STACKBYTES);
+  upb_env_setallocfunc(
+      &se->env, upb_seededalloc_getallocfunc(&se->alloc), &se->alloc);
+  upb_env_seterrorfunc(&se->env, env_error_func, se);
+}
+
+static void stackenv_uninit(stackenv* se) {
+  upb_env_uninit(&se->env);
+  upb_seededalloc_uninit(&se->alloc);
+}
+
 /*
  * call-seq:
  *     MessageClass.decode(data) => message
@@ -631,7 +674,7 @@ static const upb_pbdecodermethod *msgdef_decodermethod(Descriptor* desc) {
  * and returns a message object with the corresponding field values.
  */
 VALUE Message_decode(VALUE klass, VALUE data) {
-  VALUE descriptor = rb_iv_get(klass, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(klass, descriptor_instancevar_interned);
   Descriptor* desc = ruby_to_Descriptor(descriptor);
   VALUE msgklass = Descriptor_msgclass(descriptor);
 
@@ -645,21 +688,17 @@ VALUE Message_decode(VALUE klass, VALUE data) {
 
   const upb_pbdecodermethod* method = msgdef_decodermethod(desc);
   const upb_handlers* h = upb_pbdecodermethod_desthandlers(method);
-  upb_pbdecoder decoder;
-  upb_sink sink;
-  upb_status status = UPB_STATUS_INIT;
+  stackenv se;
+  stackenv_init(&se, "Error occurred during parsing: %s");
 
-  upb_pbdecoder_init(&decoder, method, &status);
+  upb_sink sink;
   upb_sink_reset(&sink, h, msg);
-  upb_pbdecoder_resetoutput(&decoder, &sink);
+  upb_pbdecoder* decoder =
+      upb_pbdecoder_create(&se.env, method, &sink);
   upb_bufsrc_putbuf(RSTRING_PTR(data), RSTRING_LEN(data),
-                    upb_pbdecoder_input(&decoder));
+                    upb_pbdecoder_input(decoder));
 
-  upb_pbdecoder_uninit(&decoder);
-  if (!upb_ok(&status)) {
-    rb_raise(rb_eRuntimeError, "Error occurred during parsing: %s.",
-             upb_status_errmsg(&status));
-  }
+  stackenv_uninit(&se);
 
   return msg_rb;
 }
@@ -673,7 +712,7 @@ VALUE Message_decode(VALUE klass, VALUE data) {
  * and returns a message object with the corresponding field values.
  */
 VALUE Message_decode_json(VALUE klass, VALUE data) {
-  VALUE descriptor = rb_iv_get(klass, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(klass, descriptor_instancevar_interned);
   Descriptor* desc = ruby_to_Descriptor(descriptor);
   VALUE msgklass = Descriptor_msgclass(descriptor);
 
@@ -688,21 +727,16 @@ VALUE Message_decode_json(VALUE klass, VALUE data) {
   MessageHeader* msg;
   TypedData_Get_Struct(msg_rb, MessageHeader, &Message_type, msg);
 
-  upb_status status = UPB_STATUS_INIT;
-  upb_json_parser parser;
-  upb_json_parser_init(&parser, &status);
+  stackenv se;
+  stackenv_init(&se, "Error occurred during parsing: %s");
 
   upb_sink sink;
   upb_sink_reset(&sink, get_fill_handlers(desc), msg);
-  upb_json_parser_resetoutput(&parser, &sink);
+  upb_json_parser* parser = upb_json_parser_create(&se.env, &sink);
   upb_bufsrc_putbuf(RSTRING_PTR(data), RSTRING_LEN(data),
-                    upb_json_parser_input(&parser));
+                    upb_json_parser_input(parser));
 
-  upb_json_parser_uninit(&parser);
-  if (!upb_ok(&status)) {
-    rb_raise(rb_eRuntimeError, "Error occurred during parsing: %s.",
-             upb_status_errmsg(&status));
-  }
+  stackenv_uninit(&se);
 
   return msg_rb;
 }
@@ -813,7 +847,7 @@ static void putsubmsg(VALUE submsg, const upb_fielddef *f, upb_sink *sink,
   if (submsg == Qnil) return;
 
   upb_sink subsink;
-  VALUE descriptor = rb_iv_get(submsg, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(submsg, descriptor_instancevar_interned);
   Descriptor* subdesc = ruby_to_Descriptor(descriptor);
 
   upb_sink_startsubmsg(sink, getsel(f, UPB_HANDLER_STARTSUBMSG), &subsink);
@@ -935,7 +969,8 @@ static void putmap(VALUE map, const upb_fielddef *f, upb_sink *sink,
     VALUE value = Map_iter_value(&it);
 
     upb_sink entry_sink;
-    upb_sink_startsubmsg(&subsink, getsel(f, UPB_HANDLER_STARTSUBMSG), &entry_sink);
+    upb_sink_startsubmsg(&subsink, getsel(f, UPB_HANDLER_STARTSUBMSG),
+                         &entry_sink);
     upb_sink_startmsg(&entry_sink);
 
     put_ruby_value(key, key_field, Qnil, depth + 1, &entry_sink);
@@ -956,7 +991,7 @@ static void putmsg(VALUE msg_rb, const Descriptor* desc,
 
   // Protect against cycles (possible because users may freely reassign message
   // and repeated fields) by imposing a maximum recursion depth.
-  if (depth > UPB_SINK_MAX_NESTING) {
+  if (depth > ENCODE_MAX_NESTING) {
     rb_raise(rb_eRuntimeError,
              "Maximum recursion depth exceeded during encoding.");
   }
@@ -1065,7 +1100,7 @@ static const upb_handlers* msgdef_json_serialize_handlers(Descriptor* desc) {
  * wire format.
  */
 VALUE Message_encode(VALUE klass, VALUE msg_rb) {
-  VALUE descriptor = rb_iv_get(klass, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(klass, descriptor_instancevar_interned);
   Descriptor* desc = ruby_to_Descriptor(descriptor);
 
   stringsink sink;
@@ -1074,15 +1109,16 @@ VALUE Message_encode(VALUE klass, VALUE msg_rb) {
   const upb_handlers* serialize_handlers =
       msgdef_pb_serialize_handlers(desc);
 
-  upb_pb_encoder encoder;
-  upb_pb_encoder_init(&encoder, serialize_handlers);
-  upb_pb_encoder_resetoutput(&encoder, &sink.sink);
+  stackenv se;
+  stackenv_init(&se, "Error occurred during encoding: %s");
+  upb_pb_encoder* encoder =
+      upb_pb_encoder_create(&se.env, serialize_handlers, &sink.sink);
 
-  putmsg(msg_rb, desc, upb_pb_encoder_input(&encoder), 0);
+  putmsg(msg_rb, desc, upb_pb_encoder_input(encoder), 0);
 
   VALUE ret = rb_str_new(sink.ptr, sink.len);
 
-  upb_pb_encoder_uninit(&encoder);
+  stackenv_uninit(&se);
   stringsink_uninit(&sink);
 
   return ret;
@@ -1095,7 +1131,7 @@ VALUE Message_encode(VALUE klass, VALUE msg_rb) {
  * Encodes the given message object into its serialized JSON representation.
  */
 VALUE Message_encode_json(VALUE klass, VALUE msg_rb) {
-  VALUE descriptor = rb_iv_get(klass, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(klass, descriptor_instancevar_interned);
   Descriptor* desc = ruby_to_Descriptor(descriptor);
 
   stringsink sink;
@@ -1104,64 +1140,18 @@ VALUE Message_encode_json(VALUE klass, VALUE msg_rb) {
   const upb_handlers* serialize_handlers =
       msgdef_json_serialize_handlers(desc);
 
-  upb_json_printer printer;
-  upb_json_printer_init(&printer, serialize_handlers);
-  upb_json_printer_resetoutput(&printer, &sink.sink);
+  stackenv se;
+  stackenv_init(&se, "Error occurred during encoding: %s");
+  upb_json_printer* printer =
+      upb_json_printer_create(&se.env, serialize_handlers, &sink.sink);
 
-  putmsg(msg_rb, desc, upb_json_printer_input(&printer), 0);
+  putmsg(msg_rb, desc, upb_json_printer_input(printer), 0);
 
   VALUE ret = rb_str_new(sink.ptr, sink.len);
 
-  upb_json_printer_uninit(&printer);
+  stackenv_uninit(&se);
   stringsink_uninit(&sink);
 
   return ret;
 }
 
-/*
- * call-seq:
- *     Google::Protobuf.encode(msg) => bytes
- *
- * Encodes the given message object to protocol buffers wire format. This is an
- * alternative to the #encode method on msg's class.
- */
-VALUE Google_Protobuf_encode(VALUE self, VALUE msg_rb) {
-  VALUE klass = CLASS_OF(msg_rb);
-  return Message_encode(klass, msg_rb);
-}
-
-/*
- * call-seq:
- *     Google::Protobuf.encode_json(msg) => json_string
- *
- * Encodes the given message object to its JSON representation. This is an
- * alternative to the #encode_json method on msg's class.
- */
-VALUE Google_Protobuf_encode_json(VALUE self, VALUE msg_rb) {
-  VALUE klass = CLASS_OF(msg_rb);
-  return Message_encode_json(klass, msg_rb);
-}
-
-/*
- * call-seq:
- *     Google::Protobuf.decode(class, bytes) => msg
- *
- * Decodes the given bytes as protocol buffers wire format under the
- * interpretation given by the given class's message definition. This is an
- * alternative to the #decode method on the given class.
- */
-VALUE Google_Protobuf_decode(VALUE self, VALUE klass, VALUE msg_rb) {
-  return Message_decode(klass, msg_rb);
-}
-
-/*
- * call-seq:
- *     Google::Protobuf.decode_json(class, json_string) => msg
- *
- * Decodes the given JSON string under the interpretation given by the given
- * class's message definition. This is an alternative to the #decode_json method
- * on the given class.
- */
-VALUE Google_Protobuf_decode_json(VALUE self, VALUE klass, VALUE msg_rb) {
-  return Message_decode_json(klass, msg_rb);
-}

data/ext/google/protobuf_c/extconf.rb

@@ -2,7 +2,9 @@
 
 require 'mkmf'
 
-$CFLAGS += " -O3 -std=c99 -Wno-unused-function -DNDEBUG "
+$CFLAGS += " -O3 -std=c99 -Wno-unused-function " +
+           "-Wno-declaration-after-statement -Wno-unused-variable " +
+           "-Wno-sign-compare -DNDEBUG "
 
 $objs = ["protobuf.o", "defs.o", "storage.o", "message.o",
          "repeated_field.o", "map.o", "encode_decode.o", "upb.o"]

data/ext/google/protobuf_c/message.c

@@ -53,7 +53,7 @@ rb_data_type_t Message_type = {
 };
 
 VALUE Message_alloc(VALUE klass) {
-  VALUE descriptor = rb_iv_get(klass, kDescriptorInstanceVar);
+  VALUE descriptor = rb_ivar_get(klass, descriptor_instancevar_interned);
   Descriptor* desc = ruby_to_Descriptor(descriptor);
   MessageHeader* msg = (MessageHeader*)ALLOC_N(
       uint8_t, sizeof(MessageHeader) + desc->layout->size);
@@ -63,7 +63,7 @@ VALUE Message_alloc(VALUE klass) {
   // a collection happens during object creation in layout_init().
   VALUE ret = TypedData_Wrap_Struct(klass, &Message_type, msg);
   msg->descriptor = desc;
-  rb_iv_set(ret, kDescriptorInstanceVar, descriptor);
+  rb_ivar_set(ret, descriptor_instancevar_interned, descriptor);
 
   layout_init(desc->layout, Message_data(msg));
 
@@ -86,7 +86,7 @@ static VALUE which_oneof_field(MessageHeader* self, const upb_oneofdef* o) {
   size_t case_ofs =
       self->descriptor->layout->
       fields[upb_fielddef_index(first_field)].case_offset;
-  uint32_t oneof_case = *((uint32_t*)(Message_data(self) + case_ofs));
+  uint32_t oneof_case = *((uint32_t*)((char*)Message_data(self) + case_ofs));
 
   if (oneof_case == ONEOF_CASE_NONE) {
     return Qnil;
@@ -329,6 +329,31 @@ VALUE Message_inspect(VALUE _self) {
   return str;
 }
 
+
+VALUE Message_to_h(VALUE _self) {
+  MessageHeader* self;
+  TypedData_Get_Struct(_self, MessageHeader, &Message_type, self);
+
+  VALUE hash = rb_hash_new();
+
+  upb_msg_field_iter it;
+  for (upb_msg_field_begin(&it, self->descriptor->msgdef);
+       !upb_msg_field_done(&it);
+       upb_msg_field_next(&it)) {
+    const upb_fielddef* field = upb_msg_iter_field(&it);
+    VALUE msg_value = layout_get(self->descriptor->layout, Message_data(self),
+                                 field);
+    VALUE msg_key   = ID2SYM(rb_intern(upb_fielddef_name(field)));
+    if (upb_fielddef_label(field) == UPB_LABEL_REPEATED) {
+      msg_value = RepeatedField_to_ary(msg_value);
+    }
+    rb_hash_aset(hash, msg_key, msg_value);
+  }
+  return hash;
+}
+
+
+
 /*
  * call-seq:
  *     Message.[](index) => value
@@ -376,7 +401,7 @@ VALUE Message_index_set(VALUE _self, VALUE field_name, VALUE value) {
  * message class's type.
  */
 VALUE Message_descriptor(VALUE klass) {
-  return rb_iv_get(klass, kDescriptorInstanceVar);
+  return rb_ivar_get(klass, descriptor_instancevar_interned);
 }
 
 VALUE build_class_from_descriptor(Descriptor* desc) {
@@ -397,8 +422,14 @@ VALUE build_class_from_descriptor(Descriptor* desc) {
       // their own toplevel constant class name.
       rb_intern("Message"),
       rb_cObject);
-  rb_iv_set(klass, kDescriptorInstanceVar, get_def_obj(desc->msgdef));
+  rb_ivar_set(klass, descriptor_instancevar_interned,
+              get_def_obj(desc->msgdef));
   rb_define_alloc_func(klass, Message_alloc);
+  rb_require("google/protobuf/message_exts");
+  rb_include_module(klass, rb_eval_string("Google::Protobuf::MessageExts"));
+  rb_extend_object(
+      klass, rb_eval_string("Google::Protobuf::MessageExts::ClassMethods"));
+
   rb_define_method(klass, "method_missing",
                    Message_method_missing, -1);
   rb_define_method(klass, "initialize", Message_initialize, -1);
@@ -407,6 +438,8 @@ VALUE build_class_from_descriptor(Descriptor* desc) {
   rb_define_method(klass, "clone", Message_dup, 0);
   rb_define_method(klass, "==", Message_eq, 1);
   rb_define_method(klass, "hash", Message_hash, 0);
+  rb_define_method(klass, "to_h", Message_to_h, 0);
+  rb_define_method(klass, "to_hash", Message_to_h, 0);
   rb_define_method(klass, "inspect", Message_inspect, 0);
   rb_define_method(klass, "[]", Message_index, 1);
   rb_define_method(klass, "[]=", Message_index_set, 2);
@@ -415,6 +448,7 @@ VALUE build_class_from_descriptor(Descriptor* desc) {
   rb_define_singleton_method(klass, "decode_json", Message_decode_json, 1);
   rb_define_singleton_method(klass, "encode_json", Message_encode_json, 1);
   rb_define_singleton_method(klass, "descriptor", Message_descriptor, 0);
+
   return klass;
 }
 
@@ -427,7 +461,7 @@ VALUE build_class_from_descriptor(Descriptor* desc) {
  */
 VALUE enum_lookup(VALUE self, VALUE number) {
   int32_t num = NUM2INT(number);
-  VALUE desc = rb_iv_get(self, kDescriptorInstanceVar);
+  VALUE desc = rb_ivar_get(self, descriptor_instancevar_interned);
   EnumDescriptor* enumdesc = ruby_to_EnumDescriptor(desc);
 
   const char* name = upb_enumdef_iton(enumdesc->enumdef, num);
@@ -447,7 +481,7 @@ VALUE enum_lookup(VALUE self, VALUE number) {
  */
 VALUE enum_resolve(VALUE self, VALUE sym) {
   const char* name = rb_id2name(SYM2ID(sym));
-  VALUE desc = rb_iv_get(self, kDescriptorInstanceVar);
+  VALUE desc = rb_ivar_get(self, descriptor_instancevar_interned);
   EnumDescriptor* enumdesc = ruby_to_EnumDescriptor(desc);
 
   int32_t num = 0;
@@ -467,7 +501,7 @@ VALUE enum_resolve(VALUE self, VALUE sym) {
  * EnumDescriptor corresponding to this enum type.
  */
 VALUE enum_descriptor(VALUE self) {
-  return rb_iv_get(self, kDescriptorInstanceVar);
+  return rb_ivar_get(self, descriptor_instancevar_interned);
 }
 
 VALUE build_module_from_enumdesc(EnumDescriptor* enumdesc) {
@@ -492,7 +526,8 @@ VALUE build_module_from_enumdesc(EnumDescriptor* enumdesc) {
   rb_define_singleton_method(mod, "lookup", enum_lookup, 1);
   rb_define_singleton_method(mod, "resolve", enum_resolve, 1);
   rb_define_singleton_method(mod, "descriptor", enum_descriptor, 0);
-  rb_iv_set(mod, kDescriptorInstanceVar, get_def_obj(enumdesc->enumdef));
+  rb_ivar_set(mod, descriptor_instancevar_interned,
+              get_def_obj(enumdesc->enumdef));
 
   return mod;
 }

data/ext/google/protobuf_c/protobuf.c

@@ -64,6 +64,15 @@ rb_encoding* kRubyStringUtf8Encoding;
 rb_encoding* kRubyStringASCIIEncoding;
 rb_encoding* kRubyString8bitEncoding;
 
+// Ruby-interned string: "descriptor". We use this identifier to store an
+// instance variable on message classes we create in order to link them back to
+// their descriptors.
+//
+// We intern this once at module load time then use the interned identifier at
+// runtime in order to avoid the cost of repeatedly interning in hot paths.
+const char* kDescriptorInstanceVar = "descriptor";
+ID descriptor_instancevar_interned;
+
 // -----------------------------------------------------------------------------
 // Initialization/entry point.
 // -----------------------------------------------------------------------------
@@ -71,6 +80,7 @@ rb_encoding* kRubyString8bitEncoding;
 // This must be named "Init_protobuf_c" because the Ruby module is named
 // "protobuf_c" -- the VM looks for this symbol in our .so.
 void Init_protobuf_c() {
+  descriptor_instancevar_interned = rb_intern(kDescriptorInstanceVar);
   VALUE google = rb_define_module("Google");
   VALUE protobuf = rb_define_module_under(google, "Protobuf");
   VALUE internal = rb_define_module_under(protobuf, "Internal");
@@ -86,13 +96,6 @@ void Init_protobuf_c() {
   RepeatedField_register(protobuf);
   Map_register(protobuf);
 
-  rb_define_singleton_method(protobuf, "encode", Google_Protobuf_encode, 1);
-  rb_define_singleton_method(protobuf, "decode", Google_Protobuf_decode, 2);
-  rb_define_singleton_method(protobuf, "encode_json",
-                             Google_Protobuf_encode_json, 1);
-  rb_define_singleton_method(protobuf, "decode_json",
-                             Google_Protobuf_decode_json, 2);
-
   rb_define_singleton_method(protobuf, "deep_copy",
                              Google_Protobuf_deep_copy, 1);