Question

I am working on a project that uses google proto buffers for internal data exchanges. It works fine as advertised, however it doesn’t works as expected for messages with repeated fields in them. Here is an example:

Sample test.proto file:

message Test {
    optional string t = 1;
}

message Tests {
    repeated Test testsList = 1;
}

Generate erlang code:

1> protobuffs_compile:scan_file("test.proto").
=INFO REPORT==== 14-Sep-2012::16:38:25 ===
Writing header file to "test_pb.hrl"

=INFO REPORT==== 14-Sep-2012::16:38:25 ===
Writing beam file to "test_pb.beam"
ok

Generated test_pb.hrl:

-ifndef(TEST_PB_H).
-define(TEST_PB_H, true).
-record(test, {
    t
}).
-endif.

-ifndef(TESTS_PB_H).
-define(TESTS_PB_H, true).
-record(tests, {
    testslist = []
}).
-endif.

Encode:

5> test_pb:encode_tests({tests, [{test, <<"t">>}]}).
<<10,3,10,1,116>>

Decode:

6> test_pb:decode_tests(<<10,3,10,1,116>>).
{[{test,"t"}],undefined}

As in the above example, decoding doesn’t give back the expected record tuple:

{tests, [{test, <<"t">>}]}

Has someone faced a similar issue before? Where am I missing the trick?
Any pointers and help will be greatly appreciated.

For versioning info, deps line from my rebar.config:

{protobuffs, "0.7.0", {git, "git://github.com/basho/erlang_protobuffs.git", {tag, "0.7.0"}}}

Answer 1

Just for an update.

0.7.0, 0.6.0 and HEAD were all giving me the same problem as described above.

After discussion on #riak IRC, I realized we need to use stable release marked by this commit hash: e0f5f6ea4c3dcb4e7b824496d2b48333fbd5a8c8 which solves the above issue.