Best way to send binary data with Thrift

Question

I have a structure in c++ which stores bytes like this:

struct RemoteData 
{
    /// some other fields here

    unsigned char* buf;
    int bufLen;     
};

And I need to send this data to remote service, written in C++, via thrift. I found three ways how to map this structure to thrift idl:

Using container types like this:

struct RemoteData 
{
    1: list<BYTE> buf,
    ...
}

Using binary type:

struct RemoteData 
{
    1: binary buf,
    ...
}

Storing data in string type:

struct RemoteData 
{
    1: string buf,
    ...
}

What is the best way?

score 17 · Accepted Answer · answered Dec 14 '12 at 22:23

17

The value contained in thrift string type must be UTF8-encoded, otherwise some client won't be able to read it (Java thrift client for example).

Thrift list<byte> type will be converted into std::vector<int8_t> in c++, but in other languages it will be not so good (for example, in java it will be compiled into suboptimal List<Byte>.

The binary type is the best choice in terms of interoperability with clients in other languages and the correctness of protocol definition.

Since all 3 definitions produce messages of roughly the same size, I'd go with binary: even if you don't need interoperability with other languages now, you may need it in future.

answered Dec 14 '12 at 22:23

Wildfire

6,358
2
34
50

1

But in thrift documentation http://thrift.apache.org/docs/types/ written about binary type "N.B.: This is currently a specialized form of the string type above, added to provide better interoperability with Java. The current plan-of-record is to elevate this to a base type at some point." – DanilaNV Dec 17 '12 at 05:43
6

@DanilaNV This documentation is probably outdated: Thrift binary type is as stable, convenient and widely used as other types, so I don't see why it is not in the list of basic types yet. – Wildfire Dec 17 '12 at 12:08

Best way to send binary data with Thrift

1 Answers1