[THRIFT-3839] Performance issue with big message deserialization using php extension

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.9.3
• Fix Version/s: 0.10.0
• Component/s: PHP - Library
• Assignee: Myroslav Kosinskyi
• Reporter: Myroslav Kosinskyi
• Created: 23/May/16 08:57
• Updated: 28/Sep/16 13:28
• Resolved: 25/Sep/16 17:39

Description

I have found performance issue when tried to deserialize big thrift binary message with enabled thrift_protocol php extension.

Messsage size was 10 mb and it took about 30 seconds to deserialize it.

When i have done debug of php extension and php library i have found that issue is because small read buffer is used in TBufferedTransport and i cannot change it from TBinarySerializer::deserialize method.

So i have added parameter $buffer_size to function thrift_protocol_read_binary from php extension.

And also this parameter i have added to method TBinarySerializer::deserialize from php library.

And i extended class Thrift\Transport\TMemoryBuffer by method putBack, so this class will be used for desearilization without TBufferedTransport warapper.

After these changes it takes less than a second to deserizlize message with 10 mb size id read buffer 512 kb.

Here is the pull request https://github.com/apache/thrift/pull/1014