JSON, Protobuf-c++,pbc,sproto performance testing and Solutions

The purpose of this paper is to compare the advantages and disadvantages of these four methods and to give a solution.

1. Test Environment Description CPU:I3-2310m CPU @ 2.10GHzVersion of the Os:linux (64-bit) Library: lua5.1, luajit2.0, note: 1. The Sproto library author uses the lua5.2 new Library bitwise Operators in the code, and does not match the lua5.1 environment we use, so we can replace lua5.2 with lua5.1 and Bit.numberlua. Operators.
2. Also tried to use the Protoc-gen-lua, but the solution of the table needs to be manually parsed, and time-consuming. 2. Test 2.0 test Function Interface Description

Library	Function interface	Input/output parameter	Note
Sproto	1.en=cookie.encode (AB) 2.de=cookie.decode (en)	1. Input:ab is table2. Input:en is binary string	1. The schema of the Sproto is very strong and flexible2. Writing proto files is very simple and very user-friendly 3. Easy to use
Sproto (Nopack)	Similar with Sproto	Similar with Sproto
Pbc-lua	1.en=protobuf.encode ("Cookie.cookievalue", Cookievalue) 2.de=protobuf.decode ("Cookie.cookievalue", en)	1. Input:cookievalue is table,2. Input en is binary	This decode out of the table is lazy expansion, access to the table can be solved, or it will appear garbled
Protobuf-c++	1. Str=bilin. Serialize (TB, ' cookie ') 2.bilin. Parse (str, ' Cookie ', #str)	1. INPUT:TB is a table,2.input:str is binary string	1. A change, more difficult to maintain 2. Time-consuming and compression-ratio effect is good, compared to other methods
Lua-cjson	1. En=cjson.encode (TB) 2.de=cjson.decode (en)	1. INPUT:TB is a table,2. Input:en is Cjson string	1. The space consumption is very large 2. It takes too much time to 3. Advantages: (For our existing use) No need to convert the program, that is, the solution is used

2.1 Test Results

Library	encode times (1M)	decode Times (1M)	size (byets)
Sproto	4. 6501181125641s	one. 385328769684s	139
sproto (nopack)	4. 0319328308105s	9. 9806959629059s	272
Pbc-lua	9. 3938179016113s	7. 0795350074768s	117
protobuf-c++	6. 0948710441589s	9. 6896359920502s	117
lua-cjson	. 541377067566s	. 335454940796s	437

Test data: {"Pbid": [{"id":100188, "value": 3},{"id":100189, "Value": 3}, {"id":100190, "Value": 3}], "daily": [{" daily_id ": 1125," date_id ": [{" id ":100188," value ": 1}]},{" daily_id ": 1126," date_id ": [{" id ":100118," value ": 1 }]}], "Total": [{"id":100188, "value": 1}], "segments": [{"Type": 1, "ct":1415384160, "ttl":1415384160, "id" :100001}], "imp": [{"Win": 2, "bid": 3, "Time": 8160, "id":100188},{"Win": 2, "bid": 3, "Time": 8155, "id": 100182},{"Win": 2, "bid": 3, "Time": 8157, "id":100181}]}

3. Results analysis in the "2.0 Test Function Interface description" in the "note" column, has pointed out the advantages and disadvantages of these five methods, you can choose two scenarios as our solution to this problem:

• Scenario One: protobuf-c++, features: In time and space consumption, the advantages are obvious, but once there are new needs to change, maintenance is more complex;
• Scenario Two: Sproto, features: In time and space consumption, but also very advantageous, maintenance is very convenient, please refer to Sproto:https://github.com/cloudwu/sproto, PBC please refer to: https:// GITHUB.COM/CLOUDWU/PBC; Thank you.

JSON, Protobuf-c++,pbc,sproto performance testing and Solutions