English | 中文
minipbjs is a command line tool based on pbjs
from protobuf.js project, which targets on minimizing the generated javascript code while keeping the same functionality. Compatible to Node.js
, browsers, wechat miniprogram and other instant-games. Making it perfectly suitable for the scenarios like wechat miniprogram which requires code size restrictions.
Installation & Example
installing
# install `minipbjs` globally with option -g
$> npm install minipbjs -g
Generating Javascript code
# With same usage with pbjs
$> minipbjs --keep-case # Keeps field casing instead of converting to camel case.
--root PB # Specifies an alternative protobuf.roots name.
# Adds a directory to the include path. Following with target protofiles
--path /path/to/protofiles a.proto b.proto c.proto ...
# Notice: --out option in pbjs refers to a file, but it refers to a dir in minipbjs
--out /path/to/protobuf-bundles/
# Additional options in minipbjs:
# --name Specifies the output Javascript filename,'protobuf-bundles' by default.
--name protobuf-bundles-mini
# Automatically generates Javascript files specified with --name options.
# Along with the uglified min.js.
$> ls -al /path/to/protobuf-bundles/
> -rw-r--r-- 1 mustime staff xxxx protobuf-bundles-mini.js
> -rw-r--r-- 1 mustime staff xxxx protobuf-bundles-mini.min.js
Using in program
// require the necessary protobuf-library first
require('protobuf-library-minimal.js');
require('protobuf-bundles-mini.js');
// a simple test(PB variable is specified by `--root PB` option)
var payload = { 'a': 1, 'b': 'test' };
var foo = PB.foo.Foo.create(payload);
var bytes = PB.foo.Foo.encode(foo).finish();
var foo2 = PB.foo.Foo.decode(bytes);
var foo3 = PB.foo.Foo.fromObject(PB.foo.Foo.toObject(foo2));
// they should print the same
console.log(foo.a, foo2.a, foo3.a);
console.log(foo.b, foo2.b, foo3.b);
// see futher tests under 'minipbjs/tests'
Comparison with pbjs
Under the hood
The Javascript static code generated by pbjs --target static
contains function stubs like constructor
, create
, encode
, decode
, fromObject
and toObject
, etc. These function stubs are generated individually for each message. Which means that these function stubs are distinguish to each other and unlikely be able to share with other messages. Normally we have to strip some less frequently used functionality like create
/fromObject
/toObject
by --no-create
/--no-convert
, etc., but the resulting min.js is usually remarkable huge.
However, minipbjs
pulls basic info from each message(id, name, default value, options, etc.) into a single map. And providing major functionality(currently supports create
/encode
/decode
/encodeDelimited
/decodeDelimited
/fromObject
/toObject
/toJSON
/verify
) as shared implements for all message. No need to care about the negligible code size anymore.
Comparison
Take my recent project as example, which is a wechat minigame contains more than 2500 messages. On my 2019 RMBP-15, it costs more than 40s to run pbjs --target static
, resulting in a 5.2+m min.js file. Meanwhile, running with minipbjs
(remove --target static
option as well, or minipbjs will perform the same as pbjs) costs only 1.6s, resulting with a 160+kB min.js.
Liscense
This project is liscensed under The MIT Liscense
.
Enjoy. Let me know if you have any suggestion or encountering with any problem : ).