MOTIVATION.md

Motivation

The process of building a project can be very much more complicated than you can imagine. Look at the following graph:

This could exist in reality.

Now

Grunt's tasks require reading/writing files, which bring tedious directory-making/removing and heavy disk IO. And it's difficult to tell a specific executing process from gruntfile.js.

Gulp is much better, file contents are transmited in Stream from plugin to plugin. Chained process definition is much more intelligible, but:

1. Gulp cannot make sure a file is read once at most in a building lifecycle

That means index.js will be read twice:

gulp.src('*.js')
gulp.src('index.js')

This should be optimized.

2. Gulp caches results in tasks but not files, which leads to many meaningless work

Developers always use gulp-watch to execute an increment building.

watch('**/*.js', function() {
    gulp.src('**/*.js')
});

All js files will be rebuild.

3. Gulp do not snap the "left" files, which are hard to selected

gulp.src() can only select known file types. Apparently, you will not use gulp.src('*').

4. Last but not least, we need more powerful functions to support almost any building process

We will see.

Improvement

Panto is trying to fixed these issues, in the following ways:

1. Process described by directed acyclic graph

This is the exact way to describe a building process, with universality.

2. Cache by hash of file content

If file content is not changed, result will not change either.

3. Reading into a shared file object

Will not read a file more than once.

4. Builtin dependencies graph

Minimum incremental building.