Add jamen/pull-fs-meta

[jamen]

Hello, back with another module, pull-fs-meta. This one is experimental, but seems to be pretty cool so far. It's my approach to an fs module on pull-stream.

It appears to operate much like Vinylinside gulp, except it actually just streams the file contents, and manages the metadata behind the scenes. That way you can have reusable pull-* modules that compile contents (like pull-css) operate like a build system with this, without need to create wrappers and whatnot:

var fs = require('pull-fs-meta')()
var css = require('pull-css')

pull(
  // Read file contents:
  fs.read('out/*.css'),
  // Parse contents as CSS:
  css(),
  // Then transform AST, like PostCSS
  // Stringify it after transformed:
  css.stringify(),
  // And write out
  fs.write('out')
)

[dominictarr]

curious: does this return a stream of streams, or does it concatenate all the streams?

[dominictarr]

it might be interesting to use with http://npm.im/tar-fs

[jamen]

does this return a stream of streams, or does it concatenate all the streams?

I don't understand completely 😅. But I'm pretty sure it doesn't return streams of streams. Maybe I can explain.

The functions are created in a factory:

var fs = require('pull-fs-meta')()

They share this private array called meta in that scope. This avoid having to put it on/around the contents itself (like Vinyl).

fs.read resolves globs to paths then reads to contents, and pushes metadata on meta about the file is just read. (For now just [path, base]). Each read correlates with an index on meta.

fs.files is a passthrough stream that is made for changing meta. (For things like extension name, moving files, clearing the metadata, etc.)

fs.write keeps an index for each write call so it can pick the info from meta with the correlating contents it read. (Because the 1:1 read/write ratio)

[jamen]

I'm going to redo the name and stuff. I'll open up a new PR later.