cache-crusher

A cache-buster for gulp, that integrates itself neatly into your tasks.

Usage

Let's say, you have this simplified directory structure:

gulpfile.js
src
├── client
│   ├── pages
│   │   └── main.html
│   └── styles
│       └── main.css
└── server
    └── scripts
        └── start.js

You may find the whole example project under examples/simple.

The following gulpfile.js effectively just copies the src-tree into dist and runs an express-server that statically serves the contents of dist/client under the url /static.

import gulp from 'gulp';
import crusherFactory from 'cache-crusher';

const crusher = crusherFactory({
  mapper: {
    counterparts: [{urlRoot: '/static', tagRoot: 'src/client'}]
  }
});

gulp.task('build-client-styles', function() {
  return gulp.src('src/client/styles/*.css')
  .pipe(crusher.pusher())
  .pipe(gulp.dest('dist/client/styles'));
});

gulp.task('build-client-pages', function() {
  return gulp.src('src/client/pages/*.html')
  .pipe(crusher.puller())
  .pipe(gulp.dest('dist/client/pages'));
});

gulp.task('build-server-scripts', function() {
  return gulp.src('src/server/scripts/*.js')
  .pipe(gulp.dest('dist/server/scripts'));
});

gulp.task('build', [
  'build-client-styles',
  'build-client-pages',
  'build-server-scripts']);

gulp.task('run', ['build'], function() {
  require('./dist/server/scripts/start.js');
});

gulp.task('default', ['run']);

…indeed, there is a little bit more to see here: As we pipe our resource (main.css) through a pusher, it will be automatically renamed to main-<digest>.css; as we pipe our referrer (main.html) through a puller, it will be scanned for references that start with the (configurable) urlBase /static and get that references be replaced by the appropriately renamed ones.

The only thing we must configure (but there is much more we may) is some mapping between urls and so called tags (These are path-like strings; by the default, they are generated as the vinyl-file's path relative to crusher.cwd, which in turn defaults to process.cwd()) by specifying the option mapper.counterparts.

---

A more complete development-setup that uses cache-crusher may be be created by this yeoman generator: webpack-versatile.

Architecture

A crusher object creates pusher and puller objects by means of its methods crusher.pusher and crusher.puller.

The puller side

A puller is a vinyl transform stream to pass your referrer files through. For each file it uses a extractorCatalog to find an appropriate extractor-class by the file extension. E.g. for an HTML-File this will be HtmlExtractor. This class is used to create an extractor by applying crusher.urlBase.

The extractor will provide a search pattern (a RegExp) that is used to create a stream-replacer. Each time the extractor encounters an url (which my refer a resource) it uses the mapper to convert that url to a tag. (A tag is an abstraction of a path inside some file-system.)

The mapper may reject that conversion request if that very url doesn't a refer resource to be crushed. In this case the extractor will just leave the url unmodified. If the conversion succeeds, the resulting tag will be to presented to the resolver's pull interface.

If the resolver receives a pull request, it may be either able to return the renamed tag immediately (if the corresponding push has occurred already). Otherwise that pull request will be queued until it can be resolved. When that renamed tag is available, it will be mapped back to an url, which is passed back to the extractor that will do the replacement of the referrer file's contents.

The pusher side

A pusher is a vinyl transform stream to pass your resource files through. For each file, a tagger is used to create its original tag. This tag is passed to the mapper to check if it could be converted to an url (Thus the mapper can be used to exclude files from being renamed). If this check passes, the file is piped through a hasher (provided by stream-hasher), which generates some hash-digest of this file's contents and (optionally) renames this file using that digest. A tagger is used to create a second tag after renaming. That tag pair is feed into the resolver's push interface, where this mapping is stored and may be used to resolve pending pull requests.

Optionally the pusher can be configured to not rename the resource file, but just emit the created digest together with the original tag and let the resolver append that digest as some query string.

API

var crusher = crusherFactory(options);

Creates a new crusher. Recognized options are:

• enabled (boolean, default: true): Enable the whole crusher. If this is false, puller() and pusher() will create simple pass-through-streams. You can use this to disable crusher for development builds. You can use cruser.setEnabled to change this setting afterwards.
• debug (boolean or console.log-like function): Log some useful stuff to this functions, if true, use console.error.
• cwd (a path-string, default: process.cwd()): This is used by the default tagger, that generates the tag as the file's path relative to crusher.cwd
• getExtractor (function(file)): use your extractor factory instead of crusher.prototype.getExtractor.
• getTagger (function(options)): use your tagger factory instead of crusher.prototype.getTagger.
• extractor (object, default: {urlBase: '/static/'}): the options to create the extractor.
• mapper (object): the options to create the mapper. Specify mapper._ to use your own mapper (see mapper.Mapper for required interface).
• resolver (object, default: {timeout: 10000}): the options to create the resolver. Specify resolver._ to use your own replacer (see resolver.Resolver for required interface).
• hasher (object, default: {rename: 'postfix', digestLength: 8}): the options to create the hasher.

extractor options

Set crusher.extractorOptions, which is used by getExtractor to create the extractor. Especially:

• catalog: specify an extractorCatalog, which should provide this interface:
  • catalog.getExtractor(file, options): returns an extractor.
• urlBase (default '/static/'): used by standard **extractor*s.

mapper options

• counterparts: an array of objects that define a relationship between urls and tags. Each of these may have these properties:

  • urlRoot (string):
  • tagRoot (string):
  • globs (string or array of strings):, converted to minimatch-objects to provide filtering.
  • globOptions (object): the options for minimatch creation
  • hasherOptions (object): overwrite

  To map an url to a tag, the mapper tries its counterparts in turn untill it finds a hit by this rule:

  • If the url starts with urlRoot, rel will be the remaining tail of the url.
  • If globs is provided, rel is tested against these pattern(s). If it fits:
  • The tag is created by appending rel to tagRoot.

  To map a tag to an url, the mapper tries its counterparts in turn untill it finds a hit by this rule:

  • If the tag starts with tagRoot, rel will be the remaining tail of the tag.
  • If globs is provided, rel is tested against these pattern(s). If it fits:
  • The url is created by appending rel to urlRoot.

  For example, a mapper with these counterparts:

      [ { urlRoot: '/static',
          tagRoot: 'src/client/scripts',
          globs: ['*.js'],
          globOptions: {matchBase: true}
        },
        { urlRoot: '/static',
          tagRoot: 'src/assets'
          globs: '!images/favicon.ico'
        }
      ]

  • would map the url '/static/foo/bar.js' to the tag 'src/client/scripts/foo/bar.js'.
  • would map the url '/static/foo/bar.jpg' to the tag 'src/assets/foo/bar.jpg'.
  • would not map the url '/static/images/favicon.ico', since it is excluded by globs.

resolver options

• timeout (number): pull-timeout in ms. If this is set, a TimeoutError will be thrown, when a pull-request has not been resolved for this duration.

hasher options

These options are just passed to stream-hasher:

• algorithm
• digestEncoding
• digestLength
• rename

Further there is:

• append (true, string or function. Only applicable if rename === false): Instead of renaming the resource, the url will be modified by appending some query-string constructed by 'digest:
  • true: "?digest"
  • string name: "?name=digest"
  • function fn`: "?" + fn(digest)

crusher.getTagger(options)

This is the default tagger factory. It recognizes the following options:

• _ (function(file)): use this function as tagger.
• relativeBase: create tag as file.relative relative to relativeBase.
• base: create tag as file.path relative to base.

Empty options is equivalent to {base: crusher.cwd}.

crusher.getExtractor(file)

This is the default extractor factory. It uses an extractorCatalog, specified by crusher.extractorOptions.catalog, if none is specified, a default one is provided (see later).

crusher.puller()

Create a new pull-stream (to pass referer files through).

crusher.pusher(options)

Create a new push-stream (to pass resource files through). options.tagger is used to create the tagger.

crusher.pullString(source, fileinfo, options)

An alternative non-stream interface to to crusher's puller-side. fileinfo will be used as argument to getExtractor. That extractor will be applied to source. Returns a Promise for the transformed source-string.

The default extractorCatalog

• getExtractor(file, options): uses this three-step-algorithm to create the extractor:

  • pass file's extension to getHandle look up a handle
  • pass that handle to getClass to look up an extractor-class.
  • instantiate that class (with options) to get an extractor.

• registerClass(Extractor, options): register new class Extractor by its handle. Set options.withExts to false to prevent auto extension association.

• registerExts(handle, exts): associate extension(s) exts with handle. For example: To let your .foo-files be processed by ScriptExtractor call crusher.extractorOptions.catalog.registerExts('script', '.foo').

• scanExtractors(dir, options): scan directory dir for files that define extractor-classes. options is passed to registerClass.

These basic extractor-classes are provided for now:

class	handle	extensions	comments
Extractor	'base'		looks for strings (enclosed in single or double quotes)
ScriptExtractor	'script'	'.js', '.coffee'	looks for strings in escaped quotes (as seen in compiled templates), too.
HtmlExtractor	'html'	'.html', '.xml', '.jade'	looks for strings preceded with 'src=' or 'href='

All standard extractors use option.urlBase as a required initial part of urls to be extracted. Example:

const crusher = crusherFactory({
  extractor: {
    urlRoot: '/app/assets/'
  },
  mapper: {
    counterparts: [{urlRoot: '/app', tagRoot: 'src/client'}]
  }
});

Notes:

• There has to be some correspondence between extractor.urlRoot and some counterpart.urlRoot to effectively extract anything at all.
• You may omit the closing '/' for counterpart.urlRoot and counterparts.tagRoot. It is appended automatically.
• You may just set urlRoot: '' for the cost of some performance penalty: A lot of none-url strings will be passed to mapper (and rejected here) instead of just passing them through by avoiding a RegExp-match.