Add transform and extract APIs

[bradlc]

This PR adds purge.transform and purge.extract configuration options.

purge.extract

This is an alternative way to define class extractors:

purge: {
  extract: (content) => { /* ... */ },
}

// same as:
purge: {
  options: {
    defaultExtractor: (content) => { /* ... */ },
  },
}

purge: {
  extract: {
    DEFAULT: (content) => { /* ... */ },
    html: (content) => { /* ... */ },
  },
}

// same as:
purge: {
  options: {
    defaultExtractor: (content) => { /* ... */ },
    extractors: [
      {
        extensions: ['html'],
        extractor: (content) => { /* ... */ },
      },
    ],
  },
}

purge.transform

purge.transform is a new option that allows you to specify transformation functions that run on file contents before it is passed to the class extractor. One use-case for this is converting markdown to HTML:

// One transformer for every file
purge: {
  transform: (content) => { /* ... */ },
}

// Transformer for .md files, and a (optional) default transformer for all other file types
purge: {
  transform: {
    DEFAULT: (content) => { /* ... */ },
    md: markdownToHtml,
  },
}

Notes

• The Svelte transform (content.replace(/(?:^|\s)class:/g, ' ')) is now implemented as a transformer.
• purge.options.defaultExtractor and purge.options.extractors still work and override any extractors defined using purge.extract. One thing that I noticed about this is that when using a custom defaultExtractor the purge.preserveHtmlElements option is still respected, but this isn't the case with custom extractors defined in purge.extractors. This behaviour has been maintained but I'm wondering if this was intentional?

[adamwathan]

Thanks man! I think the preserveHtmlElements thing is intentional (I vaguely remember an issue about it). We can reconsider how that should work for v3 though if it feels weird.