remix-run/remix route-pattern@0.21.0

route-pattern v0.21.0

on GitHub

Minor Changes

• BREAKING CHANGE: Remove the compareFn parameter from match and matchAll methods

  Matches always sort by specificity (most specific first). If you need a different order, sort the result of matchAll yourself.

  import * as Specificity from 'remix/route-pattern/specificity'
  
  // before
  matcher.matchAll(url, Specificity.ascending)
  
  // after
  matcher.matchAll(url).sort(Specificity.ascending)

• BREAKING CHANGE: New modular APIs and subpath exports

  Previously, this package shipped the default export and a /specificity export.
  A typical Remix app does not do any client-side matching but all the matching logic would ship to the browser anyway causing JS bloat.

  Now, features are organized into separate subpath exports, so even without a bundler, only the code you need ends up in the browser.
  For example, this reduced JS from route-pattern in demos/bookstore from 25kB (14.9kB compressed) to 8.8kb (7kB compressed) which amounts to ~65% reduction (~53% reduction compressed).

  To achieve this, we've reworked our core APIs to be simpler and more independently useful.
  So instead of a single RoutePattern class that does it all (.href, .match, ...), the new RoutePattern class is a thin layer around the parsed pattern that includes RoutePattern.parse static method for parsing and .source, .toString() and .toJSON() for serialization.

  The rest of the functionality comes from dedicated subpath exports:

  • remix/route-pattern/href : Generate hrefs for patterns with type safe params.
  • remix/route-pattern/match : Match against one pattern with type inference for params. Or match against many patterns with deterministic ranking and attached data.
  • remix/route-pattern/join : Combine two patterns into one. Override protocol, hostname, port. Join pathnames. Merge search constraints.

  remix/route-pattern/specificity remains the same as before, providing utilities for ranking matches.

  Additionally, ArrayMatcher and TrieMatcher have been replaced by createMultiMatcher (which is now always backed by trie-based matching).
  To match against only a single pattern while receiving type safe params from the match, use createMatcher.

  See the new README for details.

Patch Changes

• Encode href params so pathname params cannot inject URL path, dot segment, query, or hash syntax. Wildcard pathname params now preserve slash-separated structure while encoding each segment, and hostname params are normalized or rejected when they would inject URL authority, path, query, or hash syntax. Matchers now decode generated pathname params so reserved characters like /, ?, and # round-trip as param content.

• Do not allow partial matches for variables and wildcards in pathname

  let matcher = createMultiMatcher<string>()
  matcher.add('/files/:name.md', 'original')
  matcher.add('/files/:name.md.backup', 'backup')
  
  // before: 'original' included since `:name.md` partially matches `readme.md.backup`
  matcher.matchAll('https://example.com/files/readme.md.backup').map((match) => match.data)
  // ❌ ['backup', 'original']
  
  // after: only matches when the pattern covers the whole segment
  matcher.matchAll('https://example.com/files/readme.md.backup').map((match) => match.data)
  // ✅ ['backup']

• Fix matcher to match origin-less patterns (e.g. /, /about) against URLs that have an explicit port. Previously a pattern like / would not match http://localhost:44199/.