Strict mode

Added

• Adds a strict option to detect potential ReDOS issues

Fixed

• Fixes separator to default to suffix + prefix when not specified
• Allows separator to be undefined in TokenData
  • This is only relevant if you are building TokenData manually, previously parse filled it in automatically

Comments

• I highly recommend enabling strict: true and I'm probably releasing a V8 with it enabled by default ASAP as a necessary security mitigation

v7.0.0...v7.1.0

Wildcard, unicode, and modifier changes

Hi all! There's a few major breaking changes in this release so read carefully.

Breaking changes:

• The function returned by compile only accepts strings as values (i.e. no numbers, use String(value) before compiling a path)
  • For repeated values, when encode !== false, it must be an array of strings
• Parameter names can contain all unicode identifier characters (defined as regex \p{XID_Continue}).
• Modifiers (?, *, +) must be used after a param explicitly wrapped in {}
  • No more implied prefix of / or .
• No support for arrays or regexes as inputs
• The wildcard (standalone *) has been added back and matches Express.js expected behavior
• Removed endsWith option
• Renamed strict: true to trailing: false
• Reserved ;, ,, !, and @ for future use-cases
• Removed tokensToRegexp, tokensToFunction and regexpToFunction in favor of simplifying exports
• Enable a "loose" mode by default, so / can be repeated multiple times in a matched path (i.e. /foo works like //foo, etc)
• encode and decode no longer receive the token as the second parameter
• Removed the ESM + CommonJS dual package in favor of only one CommonJS supported export
• Minimum JS support for ES2020 (previous ES2015)
• Encode defaults to encodeURIComponent and decode defaults to decodeURIComponent

Added:

• Adds encodePath to fix an issue around encode being used for both path and parameters (the path and parameter should be encoded slightly differently)
• Adds loose as an option to support arbitrarily matching the delimiter in paths, e.g. foo/bar and foo///bar should work the same
• Allow encode and decode to be set to false which skips all processing of the parameters input/output
• All remaining methods support TokenData (exported, returned by parse) as input
  • This should be useful if you are programmatically building paths to match or want to avoid parsing multiple times

Requests for feedback:

• Requiring {} is an obvious drawback but I'm seeking feedback on whether it helps make path behavior clearer
  • Related: Removing / and . as implicit prefixes
• Removing array and regex support is to reduce the overall package size for things many users don't need
• Unicode IDs are added to align more closely with browser URLPattern behavior, which uses JS identifiers

v6.2.2...v7.0.0

Updated README

No API changes. Documentation only release.

Changed

• Fix readme example c7ec332
• Update shield URL e828000

v6.2.1...v6.2.2

Support named matching groups in `RegExp`

Added

• Add support for named matching groups (#301) 114f62d

v0.1.7...v0.1.8

Fix matching `:name*` parameter

Fixed

• Fix invalid matching of :name* parameter (#261) 762bc6b
• Compare delimiter string over regexp 86baef8

Added

• New example in documentation (#256) ae9e576
• Update demo link (#250) 77df638
• Update README encode example b39edd4

v6.2.0...v6.2.1

Named Capturing Groups

Added

• Support named capturing groups for RegExps (#225)

Fixed

• Update strict flag documentation (#227)
• Ignore test files when bundling (#220)

Use `/#?` as Default Delimiter

Fixed

• Use /#? as default delimiter to avoid matching on query or fragment parameters
  • If you are matching non-paths (e.g. hostnames), you can adjust delimiter: '.'

Custom Prefix and Suffix Groups

This release reverts the prefix behavior added in v3 back to the behavior seen in v2. For the most part, path matching is backward compatible with v2 with these enhancements:

1. Support for nested non-capturing groups in regexp, e.g. /(abc(?=d))
2. Support for custom prefix and suffix groups using /{abc(.*)def}
3. Tokens in an unexpected position will throw an error
   • Paths like /test(foo previously worked treating ( as a literal character, now it expects ( to be closed and is treated as a group
   • You can escape the character for the previous behavior, e.g. /test\(foo

Changed

• Revert using any character as prefix, support prefixes option to configure this (starts as /. which acts like every version since 0.x again)
• Add support for {} to capture prefix/suffix explicitly, enables custom use-cases like /:attr1{-:attr2}?

Remove Default Encode URI Component

No changes to path rules since 3.x, except support for nested RegEx parts in 4.x.

Changed

• Rename RegexpOptions interface to TokensToRegexpOptions
• Remove normalizePathname from library, document solution in README
• Encode using identity function as default, not encodeURIComponent

Decode URI

Removed

• Remove whitelist in favor of decodeURI (advanced behavior can happen outside path-to-regexp)