Changes:

• Avoid fall through in the keyword transform, thanks to @lrowe in #164
• Replace the keyword transform with a Map, to avoid triggering unsafe-eval Content-Security-Policy warnings, thanks to @jsharkey13 in #173
• Fix a bug in reset() relating to queuedText, thanks to @hikerpig in #169
• Improve error reporting and add additional context lines, thanks to @airportyh in #129

Documentation tweaks:

• Correct states example to use pop: 1, thanks to @evanpurkhiser in #140
• Replace itt example with modern js example, thanks to @AlansCodeLog in #147

Full Changelog: v0.5.1...v0.5.2

Published v0.5.1 to NPM.

• #116, #125 Allow Unicode flag if all RegExps use it
• #114 Allow calling formatError without a token
• 13e1157 Fix a bug with fast matching when include is used

Version 0.5.0 is now published to NPM.

Highlights:

• Add keywords function
• Add fallback tokens
• Add inheritance for stateful lexers
• Add optimization for single-character tokens
• Let type be a function
• Fix: save and restore state stack
• Rename name -> type when specifying rules in an array
• Deprecate Lexer.has()

Breaking changes:

• Keyword support is now explicit.
• Capture groups are no longer supported. Use slice() yourself, or use the value transform.
• {pop: 1} will now throw an error; use {pop: true}.
• The size property has been removed. Use text.length instead.

• Prettier syntax errors
• Remove feed()
• Include state in save()
• reset() now goes back to start state

I just published v0.2 to NPM:

• Don't re-compile RegExp in clone()
• Remove index() & remaining()