regexr 0.2.2

Fixed

• A malformed escape at the start of a pattern is rejected instead of compiling to the empty pattern, which matched every input. \e, \a, \q, \x, \u{, \p{ and a lone \ all reported success and matched everywhere; only the first token was affected, so the same escape one character later already errored.
• Inline flag negation works: (?-i), (?im-sx) and (?-x:…) were rejected as invalid groups because - never reached the flag parser.
• A mid-pattern (?flags) applies to what follows it rather than to the whole pattern. The change previously reached only the AST's single global flag set, so (?i)ab(?-i)CD matched case-sensitively throughout.
• An escape that is invalid inside a character class names itself: [\b] reported invalid escape sequence '\?'.

Added

• Extended mode (x) is implemented. Unescaped ASCII whitespace and # comments outside a character class are no longer part of the pattern; previously the flag was accepted and ignored.
• \X matches one extended grapheme cluster, following UAX #29 in full — Hangul syllables, emoji ZWJ sequences, regional-indicator flag pairs and Indic conjuncts included. Checked against the complete Unicode grapheme-break test suite.
• \Q…\E quotes everything between it literally, including metacharacters, extended-mode whitespace and character-class syntax. An unterminated \Q runs to the end of the pattern and a stray \E is ignored, as in PCRE.
• \a (alert, U+0007) and \e (escape, U+001B).
• \ followed by any non-alphanumeric ASCII character is that character, matching PCRE, Perl, Java and the regex crate. \ , \", \@ and \# previously failed to compile.

Changed

• escape also escapes # and ASCII whitespace, so its output is safe to splice into an extended-mode pattern.
• Parser::new returns Result<Self>, and Lexer::read_ident is replaced by the infallible Lexer::read_ident_rest. Both are internal parsing types exposed through regexr::parser; Regex and RegexBuilder are unaffected.

Full diff: v0.2.1...v0.2.2