All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog and this project adheres to Semantic Versioning.
- nicer set rendering with the upcoming
character_set1.8.0 (cosmetic fix)
- more lenient casting for the
target:argument
- regression in 3.10.0 that caused the loss of some escapes in character sets
- e.g.
[\^\-\.]was converted to[^-.]; fixed to convert to[\^\-.] - thanks to https://github.com/slfrnkln for the report
- e.g.
JsRegex::compatible?for checking if a Regexp can be converted- improved performance, particularly when dealing with character sets (
[...])
- limited support for whole pattern recursion (
\g<0>)
- SystemStackError with recursive group calls (
(a|b\g<1>)c)
- added an optional
target:argumenttarget: 'ES2009'- is the default
- does not make any changes from previous versions of
js_regex
target: 'ES2015'- is compatible with all modern browsers (not with IE11 or below)
- uses the
uflag in some cases to avoid lengthy escape sequences
target: 'ES2018'- is compatible with most modern browsers, but only partially with Safari
- supports lookbehinds (will cause JS errors in Safari, see README!)
- supports keep marks (
\K- again not for Safari) - adds better word-boundary anchor conversions (
\b,\B- again not for Safari) - brings more concise representations of unicode properties, posix classes, etc.
- preserves capturing group and backreference names (
(?<x>.),\k<x>)
- support for forcing the 's' flag via the
optionsargument - fixed handling of astral plane literals with a codepoint above
0xFFFFF(basically just private use chars)
- fixed handling of
{,m}quantifiers; thanks to https://github.com/serch - fixed
NoMethodErrorwhen using possessive quantifiers together with conditionals
- fixed error when using octal escapes (e.g.
\0) in JS strict mode by converting them to hex escapes instead of passing through
- added
JsRegex::Erroras a mixin to all errors (mostly parser errors) to simplify catching them - added
JsRegex::new!as a strict method that will raise if the input can not be fully converted
- added support for chained/multiplicative quantifiers (
A{4}{6}) by bumpingregexp_parserdependency
- fixed possible error with absence groups (
(?~) on Ruby 3.x by tighteningregexp_parserdependency
- improved '.' substitution so that it matches "\r" and astral plane chars like in Ruby
- fixed loss of backslashes in some edge cases when passed a
Stringsource (as possible since v3.4.0) - escaping of operational literal whitespace in x-mode is no longer carried over (cosmetic improvement)
- added full support and more concise representations for astral plane sets and properties via
character_set; thanks to https://github.com/singpolyma for the suggestion JsRegex::newcan now also be called with a sourceStringinstead of aRegexp, e.g.JsRegex.new('\h')
- '/' is now correctly escaped if it occurs within sets (this only caused issues with
JsRegex#to_soutput); thanks to https://github.com/singpolyma for the report
- added support for local encoding options ("(?u:\w)", "(?a)[[:word:]]")
- added support for absence groups ("(?~foo)")
- fixed
#to_soutput being syntactically invalid in JS when converting an empty Regexp - fixed missing codepoints in "\R" replacement (vertical tab, form feed, nextline)
- relaxed unnecessarily strict dependency on character_set 1.0.x
- added support for conditionals ("(a)?(?(1)b|c)")
- added support for forward-referring subexpression calls ("\g<+2>")
- empty alternation branches ("(a|)") are no longer removed since the potential for a zero-width match might be intended
- fixed conversion of astral plane literals with lengths greater than one
- fixed handling of locally case-insensitive literals (following e.g. "(?i)") with lengths greater than one
- fixed a rare bug where backreferences, atomic groups and subexpression calls weren't handled correctly if preceded by either the linebreak type ("\R") or an option group with children ("(?i:...)")
Major refactoring adding character_set and regexp_property_values as dependencies.
- changed default options: the
gflag is no longer automatically set, useJsRegex.new(//, options: 'g')to force it (#5) - changed handling of unicode properties (\p{...}) and posix classes ([:...:]); their replacements now match the same codepoints as the host ruby version
- changed required ruby version from 1.9.3 to 2.1.0
- added support for moderately sized astral plane properties (e.g. \p{emoji})
- added full support for set nesting (e.g. "[a[^123]]")
- added full support for set ranges (including non-literal, e.g. "[\x00-\b]")
- added support for astral plane set members
- added support for set intersections (e.g. "[a-z&&[^aeiou]]")
- added support for set intersections (e.g. "[a-z&&[^aeiou]]")
- added support for all types of subexpression calls (e.g. "(foo)\g<1>")
- added support for bell and escape shortcuts ("\a", "\e")
- added warnings for ranges and unicode properties with large astral plane parts
- fixed handling of escaped alternation chars a.k.a. pipes ("\|")
- fixed errors on new Ruby versions by upgrading regexp_parser dependency
- fixed handling of superfluous forward slash escapes; thanks to https://github.com/JasonBarnabe for the cue
- fixed quantification of astral plane literals by wrapping their surrogate pair substitution in a passive group
- added support for the most recent unicode age properties (BMP parts only)
- fixed handling of \p{Pc}, \p{Nl} and several other, older unicode properties; thanks to https://github.com/mojavelinux
- added support for option groups and switches ("(?-m:.)", "(?i)a", etc.)
- added warning for unsupported encoding options ("(?d:\w)", "(?u)", etc.)
- fixed handling of whitespace following x-switches ("(?x)", "(?-x)")
Major refactoring. Using Regexp::Parser instead of Regexp::Scanner internally allows for higher-level conversions.
- word boundaries \b and \B now cause warnings (see README for details)
- added support for possessive quantifiers ("++", "*+", etc.)
- added support for backreferences ("\2") following atomic groups
- added support for \k-style numeric backreferences ("\k'2'")
- added support for relative backreferences ("\k'-1'")
- added support for named backreferences ("\k'foo'")
- added support for the generic linebreak type \R
- added support for control and meta escapes ("\cX", "\C-X", "\M-X", "\M-\C-X", etc.)
- when dropping unsupported expressions, their quantifiers are now dropped as well
- fixed handling of hex types and backspace ("\h", "\H", "\b") within negated sets
- double-negated properties ("\P{^...}") are now correctly treated as positive
- conditionals are now replaced with passive instead of capturing alternation groups
- fixed handling of escaped parentheses ("\(" and "\)"); thanks to https://github.com/owst
- fixed handling of codepoint lists (e.g. \u{61 63 1F601}}.
- added support for hex escapes in character sets.
- fixed an exotic bug with multiple stacked quantifiers (e.g. /a{2}{3}{4}/). they are now all removed instead of breaking the regex.
- added support for astral plane literals outside of sets by converting them to surrogate pairs
- added support for the backspace pseudo set ([\b])
- added support for surrogate properties, (e.g. \p{surrogate})
- fixed a bug with atomic groups nested in atomic groups. they are now made non-atomic with a warning instead of breaking the regex.
- fixed a bug with negated properties (\p{^...} and \P{...}) in sets. they are now correctly extracted from sets.
- fixed handling of escaped bol/eol anchors (^ and $); thanks to https://github.com/tomca32
- fixed handling of carriage return escape sequences ("\r"); thanks to https://github.com/mrcthms