package quickjs

Changes

Unreleased

0.6.0

• Update QuickJS-NG to c2f37ec, adopting its register-based regexp engine and latest memory-safety fixes
• Add regexp modifiers, duplicate named capture groups, full Unicode sets, and Unicode properties of strings

0.5.1

• Add RegExp.prepare_input and exec_prepared for reusing one matching buffer and UTF-16 map across repeated matches
• Add prepared-input source range, substring and AdvanceStringIndex helpers
• Fix regexp captures whose UTF-16 range splits a surrogate pair: the unpaired half is represented as U+FFFD instead of slicing the wrong UTF-8 bytes
• Fix Unicode global and sticky regexps starting from the trailing half of a surrogate pair instead of the pair's start

0.5.0

• Breaking: RegExp.exec returns match_result option: None on no match, and a record with captures, index, input, groups and indices on match
• Breaking: RegExp captures are string option array: a group that didn't participate in the match is None, like undefined in JavaScript
• Breaking: RegExp's index, last_index and set_last_index are UTF-16 code units (like JavaScript), not byte offsets
• Breaking: RegExp.compile rejects invalid flags (unknown, duplicated, or u with v) instead of ignoring them. Adds support for the d and v flags
• Breaking: String.Prototype.match_ and match_flags return RegExp.match_result option; match_all returns RegExp.match_result list; match_global accepts ?flags
• Breaking: String.Prototype regex-based methods raise Invalid_argument on invalid patterns instead of silently returning "no match"
• Breaking: String.Prototype.split_regex returns string option array: a capture group that did not participate in the match splices None (JavaScript's undefined) instead of ""
• Breaking: Global.parse_int auto-detects the radix like JavaScript (parse_int "0x10" is Some 16) and returns None for values that don't fit in an OCaml int
• Breaking: rename Global.js_parse_options to Global.js_number_options
• Breaking: quickjs.c and quickjs.bindings are wrapped: Libregexp, Atod, Dtoa, Cutils and Bindings become Quickjs_c.* and Quickjs_bindings
• Fix RegExp.last_index tracking: no more wrong values, overlapping matches or infinite loops when patterns have capture groups
• Fix RegExp.exec updating last_index on non-global, non-sticky regexps
• Fix memory leaks: compiled regexps and character classes no longer leak
• Fix crash when regexp patterns are nested too deeply: returns Error `Stack_overflow
• Fix RegExp.flags reporting a u flag the user never passed for non-ASCII patterns
• Fix $` in String.Prototype.replace_all/replace_regex_global: it now refers to the original string
• Fix String.Prototype.split_regex to match JavaScript on empty matches and empty input
• Fix String.Prototype.split_limit to coerce its limit with ToUint32 like JavaScript: negative limits wrap to huge values and behave as "no limit" (previously they returned [||])
• Fix String.Prototype.slice/substring/substr raising Invalid_argument when their range splits a surrogate pair; the lone half becomes U+FFFD (JavaScript keeps the lone surrogate, which UTF-8 cannot represent)
• Fix String.Prototype.trim/trim_start/trim_end to trim all JavaScript whitespace (was missing U+1680)
• Fix String.Prototype.to_lower_case for the Greek final sigma: "ΑΣ" is "ας"
• Fix String.Prototype.starts_with_from raising on huge positions
• Fix Global.parse_float_partial: it returned None for virtually every input
• Fix Global.parse_float "never" returning Some nan; unparsable input is always None
• Fix invalid Number formatting options aborting the process: they raise Invalid_argument
• Global.parse_float/parse_float_partial/parse_int skip leading whitespace like JavaScript
• Number.Prototype.to_fixed/to_precision/to_exponential use JavaScript's 0-100 digit ranges
• Add RegExp match indices (the d flag, ES2022): match_result.indices carries the UTF-16 [start, end) range of every capture group (None without the flag, like JavaScript's hasIndices), plus group_indices for named groups, like JavaScript's match.indices.groups
• Add ?timeout_ms to RegExp.exec/test to bound catastrophic backtracking (raises RegExp.Timeout)
• Add String.from_char_code and String.from_code_point, JavaScript's String.fromCharCode()/String.fromCodePoint(): ToUint16 coercion, surrogate-pair composition and the RangeError contract included; unpaired surrogates become U+FFFD since UTF-8 cannot represent them
• Add String.Prototype.split_regex_limit, JavaScript's split(regexp, limit): spliced capture groups count toward the limit
• Add String.utf16_index_of_byte/byte_index_of_utf16 to convert between byte offsets and UTF-16 indices
• Add Global.parse_int_float: parseInt returning a float, exactly like JavaScript for values of any size
• Add Unicode.fold_case/fold_case_char: Unicode full case folding for caseless comparison, e.g. fold_case "Straße" = "strasse". Folds to a fixpoint, covering the capitals QuickJS's table folds one step short (fold_case "ẞ" is "ss", not "ß")
• Add Unicode.script, general_category and binary_property character-set lookups with the Unicode.CharSet module: the Unicode property tables behind JavaScript's \p{...} escapes, e.g. Unicode.script "Greek"
• Add Quickjs_c.Libunicode.char_range/char_range_free: raw bindings to libunicode's character range tables (unicode_script, unicode_general_category, unicode_prop)

0.4.2

• Rename RegExp.lastIndex -> RegExp.last_index
• Rename RegExp.setLastIndex -> RegExp.set_last_index

0.4.1

• Remove uutf

0.4.0

• Breaking: Migrated from bellard/quickjs to quickjs-ng/quickjs - the actively maintained community fork with Unicode 17.0.0 support, better spec compliance, and ongoing development
• Fix Global.parse_float for incomplete exponents ("1e", "1e+", "1e-") to return the parsed number instead of None, matching JavaScript spec. Workaround for quickjs-ng/quickjs#1259

0.3.0

• Fix openSUSE

• Breaking: Reorganized API to mirror JavaScript built-in objects for quickjs

  • Dtoa → Number.Prototype (toString, toFixed, toPrecision, toExponential, toRadix)
  • Atod → Global (parse_float, parse_float_partial)
  • IntToString → Number (of_int, of_int32, of_int64, of_int_radix, etc.)
  • New String.Prototype module (to_lower_case, to_upper_case, normalize)
  • New Unicode module for character-level operations (is_cased, is_id_start, canonicalize, etc.)
• Fix String.Prototype.lastIndexOf
• Fix String.Prototype.match

0.2.0

• Support unicode in RegExp
• Support named groups in RegExp
• Bind to dtoa (Dtoa, Atod, IntToString)

0.1.2

• Support for named groups
• Change RegExp.exec interface to return result

0.1.1

Bug fix for RegExp backtracking when not find a match

0.1.0

Initial release of quickjs. This release only includes bindings to libregexp and exposes the API of the QuickJS C library with the same shape as the JavaScript API: RegExp.