Oniguruma-To-ES
A lightweight Oniguruma to JavaScript RegExp transpiler that runs in the browser and on your server. Use it to:
- Take advantage of Oniguruma's extended regex capabilities in JavaScript.
- Run regexes intended for Oniguruma in JavaScript, such as those used in TextMate grammars (used by VS Code, Shiki syntax highlighter, etc.).
- Share regexes across your Ruby and JavaScript code.
Compared to running the actual Oniguruma C library in JavaScript via WASM bindings (e.g. via vscode-oniguruma), this library is much lighter weight and its regexes run much faster since they run as native JavaScript.
[!WARNING]
This library is currently in beta and has known bugs.
Oniguruma-To-ES deeply understands all of the hundreds of large and small differences in Oniguruma and JavaScript regex syntax and behavior across multiple JavaScript version targets. It's obsessive about precisely following Oniguruma syntax rules and ensuring that the emulated features it supports have exactly the same behavior, even in extreme edge cases. And it's battle-tested on thousands of real-world Oniguruma regexes used in TextMate grammars (via the Shiki library). A few uncommon features can't be perfectly emulated and allow rare differences, but if you don't want to allow this, you can set the accuracy
option to throw for such patterns (see details below).
π Contents
πΉοΈ Install and use
npm install oniguruma-to-es
import {toRegExp} from 'oniguruma-to-es';
In browsers:
<script type="module">
import {toRegExp} from 'https://esm.run/oniguruma-to-es';
toRegExp(String.raw`β¦`);
</script>
Using a global name (no import)
<script src="https://cdn.jsdelivr.net/npm/oniguruma-to-es/dist/index.min.js"></script>
<script>
const {toRegExp} = OnigurumaToES;
</script>
π API
toRegExp
Transpiles an Oniguruma pattern and returns a native JavaScript RegExp
.
[!TIP]
Try it in the demo REPL.
function toRegExp(
pattern: string,
options?: Options
): RegExp | EmulatedRegExp;
Type Options
type Options = {
accuracy?: 'strict' | 'default' | 'loose';
avoidSubclass?: boolean;
flags?: OnigurumaFlags;
global?: boolean;
hasIndices?: boolean;
maxRecursionDepth?: number | null;
target?: 'ES2018' | 'ES2024' | 'ESNext';
tmGrammar?: boolean;
verbose?: boolean;
};
See Options for more details.
toDetails
Transpiles an Oniguruma pattern to the parts needed to construct a native JavaScript RegExp
.
function toDetails(
pattern: string,
options?: Options
): {
pattern: string;
flags: string;
strategy?: {
name: string;
subpattern?: string;
};
};
The returned pattern
and flags
might be different than those provided, as a result of the emulation process. The returned pattern
, flags
, and strategy
can be provided as arguments to the EmulatedRegExp
constructor to produce the same result as toRegExp
.
If the only keys returned are pattern
and flags
, they can optionally be provided to JavaScript's RegExp
constructor instead. Setting option avoidSubclass
to true
ensures that this is always the case, and any patterns that are emulatable only via EmulatedRegExp
throw an error.
EmulatedRegExp
Can be provided results from toDetails
to produce the same result as toRegExp
.
class EmulatedRegExp extends RegExp {
constructor(
pattern: string | EmulatedRegExp,
flags?: string,
strategy?: {
name: string;
subpattern?: string;
}
);
};
toOnigurumaAst
Generates an Oniguruma AST from an Oniguruma pattern.
function toOnigurumaAst(
pattern: string,
options?: {
flags?: OnigurumaFlags;
}
): OnigurumaAst;
π© Options
These options are shared by functions toRegExp
and toDetails
.
accuracy
One of 'strict'
, 'default'
(default), or 'loose'
.
Sets the level of emulation rigor/strictness.
- Strict: Throw if the pattern can't be emulated with identical behavior (even in rare edge cases) for the given target.
- Default: The best choice in most cases. Permits a few close approximations of Oniguruma in order to support additional features.
- Loose: Useful for non-critical matching like syntax highlighting where having some mismatches is better than not working.
Each level of increased accuracy supports a subset of patterns supported by lower accuracies. If a given pattern doesn't produce an error for a particular accuracy, its generated result will be identical with all lower levels of accuracy (given the same target
).
More details
strict
Supports slightly fewer features, but the missing features are all relatively uncommon (see below).
default
Supports all features of strict
, plus the following additional features, depending on target
:
- All targets (
ESNext
and earlier):
- Enables use of
\X
using a close approximation of a Unicode extended grapheme cluster. - Enables recursion (e.g. via
\g<0>
) with a depth limit specified by option maxRecursionDepth
.
ES2024
and earlier:
- Enables use of case-insensitive backreferences to case-sensitive groups.
ES2018
:
- Enables use of POSIX classes
[:graph:]
and [:print:]
using ASCII-based versions rather than the Unicode versions available for ES2024
and later. Other POSIX classes are always based on Unicode.
loose
Supports all features of default
, plus the following:
- Silences errors for unsupported uses of the search-start anchor
\G
(a flexible assertion that doesnβt have a direct equivalent in JavaScript).
- Oniguruma-To-ES uses a variety of strategies to accurately emulate many common uses of
\G
. When using loose
accuracy, if a \G
assertion is found that doesn't have a known emulation strategy, the \G
is simply removed and JavaScript's y
(sticky
) flag is added. This might lead to some false positives and negatives.
avoidSubclass
Default: false
.
Disables advanced emulation strategies that rely on returning a RegExp
subclass, resulting in certain patterns not being emulatable.
flags
Oniguruma flags; a string with i
, m
, and x
in any order (all optional).
Flags can also be specified via modifiers in the pattern.
[!IMPORTANT]
Oniguruma and JavaScript both have an m
flag but with different meanings. Oniguruma's m
is equivalent to JavaScript's s
(dotAll
).
global
Default: false
.
Include JavaScript flag g
(global
) in the result.
hasIndices
Default: false
.
Include JavaScript flag d
(hasIndices
) in the result.
maxRecursionDepth
Default: 6
.
Specifies the recursion depth limit. Supported values are integers 2
to 100
and null
. If null
, any use of recursion results in an error.
Since recursion isn't infinite-depth like in Oniguruma, use of recursion also results in an error if using strict accuracy
.
More details
Using a high limit has a (usually tiny) impact on transpilation and regex performance. Generally, this is only a problem if the regex has an existing issue with runaway backtracking that recursion exacerbates.
Higher limits have no effect on regexes that don't use recursion, so you should feel free to increase this if helpful.
target
Default: 'ES2024'
.
Sets the JavaScript language version for the generated pattern and flags. Later targets allow faster processing, simpler generated source, and support for additional features.
More details
ES2018
: Uses JS flag u
.
- Emulation restrictions: Character class intersection, nested negated character classes, and Unicode properties added after ES2018 are not allowed.
- Generated regexes might use ES2018 features that require Node.js 10 or a browser version released during 2018 to 2023 (in Safari's case). Minimum requirement for any regex is Node.js 6 or a 2016-era browser.
ES2024
: Uses JS flag v
.
- No emulation restrictions.
- Generated regexes require Node.js 20 or a 2023-era browser (compat table).
ESNext
: Uses JS flag v
and allows use of flag groups and duplicate group names.
- Benefits: Faster transpilation, simpler generated source, and duplicate group names are preserved across separate alternation paths.
- Generated regexes might use features that require Node.js 23 or a 2024-era browser (except Safari, which lacks support).
tmGrammar
Default: false
.
Leave disabled unless the regex will be used in a TextMate grammar processor that merges backreferences across begin
and end
patterns.
verbose
Default: false
.
Disables optimizations that simplify the pattern when it doesn't change the meaning.
β
Supported features
Following are the supported features by target.
[!NOTE]
Targets ES2024
and ESNext
have the same emulation capabilities. Resulting regexes might have different source and flags, but they match the same strings.
Notice that nearly every feature below has at least subtle differences from JavaScript. Some features and subfeatures listed as unsupported are not emulatable using native JavaScript regexes, but support for others might be added in future versions of Oniguruma-To-ES. Unsupported features throw an error.
Feature | Example | ES2018 | ES2024+ | Subfeatures & JS differences |
---|
Flags | i | i | β
| β
|
β Unicode case folding (same as JS with flag u , v )
|
---|
m | m | β
| β
|
β Equivalent to JS flag s (dotAll )
|
x | x | β
| β
|
β Unicode whitespace ignored
β Line comments with #
β Whitespace/comments allowed between a token and its quantifier
β Whitespace/comments between a quantifier and the ? /+ that makes it lazy/possessive changes it to a chained quantifier
β Whitespace/comments separate tokens (ex: \1 0 )
β Whitespace and # not ignored in char classes
|
Flag modifiers | Group | (?im-x:β¦) | β
| β
|
β Unicode case folding for i
β Allows enabling and disabling the same flag (priority: disable)
β Allows lone or multiple -
|
---|
Directive | (?im-x) | β
| β
|
β Continues until end of pattern or group (spanning alternatives)
|
Characters | Literal | E , ! | β
| β
|
β Code point based matching (same as JS with flag u , v )
β Standalone ] , { , } don't require escaping
|
---|
Identity escape | \E , \! | β
| β
|
β Different allowed set than JS
β Allows multibyte chars
|
Escaped metachar | \\ , \. | β
| β
|
β Same as JS
|
Shorthand | \t | β
| β
|
β The JS set plus \a , \e
|
\xNN | \x7F | β
| β
|
β Allows 1 hex digit
β Above 7F , is UTF-8 encoded byte (unlike JS)
β Error for invalid encoded bytes
|
\uNNNN | \uFFFF | β
| β
|
β Same as JS with flag u , v
|
\x{β¦} | \x{A} | β
| β
|
β Allows leading 0s up to 8 total hex digits
|
Escaped num | \20 | β
| β
|
β Can be backref, error, null, octal, identity escape, or any of these combined with literal digits, based on complex rules that differ from JS
β Always handles escaped single digit 1-9 outside char class as backref
β Allows null with 1-3 0s
β Error for octal > 177
|
Control | \cA , \C-A | β
| β
|
β With A-Za-z (JS: only \c form)
|
Other (extremely rare) | β | β |
Not yet supported:
β Non-A-Za-z with \cx , \C-x
β Meta \M-x , \M-\C-x
β Octal code point \o{β¦}
β UTF-8 encoded bytes in octal
|
Character sets | Digit, word | \d , \w , etc. | β
| β
|
β Same as JS (ASCII)
|
---|
Hex digit | \h , \H | β
| β
|
β ASCII
|
Whitespace | \s , \S | β
| β
|
β ASCII (unlike JS)
|
Dot | . | β
| β
|
β Excludes only \n (unlike JS)
|
Any | \O | β
| β
|
β Any char (with any flags)
β Identity escape in char class
|
Not newline | \N | β
| β
|
β Identity escape in char class
|
Unicode property |
\p{L} ,
\P{L}
| β
[1] | β
|
β Binary properties
β Categories
β Scripts
β Aliases
β POSIX properties
β Invert with \p{^β¦} , \P{^β¦}
β Insignificant spaces, underscores, and casing in names
β \p , \P without { is an identity escape
β Error for key prefixes
β Error for props of strings
β Blocks (wontfix[2])
|
Variable-length sets | Newline | \R | β
| β
|
β Matched atomically
|
---|
Grapheme | \X | βοΈ | βοΈ |
β Uses a close approximation
β Matched atomically
|
Character classes | Base | [β¦] , [^β¦] | β
| β
|
β Unescaped - outside of range is literal in some contexts (different than JS rules in any mode)
β Fewer chars require escaping than JS
β Error for reversed range (same as JS)
|
---|
Empty | [] , [^] | β
| β
|
β Error
|
Range | [a-z] | β
| β
|
β Same as JS with flag u , v
|
POSIX class |
[[:word:]] ,
[[:^word:]]
| βοΈ[3] | β
|
β All use Unicode definitions
|
Nested class | [β¦[β¦]] | βοΈ[4] | β
|
β Same as JS with flag v
|
Intersection | [β¦&&β¦] | β | β
|
β Doesn't require nested classes for intersection of union and ranges
|
Assertions | Line start, end | ^ , $ | β
| β
|
β Always "multiline"
β Only \n as newline
|
---|
String start, end | \A , \z | β
| β
|
β Same as JS ^ $ without flag m
|
String end or before terminating newline | \Z | β
| β
|
β Only \n as newline
|
Search start | \G | βοΈ | βοΈ |
β Common uses supported
|
Word boundary | \b , \B | β
| β
|
β Unicode based (unlike JS)
|
Lookaround |
(?=β¦) ,
(?!β¦) ,
(?<=β¦) ,
(?<!β¦)
| β
| β
|
β Same as JS
β Allows variable-length quantifiers and alternation within lookbehind
|
Quantifiers | Greedy, lazy | * , +? , {2,} , etc. | β
| β
|
β Includes all JS forms
β Adds {,n} for min 0
β Explicit bounds have upper limit of 100,000 (unlimited in JS)
β Error with assertions (same as JS with flag u , v )
|
---|
Possessive | ?+ , *+ , ++ | β
| β
|
β + suffix doesn't make interval ({β¦} ) quantifiers possessive (creates a chained quantifier)
|
Chained | ** , ??+* , {2,3}+ , etc. | β
| β
|
β Further repeats the preceding repetition
|
Groups | Noncapturing | (?:β¦) | β
| β
|
β Same as JS
|
---|
Atomic | (?>β¦) | β
| β
|
β Supported
|
Capturing | (β¦) | β
| β
|
β Is noncapturing if named capture present
|
Named capturing |
(?<a>β¦) ,
(?'a'β¦)
| β
| β
|
β Duplicate names allowed (including within the same alternation path) unless directly referenced by a subroutine
β Error for names invalid in Oniguruma or JS
|
Backreferences | Numbered | \1 | β
| β
|
β Error if named capture used
β Refs the most recent of a capture/subroutine set
|
---|
Enclosed numbered, relative |
\k<1> ,
\k'1' ,
\k<-1> ,
\k'-1'
| β
| β
|
β Error if named capture used
β Allows leading 0s
β Refs the most recent of a capture/subroutine set
β \k without < ' is an identity escape
|
Named |
\k<a> ,
\k'a'
| β
| β
|
β For duplicate group names, rematch any of their matches (multiplex)
β Refs the most recent of a capture/subroutine set (no multiplex)
β Combination of multiplex and most recent of capture/subroutine set if duplicate name is indirectly created by a subroutine
|
To nonparticipating groups | βοΈ | βοΈ |
β Error if group to the right[5]
β Duplicate names (and subroutines) to the right not included in multiplex
β Fail to match (or don't include in multiplex) ancestor groups and groups in preceding alternation paths
β Some rare cases are indeterminable at compile time and use the JS behavior of matching an empty string
|
Subroutines | Numbered, relative |
\g<1> ,
\g'1' ,
\g<-1> ,
\g'-1' ,
\g<+1> ,
\g'+1'
| β
| β
|
β Allowed before reffed group
β Can be nested (any depth)
β Doesn't alter backref nums
β Reuses flags from the reffed group (ignores local flags)
β Replaces most recent captured values (for backrefs)
β \g without < ' is an identity escape
β Error if named capture used
|
---|
Named |
\g<a> ,
\g'a'
| β
| β
|
β Same behavior as numbered
β Error if reffed group uses duplicate name
|
Recursion | Full pattern |
\g<0> ,
\g'0'
| βοΈ | βοΈ |
β Has depth limit[6]
|
---|
Named, numbered, relative |
(?<a>β¦\g<a>?β¦) ,
(β¦\g<1>?β¦) ,
(β¦\g<-1>?β¦) , etc.
| βοΈ | βοΈ |
β Has depth limit[6]
|
Other | Comment group | (?#β¦) | β
| β
|
β Allows escaping \) , \\
β Comments allowed between a token and its quantifier
β Comments between a quantifier and the ? /+ that makes it lazy/possessive changes it to a chained quantifier
|
---|
Alternation | β¦|β¦ | β
| β
|
β Same as JS
|
Keep | \K | βοΈ | βοΈ |
β Supported if at top level and no top-level alternation is used
|
Absence operator | (?~β¦) | β | β |
β Some forms are supportable
|
Conditional | (?(1)β¦) | β | β |
β Some forms are supportable
|
Char sequence |
\x{1 2 β¦N} ,
\o{1 2 β¦N}
| β | β |
β Not yet supported
|
JS features unknown to Oniguruma are handled using Oniguruma syntax | β
| β
|
β \u{β¦} is an error
β [\q{β¦}] matches q , etc.
β [a--b] includes the invalid reversed range a to -
|
Invalid Oniguruma syntax | β
| β
|
β Error
|
The table above doesn't include all aspects that Oniguruma-To-ES emulates (including error handling, most aspects that work the same as in JavaScript, and many aspects of non-JavaScript features that work the same in the other regex flavors that support them).
- Target
ES2018
doesn't allow Unicode property names added in JavaScript specifications after ES2018 to be used. - Unicode blocks are easily emulatable but their character data would significantly increase library weight. They're also a deeply flawed and arguably-unuseful feature, given the ability to use Unicode scripts and other properties.
- With target
ES2018
, the specific POSIX classes [:graph:]
and [:print:]
use ASCII-based versions rather than the Unicode versions available for target ES2024
and later, and they result in an error if using strict accuracy
. - Target
ES2018
doesn't support nested negated character classes. - It's not an error for numbered backreferences to come before their referenced group in Oniguruma, but an error is the best path for Oniguruma-To-ES because (1) most placements are mistakes and can never match (based on the Oniguruma behavior for backreferences to nonparticipating groups), (2) erroring matches the behavior of named backreferences, and (3) the edge cases where they're matchable rely on rules for backreference resetting within quantified groups that are different in JavaScript and aren't emulatable. Note that it's not a backreference in the first place if using
\10
or higher and not as many capturing groups are defined to the left (it's an octal or identity escape). - The recursion depth limit is specified by option
maxRecursionDepth
. Some forms of recursion (multiple recursions in the same pattern, and recursion with backreferences) aren't yet supported. Patterns that would error in Oniguruma due to triggering infinite recursion might find a match in Oniguruma-To-ES since recursion is bounded (future versions will detect this and error at transpilation time).
γοΈ Unicode / mixed case-sensitivity
Oniguruma-To-ES fully supports mixed case-sensitivity (and handles the Unicode edge cases) regardless of JavaScript target. It also restricts Unicode properties to those supported by Oniguruma and the target JavaScript version.
Oniguruma-To-ES focuses on being lightweight to make it better for use in browsers. This is partly achieved by not including heavyweight Unicode character data, which imposes a couple of minor/rare restrictions:
- Character class intersection and nested negated character classes are unsupported with target
ES2018
. Use target ES2024
or later if you need support for these Oniguruma features. - With targets before
ESNext
, a handful of Unicode properties that target a specific character case (ex: \p{Lower}
) can't be used case-insensitively in patterns that contain other characters with a specific case that are used case-sensitively.
- In other words, almost every usage is fine, including
A\p{Lower}
, (?i:A\p{Lower})
, (?i:A)\p{Lower}
, (?i:A(?-i:\p{Lower}))
, and \w(?i:\p{Lower})
, but not A(?i:\p{Lower})
. - Using these properties case-insensitively is basically never done intentionally, so you're unlikely to encounter this error unless it's catching a mistake.
π Similar projects
JsRegex transpiles Onigmo regexes to JavaScript (Onigmo is a fork of Oniguruma with mostly shared syntax and behavior). It's written in Ruby and relies on the Regexp::Parser Ruby gem, which means regexes must be pre-transpiled on the server to use them in JavaScript. Note that JsRegex doesn't always translate edge case behavior differences.
π·οΈ About
Oniguruma-To-ES was created by Steven Levithan.
If you want to support this project, I'd love your help by contributing improvements, sharing it with others, or sponsoring ongoing development.
Β© 2024βpresent. MIT License.