regexp-tree

Regular expressions processor in JavaScript

TL;DR: RegExp Tree is a regular expressions processor, which includes parser, traversal, transformer, optimizer, and interpreter APIs.

You can get an overview of the tool in this article.

Table of Contents

• Installation
• Development
• Usage as a CLI
• Usage from Node
• Capturing locations
• Parsing options
• Using traversal API
• Using transform API
  • Transform plugins
• Using generator API
• Using optimizer API
  • Optimizer ESLint plugin
• Using compat-transpiler API
  • Compat-transpiler Babel plugin
• RegExp extensions
  • RegExp extensions Babel plugin
• Creating RegExp objects
• Executing regexes
• Using interpreter API
  • Printing NFA/DFA tables
• AST nodes specification

Installation

The parser can be installed as an npm module:

npm install -g regexp-tree

You can also try it online using AST Explorer.

Development

1. Fork https://github.com/DmitrySoshnikov/regexp-tree repo
2. If there is an actual issue from the issues list you'd like to work on, feel free to assign it yourself, or comment on it to avoid collisions (open a new issue if needed)
3. Make your changes
4. Make sure npm test still passes (add new tests if needed)
5. Submit a PR

The regexp-tree parser is implemented as an automatic LR parser using Syntax tool. The parser module is generated from the regexp grammar, which is based on the regular expressions grammar used in ECMAScript.

For development from the github repository, run build command to generate the parser module, and transpile JS code:

git clone https://github.com/<your-github-account>/regexp-tree.git
cd regexp-tree
npm install
npm run build

NOTE: JS code transpilation is used to support older versions of Node. For faster development cycle you can use npm run watch command, which continuously transpiles JS code.

Usage as a CLI

Note: the CLI is exposed as its own regexp-tree-cli module.

Check the options available from CLI:

regexp-tree-cli --help

Usage: regexp-tree-cli [options]

Options:
   -e, --expression   A regular expression to be parsed
   -l, --loc          Whether to capture AST node locations
   -o, --optimize     Applies optimizer on the passed expression
   -c, --compat       Applies compat-transpiler on the passed expression
   -t, --table        Print NFA/DFA transition tables (nfa/dfa/all)

To parse a regular expression, pass -e option:

regexp-tree-cli -e '/a|b/i'

Which produces an AST node corresponding to this regular expression:

1{
2  type: 'RegExp',
3  body: {
4    type: 'Disjunction',
5    left: {
6      type: 'Char',
7      value: 'a',
8      symbol: 'a',
9      kind: 'simple',
10      codePoint: 97
11    },
12    right: {
13      type: 'Char',
14      value: 'b',
15      symbol: 'b',
16      kind: 'simple',
17      codePoint: 98
18    }
19  },
20  flags: 'i',
21}

NOTE: the format of a regexp is / Body / OptionalFlags.

Usage from Node

The parser can also be used as a Node module:

1const regexpTree = require('regexp-tree');
2
3console.log(regexpTree.parse(/a|b/i)); // RegExp AST

Note, regexp-tree supports parsing regexes from strings, and also from actual RegExp objects (in general -- from any object which can be coerced to a string). If some feature is not implemented yet in an actual JavaScript RegExp, it should be passed as a string:

1// Pass an actual JS RegExp object.
2regexpTree.parse(/a|b/i);
3
4// Pass a string, since `s` flag may not be supported in older versions.
5regexpTree.parse('/./s');

Also note, that in string-mode, escaping is done using two slashes \\ per JavaScript:

1// As an actual regexp.
2regexpTree.parse(/\n/);
3
4// As a string.
5regexpTree.parse('/\\n/');

Capturing locations

For source code transformation tools it might be useful also to capture locations of the AST nodes. From the command line it's controlled via the -l option:

regexp-tree-cli -e '/ab/' -l

This attaches loc object to each AST node:

1{
2  type: 'RegExp',
3  body: {
4    type: 'Alternative',
5    expressions: [
6      {
7        type: 'Char',
8        value: 'a',
9        symbol: 'a',
10        kind: 'simple',
11        codePoint: 97,
12        loc: {
13          start: {
14            line: 1,
15            column: 1,
16            offset: 1,
17          },
18          end: {
19            line: 1,
20            column: 2,
21            offset: 2,
22          },
23        }
24      },
25      {
26        type: 'Char',
27        value: 'b',
28        symbol: 'b',
29        kind: 'simple',
30        codePoint: 98,
31        loc: {
32          start: {
33            line: 1,
34            column: 2,
35            offset: 2,
36          },
37          end: {
38            line: 1,
39            column: 3,
40            offset: 3,
41          },
42        }
43      }
44    ],
45    loc: {
46      start: {
47        line: 1,
48        column: 1,
49        offset: 1,
50      },
51      end: {
52        line: 1,
53        column: 3,
54        offset: 3,
55      },
56    }
57  },
58  flags: '',
59  loc: {
60    start: {
61      line: 1,
62      column: 0,
63      offset: 0,
64    },
65    end: {
66      line: 1,
67      column: 4,
68      offset: 4,
69    },
70  }
71}

From Node it's controlled via setOptions method exposed on the parser:

1const regexpTree = require('regexp-tree');
2
3const parsed = regexpTree
4  .parser
5  .setOptions({captureLocations: true})
6  .parse(/a|b/);

The setOptions method sets global options, which are preserved between calls. It is also possible to provide options per a single parse call, which might be more preferred:

1const regexpTree = require('regexp-tree');
2
3const parsed = regexpTree.parse(/a|b/, {
4  captureLocations: true,
5});

Parsing options

The parser supports several options which can be set globally via the setOptions method on the parser, or by passing them with each parse method invocation.

Example:

1const regexpTree = require('regexp-tree');
2
3const parsed = regexpTree.parse(/a|b/, {
4  allowGroupNameDuplicates: true,
5});

The following options are supported:

• captureLocations: boolean -- whether to capture AST node locations (false by default)
• allowGroupNameDuplicates: boolean -- whether to skip duplicates check of the named capturing groups

Set allowGroupNameDuplicates would make the following expression possible:

1/
2  # YYY-MM-DD date format:
3
4  (?<year>  \d{4}) -
5  (?<month> \d{2}) -
6  (?<day>   \d{2})
7
8  |
9
10  # DD.MM.YYY date format
11
12  (?<day>   \d{2}) .
13  (?<month> \d{2}) .
14  (?<year>  \d{4})
15
16/x

Using traversal API

The traverse module allows handling needed AST nodes using the visitor pattern. In Node the module is exposed as the regexpTree.traverse method. Handlers receive an instance of the NodePath class, which encapsulates node itself, its parent node, property, and index (in case the node is part of a collection).

Visiting a node follows this algorithm:

• call pre handler.
• recurse into node's children.
• call post handler.

For each node type of interest, you can provide either:

• a function (pre).
• an object with members pre and post.

You can also provide a * handler which will be executed on every node.

Example:

1const regexpTree = require('regexp-tree');
2
3// Get AST.
4const ast = regexpTree.parse('/[a-z]{1,}/');
5
6// Traverse AST nodes.
7regexpTree.traverse(ast, {
8
9  // Visit every node before any type-specific handlers.
10  '*': function({node}) {
11    ...
12  },
13
14  // Handle "Quantifier" node type.
15  Quantifier({node}) {
16    ...
17  },
18
19  // Handle "Char" node type, before and after.
20  Char: {
21    pre({node}) {
22      ...
23    },
24    post({node}) {
25      ...
26    }
27  }
28
29});
30
31// Generate the regexp.
32const re = regexpTree.generate(ast);
33
34console.log(re); // '/[a-z]+/'

Using transform API

NOTE: you can play with transformation APIs, and write actual transforms for quick tests in AST Explorer. See this example.

While traverse module provides basic traversal API, which can be used for any purposes of AST handling, transform module focuses mainly on transformation of regular expressions.

It accepts a regular expressions in different formats (string, an actual RegExp object, or an AST), applies a set of transformations, and retuns an instance of TransformResult. Handles receive as a parameter the same NodePath object used in traverse.

Example:

1const regexpTree = require('regexp-tree');
2
3// Handle nodes.
4const re = regexpTree.transform('/[a-z]{1,}/i', {
5
6  /**
7   * Handle "Quantifier" node type,
8   * transforming `{1,}` quantifier to `+`.
9   */
10  Quantifier(path) {
11    const {node} = path;
12
13    // {1,} -> +
14    if (
15      node.kind === 'Range' &&
16      node.from === 1 &&
17      !node.to
18    ) {
19      path.replace({
20        type: 'Quantifier',
21        kind: '+',
22        greedy: node.greedy,
23      });
24    }
25  },
26});
27
28console.log(re.toString()); // '/[a-z]+/i'
29console.log(re.toRegExp()); // /[a-z]+/i
30console.log(re.getAST()); // AST for /[a-z]+/i

Transform plugins

A transformation plugin is a module which exports a transformation handler. We have seen above how we can pass a handler object directly to the regexpTree.transform method, here we extract it into a separate module, so it can be implemented and shared independently:

Example of a plugin:

1// file: ./regexp-tree-a-to-b-transform.js
2
3
4/**
5 * This plugin replaces chars 'a' with chars 'b'.
6 */
7module.exports = {
8  Char({node}) {
9    if (node.kind === 'simple' && node.value === 'a') {
10      node.value = 'b';
11      node.symbol = 'b';
12      node.codePoint = 98;
13    }
14  },
15};

Once we have this plugin ready, we can require it, and pass to the transform function:

1const regexpTree = require('regexp-tree');
2const plugin = require('./regexp-tree-a-to-b-transform');
3
4const re = regexpTree.transform(/(a|c)a+[a-z]/, plugin);
5
6console.log(re.toRegExp()); // /(b|c)b+[b-z]/

NOTE: we can also pass a list of plugins to the regexpTree.transform. In this case the plugins are applied in one pass in order. Another approach is to run several sequential calls to transform, setting up a pipeline, when a transformed AST is passed further to another plugin, etc.

You can see other examples of transform plugins in the optimizer/transforms or in the compat-transpiler/transforms directories.

Using generator API

The generator module generates regular expressions from corresponding AST nodes. In Node the module is exposed as regexpTree.generate method.

Example:

1const regexpTree = require('regexp-tree');
2
3const re = regexpTree.generate({
4  type: 'RegExp',
5  body: {
6    type: 'Char',
7    value: 'a',
8    symbol: 'a',
9    kind: 'simple',
10    codePoint: 97
11  },
12  flags: 'i',
13});
14
15console.log(re); // '/a/i'

Using optimizer API

Optimizer transforms your regexp into an optimized version, replacing some sub-expressions with their idiomatic patterns. This might be good for different kinds of minifiers, as well as for regexp machines.

NOTE: the Optimizer is implemented as a set of regexp-tree plugins.

Example:

1const regexpTree = require('regexp-tree');
2
3const originalRe = /[a-zA-Z_0-9][A-Z_\da-z]*\e{1,}/;
4
5const optimizedRe = regexpTree
6  .optimize(originalRe)
7  .toRegExp();
8
9console.log(optimizedRe); // /\w+e+/

From CLI the optimizer is available via --optimize (-o) option:

regexp-tree-cli -e '/[a-zA-Z_0-9][A-Z_\da-z]*\e{1,}/' -o

Result:

Optimized: /\w+e+/

See the optimizer README for more details.

Optimizer ESLint plugin

The optimizer module is also available as an ESLint plugin, which can be installed at: eslint-plugin-optimize-regex.

Using compat-transpiler API

The compat-transpiler module translates your regexp in new format or in new syntax, into an equivalent regexp in a legacy representation, so it can be used in engines which don't yet implement the new syntax.

NOTE: the compat-transpiler is implemented as a set of regexp-tree plugins.

Example, "dotAll" s flag:

1/./s

Is translated into:

1/[\0-\uFFFF]/

Or named capturing groups:

1/(?<value>a)\k<value>\1/

Becomes:

1/(a)\1\1/

To use the API from Node:

1const regexpTree = require('regexp-tree');
2
3// Using new syntax.
4const originalRe = '/(?<all>.)\\k<all>/s';
5
6// For legacy engines.
7const compatTranspiledRe = regexpTree
8  .compatTranspile(originalRe)
9  .toRegExp();
10
11console.log(compatTranspiledRe); // /([\0-\uFFFF])\1/

From CLI the compat-transpiler is available via --compat (-c) option:

regexp-tree-cli -e '/(?<all>.)\k<all>/s' -c

Result:

Compat: /([\0-\uFFFF])\1/

Compat-transpiler Babel plugin

The compat-transpiler module is also available as a Babel plugin, which can be installed at: babel-plugin-transform-modern-regexp.

Note, the plugin also includes extended regexp features.

RegExp extensions

Some of the non-standard feature are also supported by regexp-tree.

NOTE: "non-standard" means specifically ECMAScript standard, since in other regexp egnines, e.g. PCRE, Python, etc. these features are standard.

One of such features is the x flag, which enables extended mode of regular expressions. In this mode most of whitespaces are ignored, and expressions can use #-comments.

Example:

1/
2  # A regular expression for date.
3
4  (?<year>\d{4})-    # year part of a date
5  (?<month>\d{2})-   # month part of a date
6  (?<day>\d{2})      # day part of a date
7
8/x

This is normally parsed by the regexp-tree parser, and compat-transpiler has full support for it; it's translated into:

1/(\d{4})-(\d{2})-(\d{2})/

RegExp extensions Babel plugin

The regexp extensions are also available as a Babel plugin, which can be installed at: babel-plugin-transform-modern-regexp.

Note, the plugin also includes compat-transpiler features.

Creating RegExp objects

To create an actual RegExp JavaScript object, we can use regexpTree.toRegExp method:

1const regexpTree = require('regexp-tree');
2
3const re = regexpTree.toRegExp('/[a-z]/i');
4
5console.log(
6  re.test('a'), // true
7  re.test('Z'), // true
8);

Executing regexes

It is also possible to execute regular expressions using exec API method, which has support for new syntax, and features, such as named capturing group, etc:

1const regexpTree = require('regexp-tree');
2
3const re = `/
4
5  # A regular expression for date.
6
7  (?<year>\\d{4})-    # year part of a date
8  (?<month>\\d{2})-   # month part of a date
9  (?<day>\\d{2})      # day part of a date
10
11/x`;
12
13const string = '2017-04-14';
14
15const result = regexpTree.exec(re, string);
16
17console.log(result.groups); // {year: '2017', month: '04', day: '14'}

Using interpreter API

NOTE: you can read more about implementation details of the interpreter in this series of articles.

In addition to executing regular expressions using JavaScript built-in RegExp engine, RegExp Tree also implements own interpreter based on classic NFA/DFA finite automaton engine.

Currently it aims educational purposes -- to trace the regexp matching process, transitioning in NFA/DFA states. It also allows building state transitioning table, which can be used for custom implementation. In API the module is exposed as fa (finite-automaton) object.

Example:

1const {fa} = require('regexp-tree');
2
3const re = /ab|c*/;
4
5console.log(fa.test(re, 'ab')); // true
6console.log(fa.test(re, '')); // true
7console.log(fa.test(re, 'c')); // true
8
9// NFA, and its transition table.
10const nfa = fa.toNFA(re);
11console.log(nfa.getTransitionTable());
12
13// DFA, and its transition table.
14const dfa = fa.toDFA(re);
15console.log(dfa.getTransitionTable());

For more granular work with NFA and DFA, fa module also exposes convenient builders, so you can build NFA fragments directly:

1const {fa} = require('regexp-tree');
2
3const {
4  alt,
5  char,
6  or,
7  rep,
8} = fa.builders;
9
10// ab|c*
11const re = or(
12  alt(char('a'), char('b')),
13  rep(char('c'))
14);
15
16console.log(re.matches('ab')); // true
17console.log(re.matches('')); // true
18console.log(re.matches('c')); // true
19
20// Build DFA from NFA
21const {DFA} = fa;
22
23const reDFA = new DFA(re);
24
25console.log(reDFA.matches('ab')); // true
26console.log(reDFA.matches('')); // true
27console.log(reDFA.matches('c')); // true

Printing NFA/DFA tables

The --table option allows displaying NFA/DFA transition tables. RegExp Tree also applies DFA minimization (using N-equivalence algorithm), and produces the minimal transition table as its final result.

In the example below for the /a|b|c/ regexp, we first obtain the NFA transition table, which is further converted to the original DFA transition table (down from the 10 non-deterministic states to 4 deterministic states), and eventually minimized to the final DFA table (from 4 to only 2 states).

./bin/regexp-tree-cli -e '/a|b|c/' --table all

Result:

> - starting
✓ - accepting

NFA transition table:

┌─────┬───┬───┬────┬─────────────┐
│     │ a │ b │ c  │ ε*          │
├─────┼───┼───┼────┼─────────────┤
│ 1 > │   │   │    │ {1,2,3,7,9} │
├─────┼───┼───┼────┼─────────────┤
│ 2   │   │   │    │ {2,3,7}     │
├─────┼───┼───┼────┼─────────────┤
│ 3   │ 4 │   │    │ 3           │
├─────┼───┼───┼────┼─────────────┤
│ 4   │   │   │    │ {4,5,6}     │
├─────┼───┼───┼────┼─────────────┤
│ 5   │   │   │    │ {5,6}       │
├─────┼───┼───┼────┼─────────────┤
│ 6 ✓ │   │   │    │ 6           │
├─────┼───┼───┼────┼─────────────┤
│ 7   │   │ 8 │    │ 7           │
├─────┼───┼───┼────┼─────────────┤
│ 8   │   │   │    │ {8,5,6}     │
├─────┼───┼───┼────┼─────────────┤
│ 9   │   │   │ 10 │ 9           │
├─────┼───┼───┼────┼─────────────┤
│ 10  │   │   │    │ {10,6}      │
└─────┴───┴───┴────┴─────────────┘


DFA: Original transition table:

┌─────┬───┬───┬───┐
│     │ a │ b │ c │
├─────┼───┼───┼───┤
│ 1 > │ 4 │ 3 │ 2 │
├─────┼───┼───┼───┤
│ 2 ✓ │   │   │   │
├─────┼───┼───┼───┤
│ 3 ✓ │   │   │   │
├─────┼───┼───┼───┤
│ 4 ✓ │   │   │   │
└─────┴───┴───┴───┘


DFA: Minimized transition table:

┌─────┬───┬───┬───┐
│     │ a │ b │ c │
├─────┼───┼───┼───┤
│ 1 > │ 2 │ 2 │ 2 │
├─────┼───┼───┼───┤
│ 2 ✓ │   │   │   │
└─────┴───┴───┴───┘

AST nodes specification

Below are the AST node types for different regular expressions patterns:

• Char
  • Simple char
  • Escaped char
  • Meta char
  • Control char
  • Hex char-code
  • Decimal char-code
  • Octal char-code
  • Unicode
• Character class
  • Positive character class
  • Negative character class
  • Character class ranges
• Unicode properties
• Alternative
• Disjunction
• Groups
  • Capturing group
  • Named capturing group
  • Non-capturing group
  • Backreferences
• Quantifiers
  • ? zero-or-one
  • * zero-or-more
  • + one-or-more
  • Range-based quantifiers
    • Exact number of matches
    • Open range
    • Closed range
  • Non-greedy
• Assertions
  • ^ begin marker
  • $ end marker
  • Boundary assertions
  • Lookahead assertions
    • Positive lookahead assertion
    • Negative lookahead assertion
  • Lookbehind assertions
    • Positive lookbehind assertion
    • Negative lookbehind assertion

Char

A basic building block, single character. Can be escaped, and be of different kinds.

Simple char

Basic non-escaped char in a regexp:

z

Node:

1{
2  type: 'Char',
3  value: 'z',
4  symbol: 'z',
5  kind: 'simple',
6  codePoint: 122
7}

NOTE: to test this from CLI, the char should be in an actual regexp -- /z/.

Escaped char

\z

The same value, escaped flag is added:

1{
2  type: 'Char',
3  value: 'z',
4  symbol: 'z',
5  kind: 'simple',
6  codePoint: 122,
7  escaped: true
8}

Escaping is mostly used with meta symbols:

// Syntax error
*

\*

OK, node:

1{
2  type: 'Char',
3  value: '*',
4  symbol: '*',
5  kind: 'simple',
6  codePoint: 42,
7  escaped: true
8}

Meta char

A meta character should not be confused with an escaped char.

Example:

\n

Node:

1{
2  type: 'Char',
3  value: '\\n',
4  symbol: '\n',
5  kind: 'meta',
6  codePoint: 10
7}

Among other meta character are: ., \f, \r, \n, \t, \v, \0, [\b] (backspace char), \s, \S, \w, \W, \d, \D.

NOTE: Meta characters representing ranges (like ., \s, etc.) have undefined value for symbol and NaN for codePoint.

NOTE: \b and \B are parsed as Assertion node type, not Char.

Control char

A char preceded with \c, e.g. \cx, which stands for CTRL+x:

\cx

Node:

1{
2  type: 'Char',
3  value: '\\cx',
4  symbol: undefined,
5  kind: 'control',
6  codePoint: NaN
7}

HEX char-code

A char preceded with \x, followed by a HEX-code, e.g. \x3B (symbol ;):

\x3B

Node:

1{
2  type: 'Char',
3  value: '\\x3B',
4  symbol: ';',
5  kind: 'hex',
6  codePoint: 59
7}

Decimal char-code

Char-code:

\42

Node:

1{
2  type: 'Char',
3  value: '\\42',
4  symbol: '*',
5  kind: 'decimal',
6  codePoint: 42
7}

Octal char-code

Char-code started with \0, followed by an octal number:

\073

Node:

1{
2  type: 'Char',
3  value: '\\073',
4  symbol: ';',
5  kind: 'oct',
6  codePoint: 59
7}

Unicode

Unicode char started with \u, followed by a hex number:

\u003B

Node:

1{
2  type: 'Char',
3  value: '\\u003B',
4  symbol: ';',
5  kind: 'unicode',
6  codePoint: 59
7}

When using the u flag, unicode chars can also be represented using \u followed by a hex number between curly braces:

\u{1F680}

Node:

1{
2  type: 'Char',
3  value: '\\u{1F680}',
4  symbol: '🚀',
5  kind: 'unicode',
6  codePoint: 128640
7}

When using the u flag, unicode chars can also be represented using a surrogate pair:

\ud83d\ude80

Node:

1{
2  type: 'Char',
3  value: '\\ud83d\\ude80',
4  symbol: '🚀',
5  kind: 'unicode',
6  codePoint: 128640,
7  isSurrogatePair: true
8}

Character class

Character classes define a set of characters. A set may include as simple characters, as well as character ranges. A class can be positive (any from the characters in the class match), or negative (any but the characters from the class match).

Positive character class

A positive character class is defined between [ and ] brackets:

[a*]

A node:

1{
2  type: 'CharacterClass',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'a',
7      symbol: 'a',
8      kind: 'simple',
9      codePoint: 97
10    },
11    {
12      type: 'Char',
13      value: '*',
14      symbol: '*',
15      kind: 'simple',
16      codePoint: 42
17    }
18  ]
19}

NOTE: some meta symbols are treated as normal characters in a character class. E.g. * is not a repetition quantifier, but a simple char.

Negative character class

A negative character class is defined between [^ and ] brackets:

[^ab]

An AST node is the same, just negative property is added:

1{
2  type: 'CharacterClass',
3  negative: true,
4  expressions: [
5    {
6      type: 'Char',
7      value: 'a',
8      symbol: 'a',
9      kind: 'simple',
10      codePoint: 97
11    },
12    {
13      type: 'Char',
14      value: 'b',
15      symbol: 'b',
16      kind: 'simple',
17      codePoint: 98
18    }
19  ]
20}

Character class ranges

As mentioned, a character class may also contain ranges of symbols:

[a-z]

A node:

1{
2  type: 'CharacterClass',
3  expressions: [
4    {
5      type: 'ClassRange',
6      from: {
7        type: 'Char',
8        value: 'a',
9        symbol: 'a',
10        kind: 'simple',
11        codePoint: 97
12      },
13      to: {
14        type: 'Char',
15        value: 'z',
16        symbol: 'z',
17        kind: 'simple',
18        codePoint: 122
19      }
20    }
21  ]
22}

NOTE: it is a syntax error if to value is less than from value: /[z-a]/.

The range value can be the same for from and to, and the special range - character is treated as a simple character when it stands in a char position:

// from: 'a', to: 'a'
[a-a]

// from: '-', to: '-'
[---]

// simple '-' char:
[-]

// 3 ranges:
[a-zA-Z0-9]+

Unicode properties

Unicode property escapes are a new type of escape sequence available in regular expressions that have the u flag set. With this feature it is possible to write Unicode expressions as:

1const greekSymbolRe = /\p{Script=Greek}/u;
2
3greekSymbolRe.test('π'); // true

The AST node for this expression is:

1{
2  type: 'UnicodeProperty',
3  name: 'Script',
4  value: 'Greek',
5  negative: false,
6  shorthand: false,
7  binary: false,
8  canonicalName: 'Script',
9  canonicalValue: 'Greek'
10}

All possible property names, values, and their aliases can be found at the specification.

For General_Category it is possible to use a shorthand:

1/\p{Letter}/u;   // Shorthand
2
3/\p{General_Category=Letter}/u; // Full notation

Binary names use the single value as well:

1/\p{ASCII_Hex_Digit}/u; // Same as: /[0-9A-Fa-f]/

The capitalized P defines the negation of the expression:

1/\P{ASCII_Hex_Digit}/u; // NOT a ASCII Hex digit

Alternative

An alternative (or concatenation) defines a chain of patterns followed one after another:

abc

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'a',
7      symbol: 'a',
8      kind: 'simple',
9      codePoint: 97
10    },
11    {
12      type: 'Char',
13      value: 'b',
14      symbol: 'b',
15      kind: 'simple',
16      codePoint: 98
17    },
18    {
19      type: 'Char',
20      value: 'c',
21      symbol: 'c',
22      kind: 'simple',
23      codePoint: 99
24    }
25  ]
26}

Another examples:

// 'a' with a quantifier, followed by 'b'
a?b

// A group followed by a class:
(ab)[a-z]

Disjunction

The disjunction defines "OR" operation for regexp patterns. It's a binary operation, having left, and right nodes.

Matches a or b:

a|b

A node:

1{
2  type: 'Disjunction',
3  left: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  right: {
11    type: 'Char',
12    value: 'b',
13    symbol: 'b',
14    kind: 'simple',
15    codePoint: 98
16  }
17}

Groups

The groups play two roles: they define grouping precedence, and allow to capture needed sub-expressions in case of a capturing group.

Capturing group

"Capturing" means the matched string can be referred later by a user, including in the pattern itself -- by using backreferences.

Char a, and b are grouped, followed by the c char:

(ab)c

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Group',
6      capturing: true,
7      number: 1,
8      expression: {
9        type: 'Alternative',
10        expressions: [
11          {
12            type: 'Char',
13            value: 'a',
14            symbol: 'a',
15            kind: 'simple',
16            codePoint: 97
17          },
18          {
19            type: 'Char',
20            value: 'b',
21            symbol: 'b',
22            kind: 'simple',
23            codePoint: 98
24          }
25        ]
26      }
27    },
28    {
29      type: 'Char',
30      value: 'c',
31      symbol: 'c',
32      kind: 'simple',
33      codePoint: 99
34    }
35  ]
36}

As we can see, it also tracks the number of the group.

Another example:

// A grouped disjunction of a symbol, and a character class:
(5|[a-z])

Named capturing group

A capturing group can be given a name using the (?<name>...) syntax, for any identifier name.

For example, a regular expressions for a date:

1/(?<year>\d{4})-(?<month>\d{2})-(?<day>\d{2})/u

For the group:

1(?<foo>x)

We have the following node (the name property with value foo is added):

1{
2  type: 'Group',
3  capturing: true,
4  name: 'foo',
5  nameRaw: 'foo',
6  number: 1,
7  expression: {
8    type: 'Char',
9    value: 'x',
10    symbol: 'x',
11    kind: 'simple',
12    codePoint: 120
13  }
14}

Note: The nameRaw property represents the name as parsed from the original source, including escape sequences. The name property represents the canonical decoded form of the name.

For example, given the /u flag and the following group:

1(?<\u{03C0}>x)

We would have the following node:

1{
2  type: 'Group',
3  capturing: true,
4  name: 'π',
5  nameRaw: '\\u{03C0}',
6  number: 1,
7  expression: {
8    type: 'Char',
9    value: 'x',
10    symbol: 'x',
11    kind: 'simple',
12    codePoint: 120
13  }
14}

Non-capturing group

Sometimes we don't need to actually capture the matched string from a group. In this case we can use a non-capturing group:

Char a, and b are grouped, but not captured, followed by the c char:

(?:ab)c

The same node, the capturing flag is false:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Group',
6      capturing: false,
7      expression: {
8        type: 'Alternative',
9        expressions: [
10          {
11            type: 'Char',
12            value: 'a',
13            symbol: 'a',
14            kind: 'simple',
15            codePoint: 97
16          },
17          {
18            type: 'Char',
19            value: 'b',
20            symbol: 'b',
21            kind: 'simple',
22            codePoint: 98
23          }
24        ]
25      }
26    },
27    {
28      type: 'Char',
29      value: 'c',
30      symbol: 'c',
31      kind: 'simple',
32      codePoint: 99
33    }
34  ]
35}

Backreferences

A capturing group can be referenced in the pattern using notation of an escaped group number.

Matches abab string:

(ab)\1

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Group',
6      capturing: true,
7      number: 1,
8      expression: {
9        type: 'Alternative',
10        expressions: [
11          {
12            type: 'Char',
13            value: 'a',
14            symbol: 'a',
15            kind: 'simple',
16            codePoint: 97
17          },
18          {
19            type: 'Char',
20            value: 'b',
21            symbol: 'b',
22            kind: 'simple',
23            codePoint: 98
24          }
25        ]
26      }
27    },
28    {
29      type: 'Backreference',
30      kind: 'number',
31      number: 1,
32      reference: 1,
33    }
34  ]
35}

A named capturing group can be accessed using \k<name> pattern, and also using a numbered reference.

Matches www:

1(?<foo>w)\k<foo>\1

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Group',
6      capturing: true,
7      name: 'foo',
8      nameRaw: 'foo',
9      number: 1,
10      expression: {
11        type: 'Char',
12        value: 'w',
13        symbol: 'w',
14        kind: 'simple',
15        codePoint: 119
16      }
17    },
18    {
19      type: 'Backreference',
20      kind: 'name',
21      number: 1,
22      reference: 'foo',
23      referenceRaw: 'foo'
24    },
25    {
26      type: 'Backreference',
27      kind: 'number',
28      number: 1,
29      reference: 1
30    }
31  ]
32}

Note: The referenceRaw property represents the reference as parsed from the original source, including escape sequences. The reference property represents the canonical decoded form of the reference.

For example, given the /u flag and the following pattern (matches www):

1(?<π>w)\k<\u{03C0}>\1

We would have the following node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Group',
6      capturing: true,
7      name: 'π',
8      nameRaw: 'π',
9      number: 1,
10      expression: {
11        type: 'Char',
12        value: 'w',
13        symbol: 'w',
14        kind: 'simple',
15        codePoint: 119
16      }
17    },
18    {
19      type: 'Backreference',
20      kind: 'name',
21      number: 1,
22      reference: 'π',
23      referenceRaw: '\\u{03C0}'
24    },
25    {
26      type: 'Backreference',
27      kind: 'number',
28      number: 1,
29      reference: 1
30    }
31  ]
32}

Quantifiers

Quantifiers specify repetition of a regular expression (or of its part). Below are the quantifiers which wrap a parsed expression into a Repetition node. The quantifier itself can be of different kinds, and has Quantifier node type.

? zero-or-one

The ? quantifier is short for {0,1}.

a?

Node:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: '?',
13    greedy: true
14  }
15}

* zero-or-more

The * quantifier is short for {0,}.

a*

Node:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: '*',
13    greedy: true
14  }
15}

+ one-or-more

The + quantifier is short for {1,}.

// Same as `aa*`, or `a{1,}`
a+

Node:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: '+',
13    greedy: true
14  }
15}

Range-based quantifiers

Explicit range-based quantifiers are parsed as follows:

Exact number of matches

a{3}

The type of the quantifier is Range, and from, and to properties have the same value:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: 'Range',
13    from: 3,
14    to: 3,
15    greedy: true
16  }
17}

Open range

An open range doesn't have max value (assuming semantic "more", or Infinity value):

a{3,}

An AST node for such range doesn't contain to property:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: 'Range',
13    from: 3,
14    greedy: true
15  }
16}

Closed range

A closed range has explicit max value: (which syntactically can be the same as min value):

a{3,5}

// Same as a{3}
a{3,3}

An AST node for a closed range:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: 'Range',
13    from: 3,
14    to: 5,
15    greedy: true
16  }
17}

NOTE: it is a syntax error if the max value is less than min value: /a{3,2}/

Non-greedy

If any quantifier is followed by the ?, the quantifier becomes non-greedy.

Example:

a+?

Node:

1{
2  type: 'Repetition',
3  expression: {
4    type: 'Char',
5    value: 'a',
6    symbol: 'a',
7    kind: 'simple',
8    codePoint: 97
9  },
10  quantifier: {
11    type: 'Quantifier',
12    kind: '+',
13    greedy: false
14  }
15}

Other examples:

a??
a*?
a{1}?
a{1,}?
a{1,3}?

Assertions

Assertions appear as separate AST nodes, however instread of manipulating on the characters themselves, they assert certain conditions of a matching string. Examples: ^ -- beginning of a string (or a line in multiline mode), $ -- end of a string, etc.

^ begin marker

The ^ assertion checks whether a scanner is at the beginning of a string (or a line in multiline mode).

In the example below ^ is not a property of the a symbol, but a separate AST node for the assertion. The parsed node is actually an Alternative with two nodes:

^a

The node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Assertion',
6      kind: '^'
7    },
8    {
9      type: 'Char',
10      value: 'a',
11      symbol: 'a',
12      kind: 'simple',
13      codePoint: 97
14    }
15  ]
16}

Since assertion is a separate node, it may appear anywhere in the matching string. The following regexp is completely valid, and asserts beginning of the string; it'll match an empty string:

^^^^^

$ end marker

The $ assertion is similar to ^, but asserts the end of a string (or a line in a multiline mode):

a$

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'a',
7      symbol: 'a',
8      kind: 'simple',
9      codePoint: 97
10    },
11    {
12      type: 'Assertion',
13      kind: '$'
14    }
15  ]
16}

And again, this is a completely valid regexp, and matches an empty string:

^^^^$$$$$

// valid too:
$^

Boundary assertions

The \b assertion check for word boundary, i.e. the position between a word and a space.

Matches x in x y, but not in xy:

x\b

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'x',
7      symbol: 'x',
8      kind: 'simple',
9      codePoint: 120
10    },
11    {
12      type: 'Assertion',
13      kind: '\\b'
14    }
15  ]
16}

The \B is vice-versa checks for non-word boundary. The following example matches x in xy, but not in x y:

x\B

A node is the same:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'x',
7      symbol: 'x',
8      kind: 'simple',
9      codePoint: 120
10    },
11    {
12      type: 'Assertion',
13      kind: '\\B'
14    }
15  ]
16}

Lookahead assertions

These assertions check whether a pattern is followed (or not followed for the negative assertion) by another pattern.

Positive lookahead assertion

Matches a only if it's followed by b:

a(?=b)

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'a',
7      symbol: 'a',
8      kind: 'simple',
9      codePoint: 97
10    },
11    {
12      type: 'Assertion',
13      kind: 'Lookahead',
14      assertion: {
15        type: 'Char',
16        value: 'b',
17        symbol: 'b',
18        kind: 'simple',
19        codePoint: 98
20      }
21    }
22  ]
23}

Negative lookahead assertion

Matches a only if it's not followed by b:

a(?!b)

A node is similar, just negative flag is added:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Char',
6      value: 'a',
7      symbol: 'a',
8      kind: 'simple',
9      codePoint: 97
10    },
11    {
12      type: 'Assertion',
13      kind: 'Lookahead',
14      negative: true,
15      assertion: {
16        type: 'Char',
17        value: 'b',
18        symbol: 'b',
19        kind: 'simple',
20        codePoint: 98
21      }
22    }
23  ]
24}

Lookbehind assertions

NOTE: Lookbehind assertions are not yet supported by JavaScript RegExp. It is an ECMAScript proposal which is at stage 3 at the moment.

These assertions check whether a pattern is preceded (or not preceded for the negative assertion) by another pattern.

Positive lookbehind assertion

Matches b only if it's preceded by a:

(?<=a)b

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Assertion',
6      kind: 'Lookbehind',
7      assertion: {
8        type: 'Char',
9        value: 'a',
10        symbol: 'a',
11        kind: 'simple',
12        codePoint: 97
13      }
14    },
15    {
16      type: 'Char',
17      value: 'b',
18      symbol: 'b',
19      kind: 'simple',
20      codePoint: 98
21    },
22  ]
23}

Negative lookbehind assertion

Matches b only if it's not preceded by a:

(?<!a)b

A node:

1{
2  type: 'Alternative',
3  expressions: [
4    {
5      type: 'Assertion',
6      kind: 'Lookbehind',
7      negative: true,
8      assertion: {
9        type: 'Char',
10        value: 'a',
11        symbol: 'a',
12        kind: 'simple',
13        codePoint: 97
14      }
15    },
16    {
17      type: 'Char',
18      value: 'b',
19      symbol: 'b',
20      kind: 'simple',
21      codePoint: 98
22    },
23  ]
24}