Skip to main content
Deno 2 is finally here 🎉️
Learn more

lexer

deno land deno doc

Lexical analyzer for JavaScript.

Features

  • Tiny
  • Simple / Lean
  • Regex ready
  • Maximum munch algorithm
  • Composable

Basic usage

Example of simple addition.

import { Lexer } from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

enum TokenType {
  LET = "Let",
  NUMBER = "Number",
  IDENT = "Ident",
  ASSIGN = "Assign",
  PLUS = "Plus",
  SEMICOLON = "Semicolon",
}

const lexer = new Lexer({
  [TokenType.LET]: "let",
  [TokenType.NUMBER]: /\d+/,
  [TokenType.IDENT]: /[a-z]+/i,
  [TokenType.ASSIGN]: "=",
  [TokenType.PLUS]: "+",
  [TokenType.SEMICOLON]: ";",

  "WS": {
    pattern: /[\s\t]+/,
    ignore: true,
  },
});
const input = `let sum = 100 + 200;`;
const result = lexer.lex(input);
assertEquals(result, {
  tokens: [
    { type: "Let", literal: "let", offset: 0 },
    { type: "Ident", literal: "sum", offset: 3 },
    // ...,
    { type: "Semicolon", literal: ";", offset: 19 },
  ],
  done: true,
  offset: 20,
});

Map of token type and patterns

The lexer is configured by map of token type and a token patterns.

Several expressions are supported for token patterns.

  • string
  • regex
  • above with options

string pattern

String pattern recognizes strings as they are as patterns.

import { TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";

const tokenMap: TokenMap = {
  IMPORT: "import",
  FROM: "from",
  SEMICOLON: ";",
};

String patterns can be used for various tokens such as keywords, separators, etc., and are the most readable and perform the best.

regex pattern

The regex pattern defines a regular expression as a RegExp object.

import { TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";
const tokenMap: TokenMap = {
  IDENT: /[a-z]+/i,
  NUMBER: /\d+/,
};

Note that regular expressions cannot be defined as string patterns.

import { TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";
const tokenMap: TokenMap = {
  NOT_REGEX_OBJECT: "\\s+",
};

In this case, matching is performed as the string \s+.

Also, RegExp flags of g is restricted to use. This is because it matches inputs with Left to Right.

If you use the g flag, an error is thrown before execution.

Pattern with options

The two patterns above allow almost any token to be represented.

For more fine-grained control over the generation of tokens, optional patterns can be defined.

import { TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";

const tokenMap: TokenMap = {
  WS: {
    pattern: /[\s\t]+/,
    ignore: true,
  },
};
Name Required Description
pattern :white_check_mark: string | RegExp
Token matching pattern.
ignore booleanWhether the token ignore or not.

If ignore is set to true, then even if the pattern matches, it will not be added to the generated token stream.

EOF token

Scan inputs from left to right. Therefore, EOF cannot be represented as a string or regular expression.

If you want an EOF token at the end of your token stream, you must add the EOF symbol to your pattern.

The lexer treats only EOF symbols “specially”.

import {
  EOF,
  Lexer,
  TokenMap,
} from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

const tokenMap: TokenMap = {
  "<EOF>": EOF, // or Symbol("EOF"),
};
const { tokens } = new Lexer(tokenMap).lex("");
assertEquals(tokens, [{ type: "<EOF>", literal: "", offset: 0 }]);

The EOF token will only be included if all inputs can be parsed.

import {
  EOF,
  Lexer,
  TokenMap,
} from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

const tokenMap: TokenMap = {
  "<EOF>": EOF,
};
const result = new Lexer(tokenMap).lex("a");
assertEquals(result, { tokens: [], done: false, offset: 0 });

Maximum munch

The maximum munch is used in the matching algorithm.

This identifies the longest matching pattern as the token, not in the order of definition in the token map.

import { Lexer, TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

const tokenMap: TokenMap = {
  "A": "A",
  "AAA": "AAA",
  "AA": "AA",
};
const { tokens } = new Lexer(tokenMap).lex("AAAAA");
assertEquals(tokens, [
  { type: "AAA", literal: "AAA", offset: 0 },
  { type: "AA", literal: "AA", offset: 3 },
]);

The maximum munch eliminates the need to worry about definition order and reduces the potential for many bugs.

The same is true for the regex pattern.

Pattern priority

String patterns take precedence over regex patterns. If there is a match in the string pattern, no match is made in the regex pattern.

import { Lexer, TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

const tokenMap: TokenMap = {
  CONST: "const",
  IDENT: /.+/,
};
const { tokens } = new Lexer(tokenMap).lex("constant");
assertEquals(tokens, [
  { type: "CONST", literal: "const", offset: 0 },
  { type: "IDENT", literal: "ant", offset: 5 },
]);

Conclusion:

long string > short string >> long regex > short regex

Lexical analyzing result

The input is converted to a token stream by the lexer.

The result has the following structure:

Name Description
tokens Token[]
Token streams.
done boolean
Whether the lex has done or not.
offset number
Final offset. Same as the last index of the last token.

Token structure is:

Name Description
type string
Defined token type.
literal string
Actual token literal value.
offset number
FStart index of the matched in the input.

If a character not matching the pattern is encountered, it is aborted and done is returned as false along with the token stream up to that point.

Generic token types

The type of a token matches the token type defined in the token map.

The type of the token is narrowed by generics.

import { Lexer, TokenMap } from "https://deno.land/x/lexer@$VERSION/mod.ts";
import { assertEquals } from "https://deno.land/std@$VERSION/testing/asserts.ts";

const tokenMap: TokenMap = {
  CONST: "const",
  IDENT: /.+/,
};
const { tokens } = new Lexer(tokenMap).lex("constant");
tokens.forEach(({ type }) => {
  type; // infer "CONST" | "INDENT"
});

Benchmark

Benchmark script with comparison to several popular schema library is available.

deno task bench

License

Copyright © 2022-present TomokiMiyauci.

Released under the MIT license