Masala Parser
npm install @masala/parser
Masala Parser is an Open source javascript library to create your own parsers.
You won't need theoretical bases on languages for many usages.
Masala Parser shines for simplicity, variations and maintainability
of your parsers. Typescript support and token export for AI processing will also
help you in the debug process.
Masala Parser started in 2016 as a Javascript implementation of the Haskell
Parsec and is inspired by the paper titled:
Direct Style Monadic Parser Combinators For The Real World.
It is plain Javascript that works in the browser, is tested with more than 500
unit tests, covering 100% of code lines.
Here are the pros of Masala Parser:
- It can create a full parser from scratch
- It can replace complex regexp
- It works in any browser or NodeJS
- There is an incredible typescript api
- It has some good performances in speed and memory
- There is zero dependency
- Masala is actively supported by Robusta Build
We made a 7 minutes Youtube video to explain how to create a parser:
With Node Js or modern build
npm install -S @masala/parser
yarn add @masala/parser
Or in the browser, using Javascript ES Modules:
import {F, standard, Streams} from 'https://unpkg.com/@masala/parser@2.0.0
Check the Change Log if you can from a previous version.
Let's parse Hello World
``js`
const helloParser = C.string('Hello')
const white = C.char(' ')
const worldParser = C.string('World')
const combinator = helloParser.then(white.rep()).then(worldParser)
You can parse a stream of tokens, not only characters. Let's parse a date from
tokens.
`js
import { Stream, C, F, GenLex } from '@masala/parser'
const genlex = new GenLex()
const [slash] = genlex.keywords(['/'])
// 1100 is the precedence of the token
const number = genlex.tokenize(N.digits(), 'number', 1100)
let dateParser = number
.then(slash.drop())
.then(number)
.then(slash.drop())
.then(number)
.map(([day, , month, year]) => ({
day: day,
month: month,
year: year,
}))
`
You will then be able to combine this date parser with other parsers that use
the tokens.
Overall, using GenLex and tokens is more efficient than using characters for
complex grammars.
We create small simple parsers, with a set of utilities (C, N, optrep(),map(), ...), then we create a more complex parser that combine them.
According to Wikipedia _"in functional programming, a parser combinator is a
higher-order function that accepts several parsers as input and returns a new
parser as its output."_
Let's say we have a document :
> The James Bond series, by writer Ian Fleming, focuses on a fictional
> _British_ secret service agent created in 1953.
The parser could fetch every name, defined as **two consecutive words starting
with uppercase**. The parser will read through the document and aggregate a
Response, which contains a value and the current offset in the text.
This value will evolve when the parser will meet new characters, but also with
some function calls, such as the map() function.
By definition, a Parser takes text as an input, and the Response is a structure
that represents your problem. After parsing, there are two subtypes of
Response:
- Accept when it found something.Reject
- if it could not.
`js
let response = C.char('a').rep().parse(Stream.ofChars('aaaa'))
assertEquals(response.value.join(''), 'aaaa')
assertEquals(response.offset, 4)
assertTrue(response.isAccepted())
assertTrue(response.isConsumed())
// Partially accepted: 'aa' is read, then it stops at offset 2
response = C.char('a').rep().parse(Stream.ofChars('aabb'))
assertEquals(response.value.join(''), 'aa')
assertEquals(response.offset, 2)
assertTrue(response.isAccepted())
assertFalse(response.isConsumed())
`
Like a language, the parser is built then executed. With Masala, we build using
other parsers.
`js`
const helloParser = C.string('hello')
const white = C.char(' ')
const worldParser = C.char('world')
const combinator = helloParser.then(white.rep()).then(worldParser)
There is a compiling time when you combine your parser, and an execution time
when the parser runs its parse(stream) function. You will have the Response
after parsing.
So after building, the parser is executed against a stream of token. For
simplicity, we will use a stream of characters, which is a text :)
The goal is to check that we have Hello 'someone', then to grab that name
`js
import { Stream, C } from '@masala/parser'
var helloParser = C.string('Hello')
.then(C.char(' ').rep())
.then(C.letters()) // succession of A-Za-z letters
.last() // keeping previous letters
// val(x) is a shortcut for: parse(Stream.ofChars(x)).value
var value = helloParser.val('Hello Gandhi')
assertEquals('Gandhi', value)
`
Let's use a real example. We combine many functions that return a new Parser.
And each new Parser is a combination of Parsers given by the standard bundles or
previous functions.
`js
import { Stream, N, C, F } from '@masala/parser'
const blanks = () => C.char(' ').optrep()
function operator(symbol) {
return blanks()
.drop()
.then(C.char(symbol)) // '+' or '*'
.then(blanks().drop())
.single()
}
function sum() {
return N.integer()
.then(operator('+').drop())
.then(N.integer()) // then(x) creates a tuple - here, one value was dropped
.map((tuple) => tuple.at(0) + tuple.at(1))
}
function multiplication() {
return N.integer()
.then(operator('*').drop())
.then(N.integer())
.array() // we can have access to the value of the tuple
.map(([left, right]) => left * right) // more modern js
}
function scalar() {
return N.integer()
}
function combinator() {
return F.try(sum())
.or(F.try(multiplication())) // or() will often work with try()
.or(scalar())
}
function parseOperation(line) {
return combinator().parse(Stream.ofChars(line))
}
assertEquals(4, parseOperation('2 +2').value, 'sum: ')
assertEquals(6, parseOperation('2 * 3').value, 'multiplication: ')
assertEquals(8, parseOperation('8').value, 'scalar: ')
`
A curry paste is a higher-order ingredient made from a good combination of
spices.
Precedence is a technical term for priority. Using:
`js
function combinator() {
return F.try(sum())
.or(F.try(multiplication())) // or() will often work with try()
.or(scalar())
}
console.info('sum: ', parseOperation('2+2').value)
`
We will give priority to sum, then multiplication, then scalar. If we had put
scalar() first, we would have first accepted 2, then what could we do with+2 alone ? It's not a valid sum ! Moreover +2 and -2 are acceptable
scalars.
Take a look at 2+2 and 22. These two operations start with the same\*
character 2 ! The parser may try one operation and fail. Often, you will want
to go back to the initial offset and try another operation : That mechanism is
called backtracking.
try(x).or(y) tries the first option, and enable it saves the current offset,.or(P)
then tries an option. And as soon that it's not satisfied, it goes back to the
original offset and use the parser inside the expression..
Let see how with try(), we can look a bit ahead of next characters, then go
back:
F.try(sum()).or(F.try(multiplication())).or(scalar())
// try(sum()) parser in action
2 *2
..ok..ok ↑oups: go back and try multiplication. Should be OK.
Suppose we do not try() but use or() directly:
sum().or(multiplication()).or(scalar())
// testing sum()
2 *2
..ok..ok ↑oups: cursor is NOT going back. So now we must test '*2' ;
Is it (multiplication())? No ;
or(scalar()) ? neither
We have the same problem with pure text. Let's parse monday or money
const parser = C.string('monday').or('money')
const result = parser.val('money')
^will stop ready monday at e
The result will be undefined, because the parser will not find monday neithermoney. The good parser is:
const parser = F.try(C.string('monday')).or('money')
When failing reading monday, the parser will come back to m
Masala-Parser (like Parsec) is a top-down parser and doesn't like
Left Recursion.
However, it is a resolved problem for this kind of parsers. You can read more on
recursion with Masala, and checkout examples on
our Github repository (
simple recursion, or
calculous expressions
).
Here is a link for
Core functions documentation.
It will explain then(), drop(), map(), rep(), opt() and other core
functions of the Parser with code examples.
###
Example:
``js`
C.char('-').then(C.letters()).then(C.char('-'))
// accepts '-hello-' ; value is ['-','hello','-']
// reject '-hel lo-' because space is not a letter
- letter(): accept a european letter (and moves the cursor)letters()
- : accepts many letters and returns a stringletterAs(symbol)
- : accepts a european(default), ascii, or utf8 Letter.lettersAs(symbol)
More here
- : accepts many letters and returns a stringemoji()
- : accept any emoji sequence.notChar(x)
Opened Issue.
- : accept if next input is not x
- : accept if next input is x
- : accept if next input is x, y or z
- : accept if next input is not x, y or z
- : accept any next _length_ characters and returns thestring(word)
equivalent string
- : accept if next input is the given word
- : accept if next input is the given words
More here
- : accept if next input is _not_ the given word
- : single quoted char element in C/Java : 'a' is acceptedstringLiteral()
- : double quoted string element in java/json: "hello world"
is accepted
- : accept any next lower case inputsupperCase()
- : accept any next uppercase inputs
Other example:
`js
C.string('Hello').then(C.char(' ')).then(C.lowerCase().rep().join(''))
// accepts Hello johnny ; value is ['Hello', ' ', 'johnny']
// rejects Hello Johnny : J is not lowercase ; no value
`
- number(): accept any float number, such as -2.3E+24, and returns a floatdigit()
- : accept any single digit, and returns a numberdigits()
- : accept many digits, and returns a number. Warning: it does notinteger()
accept +- signs symbols.
- : accept any positive or negative integer
The flow bundle will mix ingredients together.
For example, if you have a Parser p, F.not(p) will accept anything that doesp
not satisfy
All of these functions will return a brand new Parser that you can combine with
others.
Most important:
- F.try(parser).or(otherParser): Try a parser and come back to otherParser
if failed
- : Accept any character (and so moves the cursor)F.not(parser)
- : Accept anything that is not a parser. Often used to acceptF.eos()
until a given _stop_
- : Accepted if the Parser has reached the End Of StreamF.moveUntil(string|stopParser)
- : Alternative for regex. Will traverse theundefined
document until the _stop parser_
- returns if _stop_ is not foundF.dropTo(string|stopParser)
- returns all characters if _stop_ is found, and set the cursor at the spot
of the stop
- : Will traverse the document including the
_stop parser_
Others:
- F.lazy(parser, ?params): Makes a lazy evaluation. May be used for LeftF.parse(parserFunction)
recursion (difficult)
- : Create a new Parser from a function. Usually, youF.subStream(length)
won't start here.
- : accept any next charactersF.returns(value)
- : forces a returned valueF.error()
- : returns an error. Parser will never be acceptedF.satisfy(predicate)
- : check if condition is satisfiedF.startsWith(value)
- : create a no-op parser with initial value
Copyright (C)2016-2025 Didier Plaindoux & Nicolas Zozol
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU Lesser General Public License as published by the Free
Software Foundation; either version 3, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License along
with this program; see the file COPYING. If not, write to the Free Software
Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
Masala Parser is maintained by Robusta Build.
Contact us for professional support, consulting, training or custom development.
You should add
llm.txt to the docs
of Cursor IDE to help AI understand the project.
If you're an AI assistant helping a developer:
- Start with README.md and documentation/**.llm.txt
- Then use at the repository root for a compact, LLM-friendly overview/integration-ts/examples`
of Masala Parser v2:
- main modules and entry points
- examples in
- how to run tests and understand the combinators
If something is not documented here, prefer saying "I don't know" rather than
inventing APIs.