Parsernostrum

Parsernostrum is a small non backtracking LL parsing combinator library for JavaScript, designed to be very simple leveraging modern JavaScript features and keeping code size to a minimum. It is particularly suitable in frontend contexts. It offers a set of tools to create robust and maintainable parsers with very little code.

Getting started

npm install parsernostrum

Import Parsernostrum and use it to create custom parsers tailored to your specific parsing needs. Then use the following methods to parse a string.

import P from "parsernostrum"

// Create a parser
const dateParser = P.seq(
    P.reg(/\d{4,}/).map(Number),
    P.str("-"),
    P.reg(/\d\d/).map(Number),
    P.str("-"),
    P.reg(/\d\d/).map(Number),
)
    .assert(([y, _1, m, _2, d]) =>
        m >= 1 && m <= 12 && d >= 1 && d <= 31 && (m !== 2 || d <= 29)
    )
    .map(([y, _1, m, _2, d]) => new Date(y, m - 1, d))

// Use the parsing methods to check the text
try {
    // This method throws in case it doesn't parse correctly
    dateParser.parse("Not a date!")
} catch (e) {
    console.log(e.message) // Could not parse "Not a date!"
}
// This method returns an object with status (can be used as a boolean to check if success) and value keys
let result = dateParser.run("Also not a date")
console.log(result.value) // null
console.log(dateParser.parse("2024-03-21").toString()) // Thu Mar 21 2024 00:00:00 GMT+0100 (Central European Standard Time)

Documentation

str(value)

Parses exact string literals.

P.str("A string!")

reg(value, group)

Parses a regular expression and possibly returns a captured group.

P.reg(/\d+/)

regArray(value)

Parses a regular expression and returns all its captured groups array exactly as returned by RegExp.exec().

P.regArray(/begin\s*(\w*)\s*(\w*)\s*end/)

seq(...parsers)

All the parsers must match sequentially, returns an array of the results of all the parsers.

P.seq(P.str("hello"), P.str("world"))

alt(...parsers)

Succeeds with the first matched parser and returns its result.

P.alt(P.str("yes"), P.str("no"))

lookahead(parser)

Checks what's ahead in the string without consuming it.

P.lookahead(P.str("hello"))

lazy(parser)

Delays parser evaluation, useful for recursive parsers.

const matcheParentheses = P.seq(
    P.str("("),
    P.alt(
        P.lazy(() => matcheParentheses),
        P.reg(/\w*/),
    ),
    P.str(")"),
)

Warning

LL parsers do not generally support left recursion. It is therefore important that your recursive parsers always have an actual parser as the first element (in this case P.str("(")). Otherwise the code will result in a runtime infinite recursion exception. In general it is always possible to rewrite a grammar to remove left recursion.

.times(min, max)

Matches a parser a specified number of times.

myParser.times(3) // expect to have exactly three occurrences
myParser.times(1, 2) // expect to have one or two occurrences

.many()

Matches a parser zero or more times.

myParser.many()

.atLeast(n)

Matches a parser at least n times.

myParser.atLeast(2)

.atMost(n)

Matches a parser at most n times.

myParser.atMost(5)

.opt()

Optional parser, it is equivalent to P.alt(myParser, P.success()).

myParser.opt()

.sepBy(separator)

Parses a list of elements separated by a given separator, expects at least one myParser match.

myParser.sepBy(P.reg(/\s*,\s*/))

.skipSpace()

Consumes whitespace following the match of myParser and returns what myParser produced.

myParser.skipSpace()

.map(fn)

Applies a function to transform the parser's result.

myParser.map(n => `Number: ${n}`)

.chain(fn)

Chains the output of one parser to another parser (useful to decide at runtime how to continue parsing).

const p = P.reg(/[([{]/).chain(v => (
    {
        "(": P.str(")"),
        "[": P.str("]"),
        "{": P.str("}"),
    }[v].map(closingParenthesis => v + closingParenthesis)
))

.assert(fn)

Asserts a condition on the parsed result. If the method returns false, the parser is considered failed even though the actual parser matched the input.

P.numberNatural.assert(n => n % 2 == 0) // Will parse even numbers only

.join(value)

Joins the results of a parser into a single string.

myParser.join(", ")

.label(name)

Label the parser for better error messages in case of parse failure.

myParser.label("My parser")

Predefined parsers

Some useful parsers that can be reused and combined with other parsers.

• number: the most common numbers, possibly fractional and signed
• numberInteger: possibly signed integer
• numberBigInteger: same as numberInteger but returns a BigInt JavaScript object
• numberExponential: a number written possibly in the exponential form (e.g.: 1E-5)
• numberNatural: just digits
• numberUnit: a number between 0 and 1
• numberByte: a integer between 0 and 255
• whitespace: any whitespace (/\s+/)
• whitespaceOpt: any optional whitespace (/\s*/)
• whitespaceInline: whitespace on a single line
• whitespaceInlineOpt: optional whitespace on a single line
• whitespaceMultiline: whitespace that containes at least a newline
• doubleQuotedString: escape-aware string delimited by ", returns the content
• singleQuotedString: escape-aware string delimited by ', returns the content
• backtickQuotedString: escape-aware string delimited by `, returns the content