@thi.ng/parse

[!NOTE] This is one of 199 standalone projects, maintained as part of the @thi.ng/umbrella monorepo and anti-framework.

🚀 Please help me to work full-time on these projects by sponsoring me on GitHub. Thank you! ❤️

• About
  • Features
• Status
• Related packages
• Installation
• Dependencies
• Usage examples
• API
  • Context & reader creation
  • Presets parsers
  • Primitives
    • Naming conventions / suffixes
  • Anchors
  • Combinators
  • Transformers
• Grammar definition
  • Terms
  • Repetitions
  • Discarding results
  • Lookahead
  • Term modifier ordering
  • Rule transforms
• Examples
  • S-expression DSL
  • SVG path parser example
  • RPN parser & interpreter example
• Authors
• License

About

Purely functional parser combinators & AST generation for generic inputs.

There's a 2h 45m long video tutorial (live stream) introducing this package: Building a web editor for creating/testing parse grammars

Features

• Parser generation/compilation via built-in grammar DSL
• Small API surface, easy-to-grok syntax
• All parsers implemented as composable, higher-order functions
• All state centrally kept/managed by a parser context given as arg
• Support for custom readers (currently only string & array-like numeric inputs (incl. typed arrays) supported)
• Automatic AST generation & ability to transform/prune nodes during parsing
• Node transforms are composable too
• Each AST node (optionally) retains reader information (position, line num, column) - disabled by default to save memory
• Common, re-usable preset parsers & node transforms included

Status

STABLE - used in production

Search or submit any issues for this package

Related packages

• @thi.ng/hiccup-markdown - Markdown parser & serializer from/to Hiccup format
• @thi.ng/proctext - Extensible procedural text generation engine with dynamic, mutable state, indirection, randomizable & recursive variable expansions
• @thi.ng/transducers-fsm - Transducer-based Finite State Machine transformer

Installation

yarn add @thi.ng/parse

ESM import:

import * as parse from "@thi.ng/parse";

Browser ESM import:

<script type="module" src="https://esm.run/@thi.ng/parse"></script>

JSDelivr documentation

For Node.js REPL:

const parse = await import("@thi.ng/parse");

Package sizes (brotli'd, pre-treeshake): ESM: 5.27 KB

Dependencies

• @thi.ng/api
• @thi.ng/checks
• @thi.ng/defmulti
• @thi.ng/errors
• @thi.ng/strings

Note: @thi.ng/api is in most cases a type-only import (not used at runtime)

Usage examples

Three projects in this repo's /examples directory are using this package:

Screenshot	Description	Live demo	Source
	Markdown to Hiccup to HTML parser / transformer	Demo	Source
	Mastodon API feed reader with support for different media types, fullscreen media modal, HTML rewriting	Demo	Source
	Parser grammar livecoding editor/playground & codegen	Demo	Source

Note: Please also see the dedicated wiki page collecting various grammar examples and links to their playgrounds.

API

Generated API docs

TODO

Context & reader creation

• defContext

Source: /readers

• defArrayReader
• defStringReader

Presets parsers

Source: /presets

• WS / WS0 / WS1 / NL / DNL / SPACE / SPACES / SPACES0
• ALPHA / LOWER_CASE / UPPER_CASE / ALPHA_NUM
• ESC / UNICODE
• DIGIT / DIGITS / DIGITS0
• HEX_DIGIT / HEX_DIGITS / HEX_UINT
• BIT / BINARY_UINT
• INT / UINT / REAL / FLOAT / SIGN

Primitives

Source: /prims

Naming conventions / suffixes

The D suffix is used for discarding behavior, i.e. parsers which do not retain their result in the AST after successful matching. These should be preferred whenever possible, for lower memory usage and (possibly) better performance...

The P suffix is used to indicate predicate based parsers, i.e. using user provided functions applied to individual inputs (characters) in order to decide about successful matching.

• always
• fail
• lit / litD / litP
• noneOf / noneOfD / noneOfP
• oneOf / oneOfD / oneOfP
• pass / passD
• range / rangeD / rangeP
• satisfy / satisfyD
• skipWhile
• string / stringD
• stringOf

Anchors

• anchor
• inputStart / inputEnd
• lineStart / lineEnd
• wordBoundary
• startsWith / endsWith
• entireLine
• entirely

Combinators

Source: /combinators

• alt / altS / altD
• check
• expect
• lookahead
• maybe
• not
• oneOrMore / zeroOrMore / oneOrMoreD / zeroOrMoreD
• repeat / repeatD
• seq / seqD
• xform

Transformers

Source: /xform

Syntax sugars for xform(parser, fn):

• collect - collect child results into array
• count - count number of children
• discard - discard result
• float / int / hexInt - join child results and parse as number
• hoist / hoistResult - hoist first child / child result
• join - join child results into string
• json - join child results into string and parse as JSON
• nest - apply another parser to result
• print - print AST
• replace - replace AST node result w/ pre-configured value
• trim - trim node result (string only)
• withID - assign custom AST node ID

Actual transforms:

• comp - scope transform composition
• xfCollect
• xfCount
• xfDiscard
• xfFloat
• xfHoist / xfHoistResult
• xfInt(radix)
• xfJoin
• xfNest
• xfPrint
• xfReplace
• xfTrim
• xfID

Grammar definition

Complex parsers can be constructed via defGrammar(), which accepts a string of rule definitions in the built-in (and still WIP) grammar rule definition language, similar to PEGs and regular expressions:

Example grammar

# line comment
ws:      ' '+ => discard ;
sym:     [a-z] [a-z0-9_]* => join ;
num:     [0-9]+ => int ;
program: ( <num> | <sym> | <ws> )* ;

Here, each line (apart from the first) is a single parse rule definition, with each rule consisting of a sequence of one or more:

Terms

• . - matches any single char
• 'x' - single char literal
• "abc" - multi-char string
• [a-z0-9_@] - regex style char set (incl. char range support, inversion via ^)
• <rule_id> - rule references (order independent)
• ( term | ... | ) - choice of sub-terms

Literals, strings and char sets can include \uXXXX unicode escapes (if given within a JS source string, double escaping must be used, i.e. \\uXXXX).

Repetitions

All of these terms can be immediately followed by one of these regexp style repetition specs:

• ? - zero or one occurrence
• * - zero or more
• + - one or more
• {n}, {min,} or {min,max} - fixed size, min-infinity or min-max repetitions

Discarding results

All terms can be suffixed with the ! modifier to discard their results and help produce a cleaner result AST.

link: '['! <linktitle> "]("! <linkurl> ')'! => collect ;
linktitle: [^\\u005d]+ => join ;
linkurl: [^\\u0029]+ => join ;

Given a valid input like [abc](def), the above link parser's result will just be the array ["abc", "def"] with other terms discarded from the AST.

Important: A parse rule MUST produce a result for at least one of its successfully matched terms. Use the discard rule transform to discard the entire result (instead of the ! modifier)...

Lookahead

All terms can be suffixed with an optional, capturing or non-capturing lookahead group:

• (?-...): non-capturing
• (?+...): capturing

# accept any character until end of current line
# <NL> is a built-in preset rule to match newline chars...
rest: .(?-<NL>!) ;

When using lookahead, both the main and look-ahead parsers are run repeatedly for as long as the former succeeds and UNTIL the latter passes or the end of input is reached. If the ahead parser never passes, the entire term fails and any partial matches are discarded.

Depending on capture behavior, the result of the ahead parser is captured or omitted and the final read position is adjusted accordingly.

Currently, iff capture is disabled, the ahead parser MUST discard its own result (see section above). On successful match the final read position will then be restored to the beginning of ahead pattern.

Iff capture is enabled, the ahead parser MAY discard its own result, but the final read position will not be re-adjusted as in the non-capturing version.

Important: Since the main term will be repeated automatically, DO NOT use repetition modifiers ? or *, since both of these will cause the parser to go into an infinite loop. This is expected behavior and not a bug.

Term modifier ordering

The above mentioned optional term modifiers MUST always be given in this order:

• Repetitions (?*+ etc.)
• Discard flag (!)
• Lookahead ((?-...) / (?+...))

[a-z]+!         # valid
[a-z]{4}!       # valid
[a-z]!          # valid
[a-z]!(?+[0-9]) # valid
[a-z]!+         # invalid
[a-z]!{4}       # invalid
[a-z](?+'X')!   # invalid

Rule transforms

Furthermore, each rule can specify an optional rule transform function, result string or even another parser rule, which will only be applied after the rule's parser has successfully completed. The transform is given at the end of a rule, separated by =>.

If another parser rule is specified (via <ruleid>), it will be applied to result of the main rule in a separate parse context and its own results will be transplanted back into the main AST.

If a result string is given (e.g. "foo"), it will be used as the rule's result instead and the node's children will be removed.

Custom transforms functions can be supplied via an additional arg to defGrammar(). The following default transforms are available by default (can be overwritten) and correspond to the above mentioned transforms:

• binary - parse as binary number
• collect - collect sub terms into array
• discard - discard result
• float - join & parse as floating point number
• hex - join & parse as hex integer
• hoist - replace AST node with its 1st child
• hoistR - use result of 1st child term only
• int - join & parse as integer
• join - join sub terms into single string
• json - join & parse as JSON
• print - print out node's subtree (AST)
• trim - trim result

For convenience, the following built-in parser presets are available as rule references in the grammar definition as well:

• ALPHA
• ALPHA_NUM
• BIT
• DIGIT
• DNL - discarded newline
• END - input end
• ESC - escape sequences
• FLOAT
• HEX_DIGIT
• INT
• LEND - line end
• LSTART - line start
• NL - newline chars
• START - input start
• STRING
• UNICODE
• WB - word boundary
• WS
• WS0
• WS1

Examples

S-expression DSL

(Also see @thi.ng/defmulti as a useful tool for processing/interpreting/compiling the result AST)

import { defContext, defGrammar, print } from "@thi.ng/parse";

// define language via grammar DSL
// the upper-cased rule names are built-ins
const lang = defGrammar(`
list: '('! <expr> ')'! ;
sym: ( <ALPHA_NUM> | [?!$+\\u002d*/.~#^=<>] )+ => join ;
expr: ( <FLOAT> | <STRING> | <sym> | <list> | <WS1> )* ;
`);

// define input & parser context
const ctx = defContext(`
(def hello (x) (str "hello, " x))

(print (hello 42))
`);

// parse & print AST
print(lang!.rules.expr)(ctx);
// expr: null
//   list: null
//     expr: null
//       sym: "def"
//       sym: "hello"
//       list: null
//         expr: null
//           sym: "x"
//       list: null
//         expr: null
//           sym: "str"
//           string: "hello, "
//           sym: "x"
//   list: null
//     expr: null
//       sym: "print"
//       list: null
//         expr: null
//           sym: "hello"
//           real: 42

// parse result
// true

// the two top-level s-expressions...
console.log(ctx.children);
// [
//   { id: 'list', state: null, children: [ [Object] ], result: null },
//   { id: 'list', state: null, children: [ [Object] ], result: null }
// ]

SVG path parser example

import {
    INT, WS0,
    alt, collect, defContext, discard,
    oneOf, seq, xform, zeroOrMore
} from "@thi.ng/parse";

// whitespace parser
// discard() removes matched result from AST
const wsc = discard(zeroOrMore(oneOf(" ,")));

// SVG path parser rules
// collect() collects child results in array, then removes children
// INT & WS0 are preset parsers (see readme section above)
const point = collect(seq([INT, wsc, INT]));
const move = collect(seq([oneOf("Mm"), WS0, point, WS0]));
const line = collect(seq([oneOf("Ll"), WS0, point, WS0]));
const curve = collect(
    seq([oneOf("Cc"), WS0, point, wsc, point, wsc, point, WS0])
);
// xform used here to wrap result in array
// (to produce same result format as parsers above)
const close = xform(
    oneOf("Zz"),
    (scope) => ((scope!.result = [scope!.result]), scope)
);

// main path parser
const path = collect(zeroOrMore(alt([move, line, curve, close])));

// prepare parse context & reader
const ctx = defContext("M0,1L2 3c4,5-6,7 8 9z");
// parse input into AST
console.log(path(ctx));
// true

// transformed result of AST root node
console.log(ctx.result);
// [
//   [ 'M', [ 0, 1 ] ],
//   [ 'L', [ 2, 3 ] ],
//   [ 'c', [ 4, 5 ], [ -6, 7 ], [ 8, 9 ] ],
//   [ 'z' ]
// ]

RPN parser & interpreter example

import type { Fn, FnN2 } from "@thi.ng/api";
import {
    INT, WS0,
    alt, altS, defContext, xform, zeroOrMore
} from "@thi.ng/parse";

type StackFn = Fn<number[], void>;
type Op = { arity: number; fn: StackFn };

// wrapper for pure 2-arity stack functions/ops
const defOp2 =
    (fn: FnN2): StackFn =>
    (stack) => {
        const b = stack.pop()!;
        const a = stack.pop()!;
        stack.push(fn(a, b));
    };

// operator/word implementations
const ops: Record<string, Op> = {
    "+": { arity: 2, fn: defOp2((a, b) => a + b) },
    "-": { arity: 2, fn: defOp2((a, b) => a - b) },
    "*": { arity: 2, fn: defOp2((a, b) => a * b) },
    "/": { arity: 2, fn: defOp2((a, b) => a / b) },
    // prints top stack item to console
    print: { arity: 1, fn: (stack) => console.log(stack.pop()) },
    // duplicates top stack item
    dup: { arity: 1, fn: (stack) => stack.push(stack[stack.length - 1]) },
    // swaps two topmost stack items
    swap: {
        arity: 2,
        fn: (stack) => {
            const a = stack.pop()!;
            const b = stack.pop()!;
            stack.push(a, b);
        },
    },
};

// simple RPN parser & interpreter runtime
const interpret = (src: string, debug = true) => {
    // data stack for execution
    const stack: number[] = [];

    // signed integer parser (using INT preset) with transform fn.
    // the user fn here is only used for pushing values on data stack
    // also, if a transform returns null, the parse scope will
    // be removed from the result AST
    const value = xform(INT, (scope) => {
        stack.push(scope!.result);
        return null;
    });

    // parser (with transform) for any of the registered operators
    // the transform here applies the op in RPN fashion to the data stack
    // stack underflow handling omitted for brevity
    const op = xform(altS(Object.keys(ops)), (scope) => {
        const id = scope!.result;
        const { arity, fn } = ops[id];
        if (debug) console.log(id, stack);
        if (stack.length < arity) {
            throw new Error(
                "stack underflow " +
                `("${id}" needs ${arity} args, got ${JSON.stringify(stack)})`
            );
        }
        fn(stack);
        return null;
    });

    // parser for complete RPN program, combines above two parsers
    // and the whitespace preset as alternatives
    const program = zeroOrMore(alt([value, op, WS0]));
    // apply parser to source code
    program(defContext(src));
    // return result stack
    return stack;
};

// checking operator arity
try { interpret("1 +"); } catch (e) { console.warn(e); }
// Error: stack underflow ("+" needs 2 args, got [1])

// execute: (3 * 5 + 10) * -2 / 10 (in infix notation)
interpret("10 5 3 * + -2 * 10 / print");
// * [ 10, 5, 3 ]
// + [ 10, 15 ]
// * [ 25, -2 ]
// / [ -50, 10 ]
// print [ -5 ]
// -5

// execute: ((5 + 3)**2) / 2) - (5 + 3)**2
interpret("5 3 + dup * dup 2 / swap - print");
// + [ 5, 3 ]
// dup [ 8 ]
// * [ 8, 8 ]
// dup [ 64 ]
// / [ 64, 64, 2 ]
// swap [ 64, 32 ]
// - [ 32, 64 ]
// print [ -32 ]
// -32

Authors

• Karsten Schmidt

If this project contributes to an academic publication, please cite it as:

@misc{thing-parse,
  title = "@thi.ng/parse",
  author = "Karsten Schmidt",
  note = "https://thi.ng/parse",
  year = 2020
}

License

© 2020 - 2024 Karsten Schmidt // Apache License 2.0