TS-Tokenizr

Flexible String Tokenization Library for TypeScript

Inspiration

I found this gem of a library after writing my own for a project. It was full of features, and I just didn't properly know how to implement my own. It worked, but it was... "dumb". I decided to use it in my project, and along the way, happened to convert the whole thing into TypeScript :laughing:

It was a fun exercise and really helped me understand how it worked.

Thanks for the great library rse, cheers! :beers:

About

Tokenizr is a small TypeScript library, providing powerful and flexible string tokenization functionality. It is intended to be be used as the underlying "lexical scanner" in a Recursive Descent based "syntax parser", but can be used for other parsing purposes, too. Its distinct features are:

• Efficient Iteration: It iterates over the input character string in a read-only and copy-less fashion.

• Stacked States: Its tokenization is based on stacked states for determining rules which can be applied. Each rule can be enabled for one or more particular states only.

• Regular Expression Matching: Its tokenization is based on powerful Regular Expressions for matching the input string.

• Match Repeating: Rule actions can change the state and then force the repeating of the matching process from scratch at the current input position.

• Match Rejecting: Rule actions can reject their matching at the current input position and let subsequent rules to still match.

• Match Ignoring: Rule actions can force the matched input to be ignored (without generating a token at all).

• Match Accepting: Rule actions can accept the matched input and provide one or even more resulting tokens.

• Shared Context Data: Rule actions (during tokenization) can optionally store and retrieve arbitrary values to/from their tokenization context to share data between rules.

• Token Text and Value: Tokens provide information about their matched input text and can provide a different corresponding (pre-parsed) value, too.

• Debug Mode: The tokenization process can be debugged through optional detailed logging of the internal processing.

• Nestable Transactions: The tokenization can be split into distinct (and nestable) transactions which can be committed or rolled back. This way the tokenization can be incrementally stepped back and this way support the attempt of parsing alternatives.

• Token Look-Ahead: The forthcoming tokens can be inspected, to support alternative decisions from within the parser, based on look-ahead tokens.

Installation

Node environments (with NPM package manager):

$ npm install ts-tokenizr

Usage

Suppose we have a configuration file sample.cfg:

foo {
    baz = 1 // sample comment
    bar {
        quux = 42
        hello = "hello \"world\"!"
    }
    quux = 7
}

Then we can write a lexical scanner in ECMAScript 6 (under Node.js) for the tokens like this:

import fs       from "fs"
import Tokenizr from "ts-tokenizr"

let lexer = new Tokenizr({
  debug: true
})

lexer.rule(/[a-zA-Z_][a-zA-Z0-9_]*/, (ctx, match) => {
    ctx.accept("id")
})
lexer.rule(/[+-]?[0-9]+/, (ctx, match) => {
    ctx.accept("number", parseInt(match[0]))
})
lexer.rule(/"((?:\\"|[^\r\n])*)"/, (ctx, match) => {
    ctx.accept("string", match[1].replace(/\\"/g, "\""))
})
lexer.rule(/\/\/[^\r\n]*\r?\n/, (ctx, match) => {
    ctx.ignore()
})
lexer.rule(/[ \t\r\n]+/, (ctx, match) => {
    ctx.ignore()
})
lexer.rule(/./, (ctx, match) => {
    ctx.accept("char")
})

let cfg = fs.readFileSync("sample.cfg", "utf8")

lexer.input(cfg)
lexer.tokens().forEach((token) => {
    console.log(token.toString())
})

The output of running this sample program is:

<type: id, value: "foo", text: "foo", pos: 0, line: 1, column: 1>
<type: char, value: "{", text: "{", pos: 4, line: 1, column: 5>
<type: id, value: "baz", text: "baz", pos: 10, line: 2, column: 5>
<type: char, value: "=", text: "=", pos: 14, line: 2, column: 9>
<type: number, value: 1, text: "1", pos: 16, line: 2, column: 11>
<type: id, value: "bar", text: "bar", pos: 40, line: 3, column: 5>
<type: char, value: "{", text: "{", pos: 44, line: 3, column: 9>
<type: id, value: "quux", text: "quux", pos: 54, line: 4, column: 9>
<type: char, value: "=", text: "=", pos: 59, line: 4, column: 14>
<type: number, value: 42, text: "42", pos: 61, line: 4, column: 16>
<type: id, value: "hello", text: "hello", pos: 72, line: 5, column: 9>
<type: char, value: "=", text: "=", pos: 78, line: 5, column: 15>
<type: string, value: "hello \"world\"!", text: "\"hello \\\"world\\\"!\"", pos: 80, line: 5, column: 17>
<type: char, value: "}", text: "}", pos: 103, line: 6, column: 5>
<type: id, value: "quux", text: "quux", pos: 109, line: 7, column: 5>
<type: char, value: "=", text: "=", pos: 114, line: 7, column: 10>
<type: number, value: 7, text: "7", pos: 116, line: 7, column: 12>
<type: char, value: "}", text: "}", pos: 118, line: 8, column: 1>
<type: EOF, value: "", text: "", pos: 122, line: 9, column: 1>

If you want to combine multiple single-char plaintext tokens into a multi-char plaintext token, you can use the following code fragment:

let plaintext = ""
lexer.before((ctx, match, rule) => {
    if (rule.name !== "plaintext" && plaintext !== "") {
        ctx.accept("plaintext", plaintext)
        plaintext = ""
    }
})
lexer.rule(/./, (ctx, match) => {
    plaintext += match[0]
    ctx.ignore()
}, "plaintext")
lexer.finish((ctx) => {
    if (plaintext !== "")
        ctx.accept("plaintext", plaintext)
})

With the additional help of an Abstract Syntax Tree (AST) library like ASTy and a query library like ASTq you can write powerful Recursive Descent based parsers which parse such a token stream into an AST and then query and process the AST.

Application Programming Interface (API)

Class Tokenizr

This is the main API class for establishing a lexical scanner.

• Constructor: Tokenizr(): Tokenizr Create a new tokenization instance.

• Method: Tokenizr#reset(): Tokenizr Reset the tokenization instance to a fresh one by discarding all internal state information.

• Method: Tokenizr#debug(enable: Boolean): Tokenizr Enable (or disable) verbose logging for debugging purposes.

• Method: Tokenizr#input(input: String): Tokenizr Set the input string to tokenize. This implicitly performs a reset() operation beforehand.

• Method: Tokenizr#push(state: String): Tokenizr Push a state onto the state stack.

• Method: Tokenizr#pop(): String Pop a state from the state stack. The initial (aka first or lowest) stack value (default) cannot be popped.

• Method: Tokenizr#state(state: String): Tokenizr Method: Tokenizr#state(): String Set or get the state on the top of the state stack. Use this to initialy start tokenizing with a custom state. The initial state is named default.

• Method: Tokenizr#tag(tag: String): Tokenizr Set a tag. The tag has to be matched by rules.

• Method: Tokenizr#tagged(tag: String): Boolean Check whether a particular tag is set.

• Method: Tokenizr#untag(tag: String): Tokenizr Unset a particular tag. The tag no longer will be matched by rules.

• Method: Tokenizr#before(action: (ctx: ActionContext, match: Array[String], rule: { state: String, pattern: RegExp, action: Function, name: String }) => Void): Tokenizr Configure a single action which is called directly before any rule action (configured with Tokenizr#rule()) is called. This can be used to execute a common action just before all rule actions. The rule argument is the Tokenizr#rule() information of the particular rule which is executed.

• Method: Tokenizr#after(action: (ctx: ActionContext, match: Array[String], rule: { state: String, pattern: RegExp, action: Function, name: String }) => Void): Tokenizr Configure a single action which is called directly after any rule action (configured with Tokenizr#rule()) is called. This can be used to execute a common action just after all rule actions. The rule argument is the Tokenizr#rule() information of the particular rule which is executed.

• Method: Tokenizr#finish(action: (ctx: ActionContext) => Void): Tokenizr Configure a single action which is called directly before an EOF token is emitted. This can be used to execute a common action just after the last rule action was called.

• Method: Tokenizr#rule(state?: String, pattern: RegExp, action: (ctx: ActionContext, match: Array[String]) => Void): Tokenizr Configure a token matching rule which executes its action in case the current tokenization state is one of the states (and all of the currently set tags) in state (by default the rule matches all states if state is not specified) and the next input characters match against the pattern. The exact syntax of state is <state>[ #<tag> #<tag> ...][, <state>[ #<tag> #<tag> ...], ...], i.e., it is one or more comma-separated state matches (OR-combined) and each state match has exactly one state and zero or more space-separated tags (AND-combined). The ctx argument provides a context object for token repeating/rejecting/ignoring/accepting, the match argument is the result of the underlying RegExp#exec call.

• Method: Tokenizr#token(): Tokenizr.Token Get the next token from the input. Internally, the current position of the input is matched against the patterns of all rules (in rule configuration order). The first rule action which accepts the matching leads to the token.

• Method: Tokenizr#tokens(): Array[Tokenizr.Token] Tokenizes the entire input and returns all the corresponding tokens. This is a convenience method only. Usually one takes just single tokens at a time with Tokenizr#token().

• Method: Tokenizr#skip(next?: Number): Tokenizr Get and discard the next number of following tokens with Tokenizr#token().

• Method: Tokenizr#consume(type: String, value?: String): Tokenizr Match (with Tokenizr.Token#isA) the next token. If it matches type and optionally also value, consume it. If it does not match, throw a Tokenizr.ParsingError. This is the primary function used in Recursive Descent parsers.

• Method: Tokenizr#peek(offset?: Number): Tokenizr.Token Peek at the following token at the (0-based) offset without consuming the token. This is the secondary function used in Recursive Descent parsers.

• Method: Tokenizr#begin(): Tokenizr Begin a transaction. Until Tokenizr#commit() or Tokenizr#rollback() are called, all consumed tokens will be internally remembered and be either thrown away (on Tokenizr#commit()) or pushed back (on Tokenizr#rollback()). This can be used multiple times and this way supports nested transactions. It is intended to be used for tokenizing alternatives.

• Method: Tokenizr#depth(): Number Return the number of already consumed tokens in the currently active transaction. This is useful if multiple alternatives are parsed and in case all failed, to report the error for the most specific one, i.e., the one which consumed most tokens.

• Method: Tokenizr#commit(): Tokenizr End a transaction successfully. All consumed tokens are finally gone.

• Method: Tokenizr#rollback(): Tokenizr End a transaction unsuccessfully. All consumed tokens are pushed back and can be consumed again.

• Method: Tokenizr#alternatives(...alternatives: Array[() => any]): any Utility method for parsing alternatives. It internally executes the supplied callback functions in sequence, each wrapped into its own transaction. The first one which succeeds (does not throw an exception and returns a value) leads to the successful result. In case all alternatives failed (all throw an exception), the exception of the most-specific alterative (the one with the largest transaction depth) is re-thrown. The this in each callback function points to the Tokenizr object on which alternatives was called.

• Method: Tokenizr#error(message: String): Tokenizr.ParsingError Returns a new instance of Tokenizr.ParsingError, based on the current input character stream position, and with Tokenizr.ParsingError#message set to message.

Class Tokenizr.Token

This is the class of all returned tokens.

• Property: Tokenizr.Token#type: String The type of the token as specified on Tokenizr.ActionContext#accept().

• Property: Tokenizr.Token#value: any The value of the token. By default this is the same as Tokenizr.Token#text, but can be any pre-processed value as specified on Tokenizr.ActionContext#accept().

• Property: Tokenizr.Token#text: String The corresponding input text of this token.

• Property: Tokenizr.Token#pos: Number The (0-based) position in the input.

• Property: Tokenizr.Token#line: Number The (1-based) line number in the input.

• Property: Tokenizr.Token#column: Number The (1-based) column number in the input.

• Method: Tokenizr.Token#toString(): String Returns a formatted representation of the token, usually for debugging or tracing purposes only.

• Method: Tokenizr.Token#isA(type: String, value?: any): String Checks whether token matches against a particular type and optionally a particular value. This is especially used internally by Tokenizr#consume().

Class Tokenizr.ParsingError

This is the class of all thrown exceptions related to parsing.

• Property: Tokenizr.ParsingError#name: String Always just the string ParsingError to be complaint to the JavaScript Error class specification.

• Property: Tokenizr.ParsingError#message: String The particular error message.

• Property: Tokenizr.ParsingError#pos: Number The (0-based) position in the input.

• Property: Tokenizr.ParsingError#line: Number The (1-based) line number in the input.

• Property: Tokenizr.ParsingError#column: Number The (1-based) column number in the input.

• Property: Tokenizr.ParsingError#input: String The total input itself.

• Method: Tokenizr.ParsingError#toString(): String Returns a formatted representation of the error, usually for convenient error displaying purposes.

Class Tokenizr.ActionContext

This is the class of all rule action contexts.

• Method: Tokenizr.ActionContext#data(key: String, value?: any): any Store or retrieve any user data (indexed by key) to the action context for sharing data between two or more rules.

• Method: Tokenizr.ActionContext#info(): { line: number, column: number, pos: number, len: number } Retrieve information about the current matching.

• Method: Tokenizr.ActionContext#push(state: String): Tokenizr Method: Tokenizr.ActionContext#pop(): String Method: Tokenizr.ActionContext#state(state: String): Tokenizr.ActionContext Method: Tokenizr.ActionContext#state(): String Method: Tokenizr.ActionContext#tag(tag: String): Tokenizr.ActionContext Method: Tokenizr.ActionContext#tagged(tag: String): Boolean Method: Tokenizr.ActionContext#untag(tag: String): Tokenizr.ActionContext Methods just passed-through to the attached Tokenizr object. See above for details.

• Method: Tokenizr.ActionContext#repeat(): Tokenizr.ActionContext Mark the tokenization process to repeat the matching at the current input position from scratch. You first have to switch to a different state with Tokenizr.ActionContext#state() or this will lead to an endless loop, of course!

• Method: Tokenizr.ActionContext#reject(): Tokenizr.ActionContext Mark the current matching to be rejected. The tokenization process will continue matching following rules.

• Method: Tokenizr.ActionContext#ignore(): Tokenizr.ActionContext Mark the current matching to be just ignored. This is usually used for skipping whitespaces.

• Method: Tokenizr.ActionContext#accept(type: String, value?: any): Tokenizr.ActionContext Accept the current matching and produce a token of type and optionally with a different value (usually a pre-processed variant of the matched text). This function can be called multiple times to produce one or more distinct tokens in sequence.

• Method: Tokenizr.ActionContext#stop(): Tokenizr.ActionContext Immediately stop entire tokenization. After this the Tokenizr#token() method immediately starts to return null.

RegExp Flag Support

The pattern passed to Tokenizr.{before,after,rule}() has to be a regular JavaScript RegExp objects. Internally, Tokenizr creates a copy of this object by skipping its g (global) and y (sticky) flags and taking over its m (multiline), s (dotAll), i (ignoreCase), and u (unicode) flags.

Implementation Notice

Although Tokenizr is written in ECMAScript 6, it is transpiled to ECMAScript 5 and this way runs in really all(!) current (as of 2015) JavaScript environments, of course.

Internally, Tokenizr scans the input string in a read-only fashion by leveraging RegExp's g flag (global, for ECMAScript <= 5 environments) or y flag (sticky, for ECMAScript >= 2015 environments) in combination with RegExp's lastIndex field.

Alternatives

The following alternatives are known:

• moo: A very powerful tokenizer/lexer. It provides nearly the same functionality than Tokenizr. In addition, it compiles all regular expressions into a single one and hence is one of the fastest tokenizers/lexers.

• flex-js A medium powerful tokenizer/lexer. It provides nearly the same functionality than Tokenizr.

• lex: A tokenizer/lexer, modeled after the popular C solution Flex. This small library is similar in spirit, but not as flexible as it does not provide state supports.

License

Copyright (c) 2015-2020 Dr. Ralf S. Engelschall (http://engelschall.com/)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.