A library to aide in construction of complex regular expressions that are easier to maintain and read. Install with "npm install regex_grammar".

import Grammar from 'regex_grammar';
// Note: This is for demonstration purposes only. URLSearchParams is a better approach for this particular example of parsing urls.
const URL_GRAMMAR = {
    TOP: "$proto? $domain $path? $params? $bookmark?", // In string form, all whitespace is ignored. Specifically " " is treated as "\s*"
    proto: "$prototype://",
    prototype: ["ssh","https?","git","ftp"], // Will match any of these values
    domain: /[\w\.]+/,
    path: new RegExp("[\\w\\-/]+"), // Recall: String form requires escaping of '/'
    params: /\?((?:\w+)=(?:\w+)&?)+/,
    bookmark: /#.+/
};
const urlGrammar = new Grammar(URL_GRAMMAR);
urlGrammar.describe();
console.log(urlGrammar.match("http://foo.bar.com/api/do?param=something&more=less#foo"));

This library is inspired in part by Perl6/Raku's Grammars and the regular-grammar library. Test cases have been directly adapted from the regular-grammar library.

See index.mjs for additional documentation (TODO: Export to HTML), and test.mjs for examples.

Key Features

• 'TOP' always defines the top level of a Grammar, which is the default entry point for nominal matching.
• A grammar key can be referenced from any other key using the syntax '$key'
  • Cyclic references are not supported. ie: Rule 'foo' can't refer to rule 'bar' if it refers back to foo.
  • If a rule does not exist, no substitution will be performed and a warning will be printed. Note: User may need to explicitly escape the '$' in this case for a literal match, for example '\$foo' or /$foo/ to match 'foo', otherwise regex may interpret the '$' as matching the beginning of string in some cases.
• The 'match' function will return results equivalent to String.match. This result set may be extended in future versions with additional functionality.
• At present all grammar rules must be defined as one of the following.
  • A regular expression can be defined in the normal ways as "/foo/" or "new RegExp(...)"
  • A string will be treated as a pattern used to create a new RegExp, but where whitespace is insignificant. Specifically, any whitespace in the input string will become "\s*" in the built regular expression.
  • An array will be treated as a set of rules to match against.
    • Contents of the array may be RegExp objects, or strings.
    • Note: Whitespace substitution does not apply to strings in this case.
    • Variable substitution for referenced rules will apply.
    • Each element will be wrapped in a non-capturing group to ensure reliability.
    • A rule such as "['alpha','beta']" is equivalent to writing "(?:alpha)|(?:beta)".
• Named capture groups will be automatically created for each rule/token.
  • Group names will be prefixed with the name of the parent group as applicable, and suffixed with an integer if not unique (ie: for repeated patterns)

Future Enhancement Thoughts

• Ability to escape rule variable ('$') and/or specify an alternate delimiter.
• Parse named capture groups after match into a tree-like hierarchy reflecting the original rule definition.
• Debug options to trace Regex evaluation, if possible. At a minimum, expand describe function to provide a more useful view.
• Ability to define arbitrary subrules.
  • This may take the form of a third type of rule definition consisting of an object defining
    • A RegExp rule to match, as with standard rules
    • A discrete Grammar Class, rule key, or custom function to parse this nested result
  • This functionality would be dependent on extending Grammar match results with tree-like representation of automatic capture groups
  • Subrule matches would be folded into Grammar.match results