glob-specificity v0.2.0

glob-specificity

Define and sort by how specific a glob pattern is.

The specificness of a glob pattern can be very hard to define. This approach aims for ~90% accuracy and only supports the following glob features:

POSIX character classes and Extended glob patterns are not supported.

Installation

npm install glob-specificity

Getting started

Sort an array of glob patterns by their specificty. Will sort from least specific to most specific.

import { globSpecificitySort } from 'glob-specificity';

globSpecificitySort(['foo/bar', 'foo/*', 'foo/**']);
// [ 'foo/**', 'foo/*', 'foo/bar' ]

Calculate specificy first, sort at a later time.

import { globSpecificity, sortByGlobSpecificity } from './dist/glob-specificity.mjs';

const patterns = ['foo/bar', '*/bar', 'foo/*'];
const patternsWithSpecificity = patterns.map((pattern) => {
  return {
    glob: pattern,
    specificity: globSpecificity(pattern),
  };
});

// Do other things

sortByGlobSpecificity(patternsWithSpecificity);
// [
//     { glob: 'foo/*', specificity: [ 5, 0, 0, 0, 0 ] },
//     { glob: '*/bar', specificity: [ 7, 0, 0, 0, 0 ] },
//     { glob: 'foo/bar', specificity: [ 8, 0, 0, 0, 0 ] }
// ]

Definition of specificity

It is hard to pin down what specificity is for glob patterns. Not everyone will agree with all parts of this definition. If you disagree with any part of this definition, please let me know via an issue.

We detemine how specific a glob pattern is by 5 metrics. In order of importance:

1. Segment mask
   • Segment count
   • Segment types
2. Number of star wildcards *
3. Number of questionmark wildcards ?
4. Number of characters in character matcher [chars]
5. Number of subpatterns {a,b,...}

1. Segment mask

First we seperate the pattern into segments. A pattern with more segments is more specific. Each segment is assigned one of the following types:

• Globstar segment **
• Star-segment *
• Other segment

Globstar segments are least specific and other segments are most specific.

'foo/*/bar/**'
-> ['foo', '*', 'bar', '**'] // Separate segments
-> [other, star, other, globStar] // Assign segments a type
-> [2, 1, 2, undefined] // Assign types a numeric value
-> '212' // Join values into segment mask
-> 23 // Convert base3 -> base10

2. Number of star wildcards *

Number of star * characters in a pattern. Star * characters in star segments and globstar segments are ignored (see segment types).

'foo/*/b*r/**/*or*m'
-> ['foo', 'b*r', '*or*m'] // Remove star and globstar segments
-> [0, 1, 2] // Count star characters
-> 3 // Sum

Escaped stars \* are ignored.

3. Number of questionmark wildcards ?

Number of questionmark ? characters in a pattern.

'foo/b?r/l?r?m'
-> 3 // Count questionmark characters

Escaped questionmarks \? are ignored.

4. Number of characters in character matcher [chars]

Number of characters in a character matcher.

'[f]oo/[Bb]ar/loru[mnf]'
-> ['[f]', '[Bb]', '[mnf]'] // Find matchers
-> [0, 1, 2] // Count characters - 1
-> 3 // Sum

-1 because [f]oo is as specific as foo.

A range is expanded before counting, so [a-f] counts as 6 characters.

'foo/ba[a-f]'
-> ['[a-f]'] // Find matchers
-> ['[abcdef]'] // Expand range
-> [5] // Count characters - 1
-> 5 // Sum

Escaped ranges \[chars] are ignored.

5. Number of subpatterns {a,b,...}

Number of subpatterns in a pattern.

'{baz}/foo/{bar,lorum}/{alpha,beta,gamma}'
-> ['{baz}', '{bar,lorum}', '{alpha,beta,gamma}'] // Find subpatterns
-> [0, 1, 2] // Count patterns - 1
-> 3 // Sum

-1 because {baz} is as specific as baz.

Escaped subpatterns \{a,b,...} are ignored.