Incorrect quote string matching

Overview

The quotation string matching system matches incorrectly for adjacent strings.

Example

const nlp = require('compromise');

// Try splitting the string into quoted strings.
console.log(
    nlp('My "String" "with many" adjacent "nested" \'quotes\'')
    .quotations()
    .out('array')
)

Output

[
    "string with many",
    "nested quotes"
]

Expected

[
    "string",
    "with many",
    "nested",
    "quotes"
]

Workaround

So, I solved this for my specific circumstance like so:

function quotes(string) {
    // Match a single word surrounded by (")
    return string.match(
        '/^"[^"\\\\]*(?:\\\\.[^"\\\\]*)*"$/'
    )
    
    // Match a single word surrounded by (')
    .concat(string.match(
        '/^\'[^\'\\\\]*(?:\\\\.[^\'\\\\]*)*\'$/'
    ))
    
    
    // Match a two or more word surrounded by (")
    .concat(string.match(
        '/^"[^"\\\\]*(?:\\\\.[^"\\\\]*)*$/ ' +
        '/[^"\\\\]*(?:\\\\.[^"\\\\]*)*/+? ' +
        '/^[^"\\\\]*(?:\\\\.[^"\\\\]*)*"$/'
    ))
    
    // Match a two or more word surrounded by (')
    .concat(string.match(
        '/^\'[^\'\\\\]*(?:\\\\.[^\'\\\\]*)*$/ ' +
        '/[^\'\\\\]*(?:\\\\.[^\'\\\\]*)*/+? ' +
        '/^[^\'\\\\]*(?:\\\\.[^\'\\\\]*)*\'$/'
    ));
}

Using the RegExp by Jeffrey Friedl provided by ridgerunner on stackoverflow

Edit 1 I will note my solution is not perfect as it outputs:

[
    "string",
    "nested",
    "quotes",
    "with many"
]

End: Edit 1

A side note: The regex on the match syntax wiki page should be: /rain(ing|ed)/ meaning match rain and (ing or ed). The current /rain[ing|ed]/ means match rain and (i, n, g, | e or d).

[spencermountain]

ah, thanks Sca. This is a great bug. and your workaround is very clever. i think i’ve fixed this in 11.5.2 - at least your example works. Can you take a look? There’s all sorts of issues with, unicode quoations, um and the trailing possessive apostrophe inside a quote, which might trip it still. cheers

[spencermountain]

hey, yeah thanks @Errogant. I’ve made a new issue here