🐛 Fixed escaping search terms that contain special characters

[markstos]

• There's a clear use-case for this code change, explained below
• Commit message has a short title & references relevant issues
• The build will pass (run yarn test:all and yarn lint)

Before, Sodo Search was not escaping search input before using the search terms in a regular expression, so using special characters could result in an invalid regular expression which would crash JavaScript.

As regular expressions date back to Perl, so does a standard solution for this, which called quotemeta in Perl. It doesn't exist 1:1 in JavaScript, but StackOverflow had the answer:

https://stackoverflow.com/questions/6318710/javascript-equivalent-of-perls-q-e-or-quotemeta

... where it appears I answered this question in 2014. Ha.

So a line of code is added to escape the special characters in the regex for passing them through. This is the same code that the quotemeta module on NPM would use.

Fixes: #18133

QA Log

Unfortunately, my local Ghost development environment is not set up perfectly to test this. I tried using ghost dev --search and that did link a local blog to a local version of Sodo search, but a new JavaScript error appears which I think is unrelated:

search-index.js:12 Uncaught TypeError: b.Document is not a constructor
    at new cm (search-index.js:12:27)

So I'd love if someone with a working dev environment could quickly repro the issue with a "+C" and then apply my match and test again.

Thanks!

[JaydenForYou]

I have already verified it.bug fixed.

[daniellockyer]

@markstos Seems like flexsearch has broken during the 0.7.21 -> 0.7.31 upgrade, and we haven't published a release of the package since. I'll revert that package now 🙂

[markstos]

After rebasing this branch off the lastest main with flexsearch revert, I'm able to confirm that a search for "+C" no longer fails and indeeds succeeds.

[daniellockyer]

Thanks @markstos! 🙂