Update extractor-trim.js

[DK013]

fix for error: exec is not a function

Pull Request

PR Checklist

• [✓] I have run npm test locally and all tests are passing.
• [✓] [(https://github.com/ TypeError: condition.regex.exec is not a function  #463)]

PR Description

#463
In production if someone tries to load model nlp/json file instead of trainning from corpus, the match between function in extractor-trim.js under @nlpjs/ner package throws error: TypeError: condition.regex.exec is not a function
The reason of this error is when saved in a json file the generated regex is saved as a string instead of regex data type.

FIX:

• check if condition.regex is a string
• modify the string to be used in RegExp function so it doesn't escape characters unneccesserily
• convert condition.regex to regex
• in training mode if condition.regex is already regex, the program executes as it should normally

[Apollon77]

Beside the minimal code reveiw stuff I would more ask if not the place where we ADD the condition is the best place to do this conversion once instead of here maaaany times because it is done on every match try ....

[sonarqubecloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
2 Code Smells

No Coverage information
0.0% Duplication

[DK013]

Beside the minimal code reveiw stuff I would more ask if not the place where we ADD the condition is the best place to do this conversion once instead of here maaaany times because it is done on every match try ....

I am aware of that fact. However I'm not sure of the source where the condition variable is populated. And I don't think placing the conversion where matchBetween function is called will change the current outcome. If you can just guide me to where the variable is populated I can simple move the code there so the conversion only happens once.

[Apollon77]

The code how to add a regex rule is here

nlp.js/packages/ner/src/ner.js

Lines 239 to 247 in 032d409

	addRegexRule(locale, name, srcRegex) {
	const regex =
	typeof srcRegex === 'string' ? Ner.str2regex(srcRegex) : srcRegex;
	const globalFlag = 'g';
	const fixedRegex = regex.flags.includes(globalFlag)
	? regex
	: new RegExp(regex.source, `${regex.flags}${globalFlag}`);
	this.addRule(locale, name, 'regex', fixedRegex);
	}

Adding a Between condition is done in

nlp.js/packages/ner/src/ner.js

Line 269 in 032d409

addBetweenCondition(locale, name, srcLeftWords, srcRightWords, srcOptions) {

and basically in the end addRukeis used to add new rules

If your case is more the "restore from json" then

nlp.js/packages/ner/src/ner.js

Line 511 in 032d409

fromJSON(json) {

might be the place to check if all is initialized correctly there

[AndersDJohnson]

Hi @DK013 @Apollon77 this might be fixed by #1314

[Apollon77]

@ericzon I would close this PR in favor of #1314!!

[ericzon]

As @Apollon77 suggested, closed in favor of #1314