Single Error mode only allowing exact matches under 5 characters

[ascendedguard]

Common use case I'm running into: Searching a list of names with "Nicholas", however searching for "Nick" does not work even though it's only 1 character off. Searching for "Nicko" however will find it. intraMode is set to 1 (SingleError)

You can see this on the demo by searching "Niko Eymerich":

While adding a 5th letter with "Nikol Eymerich" will show the correct results:

Using latest version 1.0.14

[leeoniya]

yeah, by default there are stricter requirements for short terms, because you generally dont want "oct" to match "cot" and "ot" substrings.

you can modify these defaults on a per-term basis using the intraRules option, (the defaults are linked below)

more discussion in: #39, #44

i'm certainly open to improving the defaults when it makes sense. for example, if you only want to do prefix matching for each term (interLft: 1 or interLft: 2) rather than arbitrary substrings, we can relax the rules for short terms. but as i've said in the other issues, i'm quite hesitant to introduce additional custom short-term behavior that must be understood under different options combos.

uFuzzy/src/uFuzzy.js

Lines 156 to 198 in ec1c44b

	intraRules = p => {
	// default is exact term matches only
	let _intraSlice = OPTS.intraSlice, // requires first char
	_intraIns = 0,
	_intraSub = 0,
	_intraTrn = 0,
	_intraDel = 0;
	// only-digits strings should match exactly, else special rules for short strings
	if (/[^\d]/.test(p)) {
	let plen = p.length;
	// prevent junk matches by requiring stricter rules for short terms
	if (plen <= 4) {
	if (plen >= 3) {
	// one swap in non-first char when 3-4 chars
	_intraTrn = Math.min(intraTrn, 1);
	// or one insertion when 4 chars
	if (plen == 4)
	_intraIns = Math.min(intraIns, 1);
	}
	// else exact match when 1-2 chars
	}
	// use supplied opts
	else {
	_intraSlice = intraSlice;
	_intraIns = intraIns,
	_intraSub = intraSub,
	_intraTrn = intraTrn,
	_intraDel = intraDel;
	}
	}
	return {
	intraSlice: _intraSlice,
	intraIns: _intraIns,
	intraSub: _intraSub,
	intraTrn: _intraTrn,
	intraDel: _intraDel,
	};
	};
	}

[ascendedguard]

Thanks, this was what I needed.

This is basically what I ended up using, with some special shenanigans so that the method return played nice in Typescript.

// custom uFuzzy rules for short strings, assuming intraMode = 1
const uFuzzyIntraRules = (p: string) => { 
    // note: this is assuming intraMode = 1
    const settings: {
        intraSlice: IntraSliceIdxs;
        intraIns: 0 | 1;
        intraSub: 0 | 1;
        intraTrn: 0 | 1;
        intraDel: 0 | 1;
    } = {
        intraSlice: [1, Infinity],
        intraIns: 1,
        intraSub: 1,
        intraTrn: 1,
        intraDel: 1,
    }

    // only-digits strings should match exactly, else special rules for short strings 
    if (/[^\d]/.test(p)) { 
        const plen = p.length;

        // some more flexible searching for 3/4 letter terms than what default allowed
        if (plen === 3 || plen === 4) {
            settings.intraSlice = [2, Infinity];
            settings.intraTrn = 0;
            settings.intraIns = 0;
            settings.intraDel = 0;
        } 
    }

    return settings;
}

const opts: uFuzzy.Options = {
    intraMode: 1, // Single errors allowed,
    intraRules: uFuzzyIntraRules,
    interLft: 2, // Loose left word boundary
};
const fuzzySearch = new uFuzzy(opts);

I may do a PR soon just with some minor documentation suggestions from figuring out the library over the last day. Thanks for the library!

[leeoniya]

👍