Errors

SearchConfigError

class SearchConfigError extends Error {
	override name = "SearchConfigError";
}

The only error type the package throws. It extends Error and sets name to "SearchConfigError", so instanceof Error and instanceof SearchConfigError both work, and the name is preserved through logging even after minification.

It is thrown for every pipeline misconfiguration:

Situation	Thrown from
Registering an algorithm name that is already registered	.register()
A dataset that is not an array	.defineDataset()
A dataset containing a non-string entry	.defineDataset()
A runOrder entry that was never registered	.runOrder()
The same name appearing more than once in a run order	.runOrder()
Calling .search() before .defineDataset()	.search()
Calling .search() with nothing registered	.search()

import { NaiveSearch, Search, SearchConfigError } from "@michaelrwalker/search";

try {
	new Search().register(NaiveSearch()).register(NaiveSearch());
} catch (error) {
	if (error instanceof SearchConfigError) {
		console.error(error.message); // 'An algorithm named "naive" is
		                               // already registered...'
	}
}

What is not a SearchConfigError

Bad input to .search(query) itself does not throw: an empty string simply matches nothing (BaseAlgorithm.run returns immediately for a zero-length query), and a query that matches no entry returns a SearchResult with matched: false and best: null rather than throwing. SearchConfigError is reserved for pipeline setup mistakes the compiler cannot catch in plain JavaScript, not for a plain “no results.”