var util = require("../util/util");
var STOPWORDS = require("./stopwords");
var parse_term = require("./parse_term");

function snippet(s, terms) {
	s = util.sanitize(s);
	var term_set = new Set(terms);

	var words = s.split(/[^a-zA-Z0-9]+/);
	var snippet = "";

	// deduper for matching @words indexes, so we don't add a word twice
	var index_matches = {};

	// words in the eventual snippet
	var words_matched = [];

	// counter for aggregating context after a match
	var aggr = 0;

	// amount of context to show, in number of words surrounding a match
	var pad = 10;

	// loop over each of the words in the string
	for (var i = 0, len = words.length; i < len; i++) {
		var word = words[i];
		var term = parse_term(word);

		// if the word matches...
		if (term && term_set.has(term) && !STOPWORDS.has(term.toLowerCase())) {
			// if we aren't already aggregating, add an ellipsis
			if (!aggr) {
				words_matched.push("...");
			}

			// look backward $pad words
			var idx;
			INNER: for (var j = -pad; j < 1; j++) {
				// create a new index from the offset
				idx = i + j;

				// is this a valid index? has it already been encountered?
				if (idx < 0) continue INNER;
				if (idx > words.length) continue INNER;
				if (index_matches[idx]) continue INNER;

				// checks out, save this word
				words_matched.push(words[idx]);

				// note the matching index in our deduper
				index_matches[idx] = 1;
			}
			// enter aggregate mode -- add the next (pad) words
			aggr = pad;
		}

		// have we been told to aggregate?
		else if (aggr) {
			// save this word
			words_matched.push(word);

			// add index to the deduper
			index_matches[i] = 1;

			// one less word to aggregate
			aggr--;
		}

		// keep snippets to a modest length
		if (words_matched.length > 30) break;
	}

	// add a trailing ellipsis
	words_matched.push("...");

	// create the snippet from the saved context words
	snippet = words_matched.join(" ");

	return snippet;
}

module.exports = snippet;