feat: search for similar packages

README.md

@@ -73,6 +73,31 @@ Available options:
 
 - `size`: The total number of results to return, defaults to `25`
 
+#### .search.similar(q, esClient, [options]) -> Promise
+
+Perform a fuzzy search for similarly named packages.
+
+Results are ranked based on a combination of analyzer weightings (`quality`, `popularity`, `maintenance`) and the `_score` returned by the [fuzzy match](https://www.elastic.co/guide/en/elasticsearch/guide/current/fuzzy-match-query.html).
+
+```js
+const queries = require('@npms/queries');
+
+// ...
+queries.search.similar('chaik', esClient)
+.then(results => {
+  // perhaps we were instead looking for chalk?
+});
+```
+
+Available options:
+
+- `size`: The total number of results to return, defaults to `10`.
+- `analyzerWeight`: How much should we weight the analyzer values by? defaults to `2.2`.
+- `scoreWeight`: How much should we weight the fuzzy score by? defaults to `1.5`.
+- `minScore`: defaults to `4.5`.
+
+_the above default values were based on trial and error examining the
+ top npm modules, they will likely change over time._
 
 ## Tests
 

index.js

@@ -2,3 +2,4 @@
 
 module.exports.search = require('./lib/search');
 module.exports.search.suggestions = require('./lib/searchSuggestions');
+module.exports.search.similar = require('./lib/searchSimilar');

lib/searchSimilar.js

@@ -0,0 +1,62 @@
+'use strict';
+
+const pick = require('lodash/pick');
+const parseQuery = require('./util/parseSearchQuery');
+const toEsClient = require('./util/toEsClient');
+
+function searchSimilar(q, esClient, options) {
+    esClient = toEsClient(esClient);
+    options = Object.assign({
+      size: 10,
+      minScore: 4.5,
+      analyzerWeight: 2.2,
+      scoreWeight: 1.5
+    }, options);
+
+    const text = parseQuery.discardQualifiers(q);
+    const script = `(doc["score.final"].value * ${options.analyzerWeight}) * (_score * ${options.scoreWeight})`;
+
+    if (!text) {
+        return Promise.resolve([]);
+    }
+
+    return Promise.resolve(esClient.search({
+        /* eslint camelcase: 0 */
+        index: 'npms-current',
+        type: 'score',
+        body: {
+            size: options.size,
+            query: {
+                function_score: {
+                    min_score: options.minScore,
+                    boost_mode: 'replace',
+                    query: {
+                        fuzzy: {
+                          'package.name.raw': {
+                              value: text
+                          }
+                        }
+                    },
+                    script_score: {
+                        lang: 'groovy',
+                        script: script,
+                        params: {},
+                    },
+                },
+            },
+        },
+    }))
+    .then((res) => res.hits.hits.map((hit) => {
+        // We can't use _fields in the query because the JSON properties order get messed up,
+        // see https://github.com/elastic/elasticsearch/issues/17639
+        // So we filter the source fields manually with pick().. this is not ideal since there's payload
+        // navigating through the network that we do not use, but it's definitively better than having order messed up
+        const result = pick(hit._source, ['package', 'flags', 'score']);
+
+        result.searchScore = hit._score;
+
+        return result;
+    }));
+}
+
+module.exports = searchSimilar;