Abbreviate search context (#15028)

src/metabase/search/config.clj

@@ -23,6 +23,10 @@
   "Results in more dashboards than this are all considered to be equally popular."
   50)
 
+(def ^:const surrounding-match-context
+  "Show this many words of context before/after matches in long search results"
+  2)
+
 (def searchable-models
   "Models that can be searched. The order of this list also influences the order of the results: items earlier in the
   list will be ranked higher."

src/metabase/search/scoring.clj

@@ -62,6 +62,18 @@
           (count search-tokens)))
        fs))
 
+(defn- tokens->string
+  [tokens abbreviate?]
+  (let [->string (partial str/join " ")
+        context  search-config/surrounding-match-context]
+    (if (or (not abbreviate?)
+            (<= (count tokens) (* 2 context)))
+      (->string tokens)
+      (str
+       (->string (take context tokens))
+       "…"
+       (->string (take-last context tokens))))))
+
 (defn- match-context
   "Breaks the matched-text into match/no-match chunks and returns a seq of them in order. Each chunk is a map with keys
   `is_match` (true/false) and `text`"
@@ -71,10 +83,11 @@
               {:text match-token
                :is_match (boolean (some #(matches? % match-token) query-tokens))}))
        (partition-by :is_match)
-       (map (fn [matches-or-misses-map]
-              {:is_match (:is_match (first matches-or-misses-map))
-               :text     (str/join " "
-                                   (map :text matches-or-misses-map))}))))
+       (map (fn [matches-or-misses-maps]
+              (let [is-match    (:is_match (first matches-or-misses-maps))
+                    text-tokens (map :text matches-or-misses-maps)]
+                {:is_match is-match
+                 :text     (tokens->string text-tokens (not is-match))})))))
 
 (def ^:const text-score-max
   "The maximum text score that could be achieved without normalization. This value is then used to normalize it down to the interval [0, 1]"

test/metabase/search/scoring_test.clj

@@ -129,6 +129,7 @@
 
 (deftest match-context-test
   (let [context  #'search/match-context
+        tokens   (partial map str)
         match    (fn [text] {:text text :is_match true})
         no-match (fn [text] {:text text :is_match false})]
     (testing "it groups matches together"
@@ -145,7 +146,20 @@
       (is (= [(no-match "aviary stats")]
              (context
                  ["rasta" "toucan"]
-                 ["aviary" "stats"]))))))
+                 ["aviary" "stats"]))))
+    (testing "it abbreviates when necessary"
+      (is (= [(no-match "one two…eleven twelve")
+              (match "rasta toucan")
+              (no-match "alpha beta…the end")]
+             (context
+                 (tokens '(rasta toucan))
+                 (tokens '(one two
+                           this should not be included
+                           eleven twelve
+                           rasta toucan
+                           alpha beta
+                           some other noise
+                           the end))))))))
 
 (deftest test-largest-common-subseq-length
   (let [subseq-length (partial #'search/largest-common-subseq-length =)]