Sort candidate items before getting similarity scores for consistent …

…results You'll notice some of the snapshots here have changed. The fuzzy match scores for these items have not changed. What changed was the order of candidate items passed into the similarity scorer once group by metrics were added to the list of candidate items. The change in order for these snapshots is due to ties in score that are returned in the order they came in. Sort inputs to ensure consistent results in the future.
dbt-labs · May 13, 2024 · 223b0c1 · 223b0c1
1 parent c838681
commit 223b0c1
Show file tree

Hide file tree

Showing 5 changed files with 9 additions and 5 deletions.
diff --git a/metricflow-semantics/metricflow_semantics/query/similarity.py b/metricflow-semantics/metricflow_semantics/query/similarity.py
@@ -23,6 +23,10 @@ def top_fuzzy_matches(
 
     Return scores from -1 -> 0 inclusive.
     """
+    # In the case of a tie in score, items will be returned in the order they were passed in.
+    # Sort candidate item inputs first for consistent results.
+    sorted_candidate_items = sorted(candidate_items)
+
     scored_items = []
 
     # Rank choices by edit distance score.
@@ -31,7 +35,7 @@ def top_fuzzy_matches(
         rapidfuzz.process.extract(
             # This scorer seems to return the best results.
             item,
-            list(candidate_items),
+            sorted_candidate_items,
             limit=max_matches,
             scorer=rapidfuzz.fuzz.token_set_ratio,
         ),

diff --git a/...ggestions.py/str/test_suggestions_for_defined_filters_in_multi_metric_query__result_0.txt b/...ggestions.py/str/test_suggestions_for_defined_filters_in_multi_metric_query__result_0.txt
@@ -16,8 +16,8 @@ Error #1:
       [
         "Dimension('listing__capacity_latest')",
         "TimeDimension('listing__created_at', 'day')",
-        "TimeDimension('listing__ds', 'day')",
         "Dimension('listing__is_lux_latest')",
+        "TimeDimension('listing__ds', 'day')",
         "TimeDimension('user__created_at', 'day')",
         "TimeDimension('user__ds_latest', 'day')",
       ]

diff --git a/...snapshots/test_suggestions.py/str/test_suggestions_for_defined_where_filter__result_0.txt b/...snapshots/test_suggestions.py/str/test_suggestions_for_defined_where_filter__result_0.txt
@@ -16,8 +16,8 @@ Error #1:
       [
         "Dimension('listing__capacity_latest')",
         "TimeDimension('listing__created_at', 'day')",
-        "TimeDimension('listing__ds', 'day')",
         "Dimension('listing__is_lux_latest')",
+        "TimeDimension('listing__ds', 'day')",
         "Dimension('listing__country_latest')",
         "TimeDimension('user__created_at', 'day')",
       ]

diff --git a/...low_semantics/snapshots/test_suggestions.py/str/test_suggestions_for_metric__result_0.txt b/...low_semantics/snapshots/test_suggestions.py/str/test_suggestions_for_metric__result_0.txt
@@ -6,7 +6,7 @@ Error #1:
     The given input does not exactly match any known metrics.
 
     Suggestions:
-      ['bookings', 'booking_fees', 'booking_value', 'instant_bookings', 'booking_payments', 'max_booking_value']
+      ['bookings', 'booking_fees', 'booking_value', 'booking_payments', 'instant_bookings', 'booking_value_p99']
 
   Query Input:
 

diff --git a/...ics/snapshots/test_suggestions.py/str/test_suggestions_for_multiple_metrics__result_0.txt b/...ics/snapshots/test_suggestions.py/str/test_suggestions_for_multiple_metrics__result_0.txt
@@ -19,7 +19,7 @@ Error #1:
         'listing__lux_listing',
         'listing__is_lux_latest',
         'listing__country_latest',
-        'listing__created_at__day',
+        'listing__capacity_latest',
       ]
 
   Query Input: