fix(translator): % Split over a trait folds to FALSE without an identity context (#18)

The trait branch of PERCENTAGE_SPLIT checked the eval context's
identity.traits and returned FALSE when the property was absent, instead of
hashing the per-row traits subcolumn. In an identity-less context (segment
membership counts/list) that dict is empty, so every trait-keyed split
compiled to FALSE. Hash the per-row value guarded by IS NOT NULL, mirroring
the engine's context_value-is-None -> False, per row.

beep boop
Claude-Session: https://claude.ai/code/session_01EgZ5iHpDASZzCapiHRxHLB

Author: khvn26

src/flagsmith_sql_flag_engine/translator.py

@@ -432,7 +432,6 @@ def translate_condition(cond: SegmentCondition, ctx: TranslateContext) -> str |
             # Engine: float() on the threshold raises → returns False.
             return "FALSE"
         threshold = float(threshold_lit)
-        identity: dict[str, object] = ctx.evaluation_context.get("identity") or {}  # type: ignore[assignment]
         kind = classification.kind
         if not prop:
             # In traditional engine implementations, this branch implies
@@ -457,13 +456,18 @@ def translate_condition(cond: SegmentCondition, ctx: TranslateContext) -> str |
             # future work and let the caller fall back to the engine.
             return None
         else:
-            # Plain trait key, or `$.identity.traits.<X>` rewritten to
-            # the bare key. Hash subject pulls from `i.traits:"<key>"`
-            # per row.
-            traits = identity.get("traits") or {}
-            if not isinstance(traits, dict) or prop not in traits:
-                return "FALSE"
-            value_expr = ctx.dialect.cast_string(ctx.trait_path(prop))
+            # Plain trait key, or `$.identity.traits.<X>` rewritten to the
+            # bare key. Hash the per-row trait value from `i.traits:"<key>"`.
+            # A row missing the trait has no value to bucket, so it cannot
+            # match — mirroring the engine's `context_value is None -> False`.
+            trait_sql = ctx.trait_path(prop)
+            split = _percentage_split_expr(
+                ctx,
+                ctx.segment_key,
+                ctx.dialect.cast_string(trait_sql),
+                threshold,
+            )
+            return f"({trait_sql} IS NOT NULL AND {split})"
         return _percentage_split_expr(ctx, ctx.segment_key, value_expr, threshold)
 
     if not prop:

tests/test_translator_unit.py

@@ -16,13 +16,6 @@
 def _ctx(env_key: str = "test-env-key", env_name: str = "Test") -> TranslateContext:
     eval_ctx: EvaluationContext = {
         "environment": {"key": env_key, "name": env_name},
-        # PERCENTAGE_SPLIT short-circuits to FALSE when the prop isn't in
-        # the eval context's traits, so seed the traits referenced below.
-        "identity": {
-            "identifier": "u",
-            "key": "k",
-            "traits": {"plan": "growth", "country": "GB", "uuid_attr": "abc"},
-        },
     }
     return TranslateContext(evaluation_context=eval_ctx, dialect=ClickHouseDialect())
 
@@ -164,6 +157,31 @@ def test_translate_segment__percentage_split_on_trait__uses_trait_subcolumn() ->
     assert "MD5(" in sql
 
 
+def test_translate_segment__percentage_split_on_trait__guards_missing_trait_per_row() -> None:
+    # Given
+    seg: SegmentContext = {
+        "key": "101",
+        "name": "s",
+        "rules": [
+            {
+                "type": "ALL",
+                "conditions": [{"operator": "PERCENTAGE_SPLIT", "property": "plan", "value": "10"}],
+            }
+        ],
+    }
+
+    # When
+    sql = translate_segment(seg, _ctx())
+
+    # Then
+    assert sql is not None
+    assert "FALSE" not in sql
+    assert "i.traits.`plan`" in sql
+    assert "IS NOT NULL" in sql
+    assert "MD5(" in sql
+    assert "<= 10.0" in sql
+
+
 def test_translate_segment__jsonpath_identity_identifier__uses_column_directly() -> None:
     # Given a condition referencing $.identity.identifier
     seg: SegmentContext = {
@@ -623,25 +641,6 @@ def test_translate_segment__percentage_split_unknown_jsonpath__returns_none() ->
     assert translate_segment(seg, _ctx()) is None
 
 
-def test_translate_segment__percentage_split_trait_not_in_context__compiles_to_false() -> None:
-    # Given a PERCENTAGE_SPLIT on a trait the eval context's identity doesn't carry
-    seg: SegmentContext = {
-        "key": "ps6",
-        "name": "s",
-        "rules": [
-            {
-                "type": "ALL",
-                "conditions": [
-                    {"operator": "PERCENTAGE_SPLIT", "property": "missing_trait", "value": "50"}
-                ],
-            }
-        ],
-    }
-
-    # When / Then the predicate collapses to FALSE
-    assert translate_segment(seg, _ctx()) == "((FALSE))"
-
-
 def test_translate_segment__is_set_on_identity_identifier__emits_true() -> None:
     # Given an IS_SET on $.identity.identifier — every IDENTITIES row IS an
     # identity, so the predicate is unconditionally true