Minor refactor

src/sentry/snuba/metrics/extraction.py

@@ -116,13 +116,6 @@
     # generic percentile is not supported by metrics layer.
 }
 
-# Maps plain Discover functions to derived metric functions which are understood by the metrics layer.
-_SEARCH_TO_DERIVED_METRIC_AGGREGATES: Dict[str, MetricOperationType] = {
-    "failure_count": "on_demand_failure_count",
-    "failure_rate": "on_demand_failure_rate",
-    "apdex": "on_demand_apdex",
-}
-
 # Mapping to infer metric type from Discover function.
 _AGGREGATE_TO_METRIC_TYPE = {
     "count": "c",
@@ -133,10 +126,6 @@
     "p75": "d",
     "p95": "d",
     "p99": "d",
-    # With on demand metrics, evaluated metrics are actually stored, thus we have to choose a concrete metric type.
-    "failure_count": "c",
-    "failure_rate": "c",
-    "apdex": "c",
 }
 
 # Query fields that on their own do not require on-demand metric extraction but if present in an on-demand query
@@ -620,13 +609,37 @@ def apdex_tag_spec(project: Project, argument: Optional[str]) -> List[TagSpec]:
     ]
 
 
-# This is used to map a metric to a function which generates a specification
-_DERIVED_METRICS: Dict[MetricOperationType, TagsSpecsGenerator] = {
-    "on_demand_failure_count": failure_tag_spec,
-    "on_demand_failure_rate": failure_tag_spec,
-    "on_demand_apdex": apdex_tag_spec,
+DERIVED_METRICS = {
+    "apdex": {
+        "tag_spec_func": apdex_tag_spec,
+        "hash_method": "op_plus_arg",
+    },
+    "failure_count": {
+        "tag_spec_func": failure_tag_spec,
+        "hash_method": "op",
+    },
+    "failure_rate": {
+        "tag_spec_func": failure_tag_spec,
+        "hash_method": "op",
+    },
 }
 
+# This is used to map a metric to a function which generates a specification
+_DERIVED_METRICS: Dict[MetricOperationType, TagsSpecsGenerator] = {}
+_HASH_METHOD = {}
+for metric in DERIVED_METRICS.keys():
+    _DERIVED_METRICS[f"on_demand_{metric}"] = DERIVED_METRICS[metric]["tag_spec_func"]
+    _HASH_METHOD[f"on_demand_{metric}"] = DERIVED_METRICS[metric]["hash_method"]
+
+# Maps plain Discover functions to derived metric functions which are understood by the metrics layer.
+_SEARCH_TO_DERIVED_METRIC_AGGREGATES: Dict[str, MetricOperationType] = {}
+for metric in DERIVED_METRICS.keys():
+    _SEARCH_TO_DERIVED_METRIC_AGGREGATES[metric] = f"on_demand_{metric}"
+
+# With on demand metrics, evaluated metrics are actually stored, thus we have to choose a concrete metric type.
+for metric in DERIVED_METRICS:
+    _AGGREGATE_TO_METRIC_TYPE[metric] = "c"
+
 
 @dataclass(frozen=True)
 class FieldParsingResult:
@@ -700,9 +713,9 @@ def _field_for_hash(self) -> Optional[str]:
         # with condition `f` and this will create a problem, since we might already have data for the `count()` and when
         # `apdex()` is created in the UI, we will use that metric but that metric didn't extract in the past the tags
         # that are used for apdex calculation, effectively causing problems with the data.
-        if self.op in ["on_demand_failure_count", "on_demand_failure_rate"]:
+        if _HASH_METHOD[self.op] == "op":
             return self.op
-        elif self.op == "on_demand_apdex":
+        elif _HASH_METHOD[self.op] == "op_plus_arg":
             return f"{self.op}:{self._argument}"
 
         return self._argument