feat: use well-defined dimensions instead of strings (#284)

Using dimensions as a list of strings causes ambiguity in other druid clients. Well-defined dimensions makes use of the utility in pydruid to create dimensions as DimensionSpec. This helps with deterministic serialization and de-serialization of the query. --------- Signed-off-by: shrivardhan <[email protected]> Co-authored-by: Avik Basu <[email protected]>
numaproj · Sep 13, 2023 · 0cdc257 · 0cdc257
1 parent de8930a
commit 0cdc257
Showing 1 changed file with 4 additions and 1 deletion.
diff --git a/numalogic/connectors/druid.py b/numalogic/connectors/druid.py
@@ -5,6 +5,7 @@
 import pandas as pd
 import pytz
 from pydruid.client import PyDruid
+from pydruid.utils.dimensions import DimensionSpec
 from pydruid.utils.filters import Filter
 from numalogic.connectors._config import Pivot
 from typing import Optional
@@ -50,13 +51,15 @@ def fetch_data(
         start_dt = end_dt - timedelta(hours=hours)
         intervals = f"{start_dt.isoformat()}/{end_dt.isoformat()}"
 
+        dimension_specs = map(lambda d: DimensionSpec(dimension=d, output_name=d), dimensions)
+
         params = {
             "datasource": datasource,
             "granularity": granularity,
             "intervals": intervals,
             "aggregations": aggregations,
             "filter": _filter,
-            "dimensions": dimensions,
+            "dimensions": dimension_specs,
         }
 
         _LOGGER.debug(