PY-180 Change type inference in neptune-query for better UX (#406)

* PY-180 Change type inference in neptune-query for better UX

1. If the type cannot be inferred because the attribute doesn't exist in the project -> show a warning and infer as string
2. If the type cannot be inferred because the attribute has multiple types across the runs and experiments of the project -> raise an exception prompting the user to specify the type explicitly
  * don't try to be smart and limit the attribute search to only runs/experiments matching some filters, etc

* Emit inference warnings for list_containers and list_attributes.

* Test warning is emmitted for missing attributes

Author: gabrys

src/neptune_query/internal/composition/fetch_metrics.py

@@ -84,20 +84,10 @@ def fetch_metrics(
             client=client,
             project_identifier=project_identifier,
             filter_=filter_,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
-            container_type=container_type,
         )
-        if inference_result.is_run_domain_empty():
-            return create_metrics_dataframe(
-                metrics_data={},
-                sys_id_label_mapping={},
-                index_column_name="experiment" if container_type == ContainerType.EXPERIMENT else "run",
-                timestamp_column_name="absolute_time" if include_time == "absolute" else None,
-                include_point_previews=include_point_previews,
-                type_suffix_in_column_names=type_suffix_in_column_names,
-            )
         inferred_filter = inference_result.get_result_or_raise()
+        inference_result.emit_warnings()
 
         metrics_data, sys_id_to_label_mapping = _fetch_metrics(
             filter_=inferred_filter,

src/neptune_query/internal/composition/fetch_series.py

@@ -81,19 +81,10 @@ def fetch_series(
             client=client,
             project_identifier=project_identifier,
             filter_=filter_,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
-            container_type=container_type,
         )
-        if inference_result.is_run_domain_empty():
-            return create_series_dataframe(
-                series_data={},
-                project_identifier=project_identifier,
-                sys_id_label_mapping={},
-                index_column_name="experiment" if container_type == ContainerType.EXPERIMENT else "run",
-                timestamp_column_name="absolute_time" if include_time == "absolute" else None,
-            )
         inferred_filter = inference_result.get_result_or_raise()
+        inference_result.emit_warnings()
 
         sys_id_label_mapping: dict[identifiers.SysId, str] = {}
 

src/neptune_query/internal/composition/fetch_table.py

@@ -79,38 +79,19 @@ def fetch_table(
             client=client,
             project_identifier=project_identifier,
             filter_=filter_,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
-            container_type=container_type,
         )
-        if inference_result.is_run_domain_empty():
-            return output_format.convert_table_to_dataframe(
-                table_data={},
-                project_identifier=project_identifier,
-                selected_aggregations={},
-                type_suffix_in_column_names=type_suffix_in_column_names,
-                index_column_name="experiment" if container_type == search.ContainerType.EXPERIMENT else "run",
-            )
         filter_ = inference_result.get_result_or_raise()
+        inference_result.emit_warnings()
 
         sort_by_inference_result = type_inference.infer_attribute_types_in_sort_by(
             client=client,
             project_identifier=project_identifier,
-            filter_=filter_,
             sort_by=sort_by,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
-            container_type=container_type,
         )
-        if sort_by_inference_result.is_run_domain_empty():
-            return output_format.convert_table_to_dataframe(
-                table_data={},
-                project_identifier=project_identifier,
-                selected_aggregations={},
-                type_suffix_in_column_names=type_suffix_in_column_names,
-                index_column_name="experiment" if container_type == search.ContainerType.EXPERIMENT else "run",
-            )
-        sort_by = sort_by_inference_result.get_result_or_raise()
+        sort_by = sort_by_inference_result.result
+        sort_by_inference_result.emit_warnings()
 
         sys_id_label_mapping: dict[identifiers.SysId, str] = {}
         result_by_id: dict[identifiers.SysId, list[att_vals.AttributeValue]] = {}
@@ -184,6 +165,7 @@ def go_fetch_sys_attrs() -> Generator[list[identifiers.SysId], None, None]:
         index_column_name="experiment" if container_type == search.ContainerType.EXPERIMENT else "run",
         flatten_aggregations=flatten_aggregations,
     )
+
     return dataframe
 
 

src/neptune_query/internal/composition/list_attributes.py

@@ -62,13 +62,10 @@ def list_attributes(
             client,
             project_identifier,
             filter_,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
-            container_type=container_type,
         )
-        if inference_result.is_run_domain_empty():
-            return []
         filter_ = inference_result.get_result_or_raise()
+        inference_result.emit_warnings()
 
         output = _components.fetch_attribute_definitions_complete(
             client=client,

src/neptune_query/internal/composition/list_containers.py

@@ -37,20 +37,15 @@ def list_containers(
     validated_context = _context.validate_context(context or _context.get_context())
     client = _client.get_client(context=validated_context)
 
-    with (
-        concurrency.create_thread_pool_executor() as executor,
-        concurrency.create_thread_pool_executor() as fetch_attribute_definitions_executor,
-    ):
+    with concurrency.create_thread_pool_executor() as fetch_attribute_definitions_executor:
         inference_result = type_inference.infer_attribute_types_in_filter(
             client=client,
             project_identifier=project_identifier,
             filter_=filter_,
-            executor=executor,
             fetch_attribute_definitions_executor=fetch_attribute_definitions_executor,
         )
-        if inference_result.is_run_domain_empty():
-            return []
         filter_ = inference_result.get_result_or_raise()
+        inference_result.emit_warnings()
 
         sys_attr_pages = search.fetch_sys_id_labels(container_type)(client, project_identifier, filter_)
         return list(sorted(attrs.label for page in sys_attr_pages for attrs in page.items))