feat(gooddata-sdk): [AUTO] Add refresh-partition endpoint for AI Lake pipe tables

[yenkins-admin]

Summary

Added refresh_partition method to CatalogAILakeService. Since RefreshPartitionRequest is not yet in the generated gooddata-api-client, the method uses a raw requests.post call following the same URL-construction pattern as GoodDataApiClient._do_post_request. Seeds a client-side UUID as the operation-id (matching the analyze_statistics pattern), sends it as an HTTP header, and returns it so callers can poll via get_operation / wait_for_operation.

Impact: new_feature | Services: afm-exec-api

Source commits (gdc-nas):

• ec3f4d3 by Dan Homola — Merge pull request #23181 from gooddata/dho/refresh-partition

Files changed

• packages/gooddata-sdk/src/gooddata_sdk/catalog/ai_lake/service.py
• packages/gooddata-sdk/tests/catalog/unit_tests/test_ai_lake_service.py
• packages/gooddata-sdk/tests/catalog/test_catalog_ai_lake.py

Agent decisions

Decisions (4)

raw HTTP instead of generated client — Use requests.post directly in CatalogAILakeService.refresh_partition

• Alternatives: Wait for api-client regeneration with RefreshPartitionRequest, Hand-roll a generated-client stub (blocked by security hook)
• Why: RefreshPartitionRequest is absent from gooddata-api-client; the task instructions explicitly recommend wrapping raw HTTP for genuinely new endpoints not yet in the generated client. The raw-HTTP layer survives the next api-client regeneration cleanly.

operation-id seeding strategy — Seed a UUID on the client side and send it as the operation-id request header, returning the same UUID to the caller

• Alternatives: Let server generate the ID and read it from the 202 response headers
• Why: Matches the existing analyze_statistics pattern: caller gets a known polling handle without inspecting response headers, enabling deterministic cassette recording.

no new init.py export — refresh_partition is a method on the already-exported CatalogAILakeService; no new class or alias needs re-exporting

• Alternatives: Export a RefreshPartitionRequest wrapper class
• Why: The method signature uses only primitive types (str, dict[str, str]); no new public class is introduced, so no init.py change is required.

integration test scope — Single test (test_refresh_partition) that supplies a fixed operation_id for cassette determinism and asserts the returned ID matches

• Alternatives: Also call get_operation in the same test to verify the operation is trackable, Add a second test for UUID auto-generation
• Why: One cassette per test (caps rule). UUID auto-generation is covered by the unit test. The integration test validates the full HTTP round-trip; the unit tests validate the logic.

Assumptions to verify (3)

• The AI Lake staging instance named 'demo-db' exists and 'fact_orders' is a valid pipe table when the cassette is recorded.
• The pre-existing ty check failures in execution.py (pyarrow not installed in this environment) and filter.py (EmptyValueHandling type mismatch) are baseline issues unrelated to this cluster's changes — they appear in git-unchanged files.
• The refresh-partition endpoint returns HTTP 202 with no body; raise_for_status() is sufficient error handling.

Risks (2)

• Raw requests.post bypasses the api-client's default-headers injection (X-Requested-With, X-GDC-VALIDATE-RELATIONS, Accept-Encoding). If the server requires these headers, the call will fail.
• If the server does not honor the client-supplied operation-id header and generates its own, the returned op_id will not match the server's tracking ID and polling via get_operation(op_id) would fail with 404.

Layers touched (2)

• public_api — New method refresh_partition on CatalogAILakeService; no new public class — method return type is str (operation ID)
  • packages/gooddata-sdk/src/gooddata_sdk/catalog/ai_lake/service.py
• tests — Updated _make_service factory to include _hostname and _token on fake_client; added TestRefreshPartition unit tests and test_refresh_partition integration test
  • packages/gooddata-sdk/tests/catalog/unit_tests/test_ai_lake_service.py
  • packages/gooddata-sdk/tests/catalog/test_catalog_ai_lake.py

OpenAPI diff

--- a/gooddata-afm-client.json
+++ b/gooddata-afm-client.json
@@ -4547,14 +4791,15 @@
           "kind": {
-            "description": "...* `analyze-statistics` — Running ANALYZE TABLE...",
+            "description": "...* `refresh-partition` — Refreshing a specific Hive partition (delete + re-load from S3).",
             "enum": [
               "provision-database",
               "deprovision-database",
               "run-service-command",
               "create-pipe-table",
               "delete-pipe-table",
-              "analyze-statistics"
+              "analyze-statistics",
+              "refresh-partition"
             ]
@@ -5416,6 +5661,23 @@
+      "RefreshPartitionRequest": {
+        "description": "Request to refresh a specific Hive partition in a pipe-backed OLAP table",
+        "properties": {
+          "partitionSpec": {
+            "additionalProperties": { "type": "string" },
+            "description": "Partition column values identifying the partition to refresh.",
+            "type": "object"
+          }
+        },
+        "required": ["partitionSpec"],
+        "type": "object"
+      },
@@ -9874,6 +10215,95 @@
+    "/api/v1/ailake/database/instances/{instanceId}/pipeTables/{tableName}/refresh": {
+      "post": {
+        "description": "(BETA) Deletes all rows for the specified Hive partition and re-loads them from S3.",
+        "operationId": "refreshAiLakePipeTablePartition",
+        "parameters": [
+          { "name": "instanceId", "in": "path", "required": true },
+          { "name": "tableName", "in": "path", "required": true },
+          { "name": "operation-id", "in": "header", "required": false }
+        ],
+        "requestBody": { "$ref": "#/components/schemas/RefreshPartitionRequest" },
+        "responses": {
+          "202": {
+            "headers": {
+              "operation-id": { "description": "Operation ID to use for polling.", "required": true },
+              "operation-location": { "description": "Operation location URL.", "required": true }
+            }
+          }
+        },
+        "summary": "(BETA) Refresh a pipe table partition",
+        "tags": ["AI Lake", "AI Lake - Pipe Tables"]
+      }
+    }

Workflow run

---

Generated by SDK OpenAPI Sync workflow