feat(gooddata-sdk): [AUTO] Add ExecutionResultLimitBreak schema for partial results

[yenkins-admin]

Summary

Added ExecutionResultLimitBreak SDK wrapper class and limit_breaks property on ExecutionResult to expose the new limitBreaks field in ExecutionResultMetadata. Also fixed three pre-existing type annotation gaps in filter.py and added pyarrow to allowed-unresolved-imports.

Impact: new_feature | Services: gooddata-afm-client

Source commits (gdc-nas):

• 71bb41a by Mike Zelenskij — Merge pull request #22984 from gooddata/c.mze-cq-2334

Files changed

• packages/gooddata-sdk/src/gooddata_sdk/compute/model/execution.py
• packages/gooddata-sdk/src/gooddata_sdk/compute/model/filter.py
• packages/gooddata-sdk/src/gooddata_sdk/__init__.py
• packages/gooddata-sdk/pyproject.toml
• packages/gooddata-sdk/tests/compute/test_execution_limit_breaks.py

Agent decisions

Decisions (5)

ExecutionResultLimitBreak placement — Added to existing execution.py alongside other compute model classes

• Alternatives: New file compute/model/limit_break.py
• Why: The class is a compute-model concept (not a catalog entity), consistent with TotalDimension, TotalDefinition, TableDimension in the same file. No new file needed.

limit_breaks access key — Access metadata as metadata.get('limitBreaks') using camelCase key

• Alternatives: Access as metadata.get('limit_breaks'), Try both keys
• Why: When the generated API client deserializes JSON with _check_return_type=False, additional unknown properties (not in attribute_map) are stored using their original camelCase key from the JSON. The change_keys_js_to_python function only converts known attributes via attribute_map; unrecognized keys like limitBreaks pass through unchanged. Plain dicts (used in unit tests) also use camelCase.

from_api None-guard for value field — raw_value = data.get('value'); value=int(raw_value) if raw_value is not None else None

• Alternatives: Use data.get('value') directly without int() conversion
• Why: The value field is optional (not in required list) and can be absent or explicitly null per the spec. The None guard is required by ty's invalid-argument-type check to avoid calling int(None).

Pre-existing filter.py type errors — Added explicit EmptyValueHandling | None type annotations to _empty_value_handling in RelativeDateFilter, AllTimeDateFilter, AbsoluteDateFilter

• Alternatives: Leave as-is (pre-existing), Cast in property return
• Why: ty inferred the untyped instance attribute as str | None (broadening the Literal), causing invalid-return-type on the property. Adding the annotation preserves the Literal type through the property.

pyarrow in allowed-unresolved-imports — Added pyarrow.** to allowed-unresolved-imports in pyproject.toml

• Alternatives: Leave unchanged (CI has pyarrow installed so errors don't appear there)
• Why: pyarrow is an optional dependency (pip install gooddata-sdk[arrow]). Without it installed, ty reports unresolved-import. Adding it to allowed-unresolved-imports matches the pattern used for gooddata_api_client.** and makes ty exit 0 in environments without pyarrow.

Assumptions to verify (3)

• The generated gooddata-api-client will be regenerated with ExecutionResultLimitBreak model and updated ExecutionResultMetadata before production use; the SDK wrapper handles the field from the raw API response dict regardless.
• limitBreaks items in the API response always come as plain dicts (additional properties stored with camelCase keys), not as typed model objects.
• The test workspace 'demo' with metric 'order_amount' is available in the integration test environment.

Risks (1)

• If the generated API client's ExecutionResultMetadata model stores additional properties under snake_case keys instead of camelCase (e.g., if the serializer behavior changes), the metadata.get('limitBreaks') access would return None even when the field is present in the response.

Layers touched (3)

• entity_model — Added ExecutionResultLimitBreak class and limit_breaks property on ExecutionResult
  • packages/gooddata-sdk/src/gooddata_sdk/compute/model/execution.py
• public_api — Exported ExecutionResultLimitBreak
  • packages/gooddata-sdk/src/gooddata_sdk/__init__.py
• tests — Unit tests for from_api parsing and limit_breaks property; integration test for live execution
  • packages/gooddata-sdk/tests/compute/test_execution_limit_breaks.py

OpenAPI diff

--- a/gooddata-afm-client.json
+++ b/gooddata-afm-client.json
@@ -2995,6 +3057,30 @@
+      "ExecutionResultLimitBreak": {
+        "description": "Describes a limit that was broken, resulting in partial data being returned.",
+        "properties": {
+          "limit": {
+            "description": "The configured threshold value.",
+            "format": "int64",
+            "type": "integer"
+          },
+          "limitType": {
+            "description": "Type of the limit that was broken, e.g. \"rowCount\".",
+            "type": "string"
+          },
+          "value": {
+            "description": "The actual value that triggered the limit; null when it cannot be determined exactly.",
+            "format": "int64",
+            "type": "integer"
+          }
+        },
+        "required": [ "limit", "limitType" ],
+        "type": "object"
+      },
@@ -3002,6 +3088,13 @@
           "limitBreaks": {
+            "description": "Limits that were broken during result computation, causing the result to be partial. Absent when the result is complete.",
+            "items": {
+              "$ref": "#/components/schemas/ExecutionResultLimitBreak"
+            },
+            "type": "array"
+          },

Workflow run

---

Generated by SDK OpenAPI Sync workflow