Support Python 3.15, drop Python 3.10, require pandas 3

[hkad98]

Adds Python 3.15 support, drops Python 3.10 (EOL October 2026), requires pandas 3 and migrates (str, Enum) classes to StrEnum. These are breaking changes; the next release should be 2.0.0.

Draft: the py315 CI job is expected to fail until pyarrow 26.0.0 (first release with cp315 wheels, currently in its release vote as 26.0.0 RC0) is on PyPI. Once it is, a uv lock --upgrade-package pyarrow commit finishes this PR.

Breaking changes

• Python 3.11+ required. Python 3.10 users stay on 1.x.
• gooddata-pandas requires pandas >=3.0.0,<4.0.0. The DataFrames and Series it returns follow pandas 3 semantics: string columns and index levels use the default str dtype instead of object, missing values in string columns stay missing, and copy-on-write is always on. Date/datetime attribute columns keep datetime64[ns] as before.
• StrEnum migration. str() and format() of these enums now return the member value (e.g. "asc") instead of "SortDirection.ASC". Equality with plain strings and serialization are unchanged.
  • gooddata-sdk: SortDirection, LocatorItemType (UpsertOutcome already behaved this way)
  • gooddata-pipelines: CustomFieldType, ColumnDataType, EntityType, LoadType, WorkflowType
  • gooddata-eval: AnomalyDetectionGranularity, LoopExit

Commits

Override cachebox to >=6

cachebox (sdk test group, via deepdiff) publishes cp315 wheels only from 6.x, but deepdiff 9.1.0 still caps it <6. A uv override-dependencies entry forces 6.2.7; drop it once deepdiff allows cachebox 6.

Support Python 3.15 and drop Python 3.10

• 3.15 added to classifiers, tox envlists, the CI matrix and ci_tests.mk (new test-ci-py315-build target on python:3.15.0rc3-slim-bookworm; bump to the final tag once released).
• requires-python >=3.11 everywhere; the 3.10 classifier, tox env, CI entry and Docker target removed; READMEs, AGENTS.md, CONTRIBUTING.md and Sonar settings updated.
• ruff target-version = "py311": applied the safe UP017 (datetime.UTC) and import-order fixes.

Require pandas 3 in gooddata-pandas

pandas 2 has no cp315 wheels (2.3.3 built from source segfaults inside pd.to_datetime on 3.15), and pandas 3's behavior changes reach users through gooddata-pandas' return values, so the pin moves to >=3.0.0,<4.0.0 as a breaking change rather than silently allowing both majors.

• pd.to_datetime infers resolution in pandas 3 (datetime64[s] for dates). Date/datetime conversion in both the JSON and Arrow paths now goes through to_datetime_ns(), keeping the datetime64[ns] dtype gooddata-pandas returned before.
• Index levels built from Arrow field metadata (transposed results) were never Arrow-mapped and now get the default str dtype; the ARROW_STRINGS test skips them as it skips object levels. Converter output is unchanged.

Migrate (str, Enum) classes to StrEnum

With 3.11 as the floor, StrEnum is available and several enums already carried notes to switch once 3.10 was dropped. ruff UP042 is enforced.

Points for review

• Keeping date columns at datetime64[ns] vs. adopting pandas 3's inferred resolution.
• Release as 2.0.0, with release notes covering the breaking changes above.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

_{Comment @coderabbitai help to get the list of available commands.}