[dasgoutam]

Short description

When page/s are saved as draft (bulk action or in the editor), some of them still show up in the Page Based Statistics until the Redis cache is flushed.

Proposed changes

• When Redis is enabled, use cacheops invalidated_update() instead of update(), which clears the cached queries. Without Redis the update() stays. This follows the existing pattern in other parts of the code.
• Applied in change_publication_status (bulk action "Save as draft") and in PageFormView

Side effects

• Hopefully None

Faithfulness to issue description and design

There are no intended deviations from the issue and design.

How to test

Follow steps in Issue Description

• Make sure that there are 1+ published pages
• Go to 'Statistics' verify the pages are visible in Page Based Statistics. Go to 'Pages' and select all pages
• Select bulk action 'Save as draft' and execute bulk action
• Go to 'Statistics'. The Pages should no longer be visible after choosing a language
• Check other conditions by adding and removing Draft status from a single page, while keeping the remaining pages published

Resolved issues

Fixes: #4139

---

Pull Request Review Guidelines

[MizukiTemma]

Looks good 😸

💡 We can add a release note 📜

[deliverino]

LLM PR Review (verdigado-code)

Analyzed diff, but not the whole repository. Important context might be missing.

• The PR body says the change is used “when Redis is enabled”, but the diff gates on hasattr(..., "invalidated_update") in integreat_cms/cms/views/utils/publication_status.py:60 and integreat_cms/cms/views/pages/page_form_view.py:547. That checks whether cacheops has monkey-patched QuerySet, not whether Redis is configured or active. Either the description or the condition should be aligned.

• The PageFormView change only modifies set_dependent_translations_to_draft(). If the editor’s primary page/translation draft-save path performs another queryset update() outside this diff, that path is not covered by the change.

• The fix only invalidates cacheops-managed queryset caches. If the page-based statistics endpoint uses a separately cached aggregate or cache key outside cacheops, this may not clear it; that cannot be verified from the diff alone.

• invalidated_update is not a standard Django QuerySet method. The diff shows no type stub, protocol, or # type: ignore[...]; confirm that strict mypy accepts these calls.

• Using hasattr to detect cacheops is a runtime implementation detail. If cacheops is enabled with a non-Redis backend, the code will still call invalidated_update, contrary to the statement that “Without Redis the update() stays”.

• No tests are included. The stale-statistics behavior should be covered by a test that saves pages as draft and asserts they no longer appear in page-based statistics, or at least by a test for the cacheops-aware fallback behavior.

• No new views, permission checks, API response fields, model/migration changes, or new gettext strings are visible in the diff.

• Release note integreat_cms/release_notes/current/unreleased/4139.yml is present with en and de. Nitpick: “page based” could be “page-based” if that is the term used in the UI.

• Commit messages add release note and change release note message are vague. The first commit says “when redis is available”, but the current code checks for cacheops instead.