feat: support excluding anomalous dates from static analytics reports (#4907)

[NoopDog]

Closes #4907

What changed

• analytics/analytics_package/analytics/static_site/fetch.py: fetch_data() gains an optional exclude_dates parameter (list of YYYY-MM-DD strings). When set, the dates are compiled into a GA4 dimension filter via the package's existing filter DSL (date!=YYYYMMDD;... through parse_filter_expressions) and AND-merged into base_dimension_filter, so every query the dashboard makes excludes those days.
• analytics/analytics_package/analytics/static_site/generator.py: passes the new parameter through generate_site().
• analytics/anvil-explorer-sheets/constants.py + generate_static_site.py: new EXCLUDE_BOT_TRAFFIC_DATES = ["2025-02-10", "2025-02-11"] constant, passed as exclude_dates.
• gh-pages/anvil-explorer/: regenerated. Only monthly_traffic.json (Feb 2025: 9,480 → 277 users, 19,951 → 2,239 pageviews) and meta.json (timestamp) changed; all other data files are byte-identical.

Why

GA4 recorded a synthetic traffic burst on 2025-02-10/11: ~9,160 "users" sharing one fingerprint (headless-Chrome 800×600 viewport, single geolocation, every session a first visit) hammering the site — likely a load test. It shows as a bogus spike in the dashboard's monthly traffic chart. GA4 cannot retroactively remove data, so the report queries exclude those dates instead. Excluding at query time (rather than post-processing) keeps GA's user deduplication correct — daily user counts are not additive, so client-side subtraction would corrupt user metrics.

Assumptions I made

• Date exclusion (losing the ~50–70 real users on those two days) was chosen over fingerprint-based filtering, which would silently drop matching traffic forever; the two days are ~0.1% of the chart's range.
• The dates live in per-site constants (property-specific facts), not in the package.
• The exclusion applies to GA4 queries only, not to historic_data_path (pre-GA4 UA history) — documented in both docstrings; no current caller combines the two in an affected range.
• The regenerated data was verified equivalent to a direct GA4 Data API query with the same filter before committing, so no other data files needed to change.
• anvil-portal will consume this via its analytics dependency on this repo's main for the same fix there (chore: exclude 2025-02-10 bot traffic from portal analytics report anvilproject/anvil-portal#4077).

How to verify

Definition of done from #4907: (1) the package supports excluding dates from report queries, (2) the AnVIL Explorer report excludes 2025-02-10/11, (3) the regenerated site no longer shows the spike.

1. Open gh-pages/anvil-explorer/data/monthly_traffic.json and find the 2025-02 entry: it should read 277 users / 2,239 pageviews (previously 9,480 / 19,951). Cross-check in the GA4 UI: AnVIL Explorer property, Feb 2025, exclude Feb 10–11.
2. Serve the site locally (cd gh-pages/anvil-explorer && python -m http.server 8080) and look at the Monthly Traffic Over Time chart: February 2025 should look like an ordinary month, no spike.
3. Confirm every other file under gh-pages/anvil-explorer/data/ is unchanged in this diff (only meta.json's generated_at moved) — the exclusion touched nothing outside those two days.
4. Optionally regenerate end-to-end: cd analytics/anvil-explorer-sheets && python generate_static_site.py (requires GA4 OAuth); the console prints "Excluding dates from all queries: 2025-02-10, 2025-02-11".

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds first-class support in the analytics static-site generator for excluding specific anomalous dates from all GA4 queries, and applies it to the AnVIL Explorer report to remove a known synthetic traffic spike in Feb 2025.

Changes:

• Add an exclude_dates parameter to static-site generation and fetch logic, merged into the existing GA4 dimension filter via the filter DSL.
• Wire exclude_dates through the AnVIL Explorer generation script using a new per-site constant for known bot-traffic dates.
• Regenerate the published AnVIL Explorer static site data to reflect the exclusion (notably Feb 2025 monthly traffic).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
analytics/analytics_package/analytics/static_site/fetch.py	Adds date-exclusion support by AND-merging a date!=YYYYMMDD filter into the base GA4 dimension filter.
analytics/analytics_package/analytics/static_site/generator.py	Plumbs exclude_dates through generate_site() into fetch_data().
analytics/anvil-explorer-sheets/constants.py	Introduces an AnVIL Explorer-specific constant for excluded bot-traffic dates.
analytics/anvil-explorer-sheets/generate_static_site.py	Passes the excluded dates into generate_site() for AnVIL Explorer.
gh-pages/anvil-explorer/data/monthly_traffic.json	Updates Feb 2025 monthly rollup to remove the anomalous spike.
gh-pages/anvil-explorer/data/meta.json	Updates generated timestamp metadata.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

Suppressed comments (1)

analytics/analytics_package/analytics/static_site/fetch.py:368

• Optional robustness: the regex check ensures YYYY-MM-DD formatting, but it still accepts non-existent calendar dates like 2025-02-31 or 2025-13-01. Validating with date.fromisoformat() will fail fast with a clear error rather than silently generating a GA4 filter that matches nothing (or triggers an API error).

    if exclude_dates:
        for d in exclude_dates:
            if not re.fullmatch(r"\d{4}-\d{2}-\d{2}", d):
                raise ValueError(f"exclude_dates entries must be YYYY-MM-DD, got: {d!r}")

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.