A court holiday calendar decides whether a filing deadline moves. Our database holds holiday dates for 813 covered court listings. Only 360 of those listings have any date for 2027 so far. We counted what the stored calendars show. Here is the method.

The unit and the population

The unit is a court holiday calendar, keyed by a canonical district_id and evaluated for calendar year 2027. One row in the data file stands for one court listing with at least one stored holiday date, including listings with no target-year date. A statewide source copied to county listings contributes a calendar to each listing. That copy does not establish separate decisions by each court. Courts at different levels can serve the same place.

We read the production database on October 22, 2026, Pacific time. Every request was an HTTP GET that translates to a SELECT. No write, RPC or DDL ran. We excluded the redirect aliases in the site's duplicate-court-stubs.json and the California statewide rules-and-forms record, which is not a court listing. Those scope decisions stay in reviews.json.

The live canonical registry includes 979 court listings. Of those, 813 have holiday dates from any stored year. The per-court file includes every canonical listing, including the ones with no stored calendar. That keeps courts without holiday data from dropping out of the census.

The fields we read

The holiday contract stores direct fields rather than a rule-type flag, such as district_id, calendar_year, holiday_date, holiday_name, source_url, scraped_at, created_at and the holiday row ID. The table is unique on (district_id, holiday_date). There is no document date, publication date, annual-calendar completeness flag or part-day flag. scraped_at is a database retrieval timestamp. It is not a source publication date.

The builder pages five GET/SELECT endpoints in a stable order with an offset. The endpoints are v2_courts, v2_court_holidays, v2_judges, v_rule_sources_current and v2_rule_counts_current. It fetches all stored holiday years for the denominator, then filters 2027 locally. It joins holiday rows to courts on district_id and asserts that each calendar year agrees with its date and that dates do not repeat within a court. The exact SELECTs are in build.py and internal/search-audit.json.

The comparison baseline

We used the OPM observed-holiday lists as a comparison baseline, read October 22. A weekday counts as extra when it falls outside the federal observed-date list for the target year. Christmas Eve and New Year's Eve are already federal observed dates in 2027. We do not count either as an extra weekday just because its label says Eve. The baseline is a convention. It is not a statement that every federal court follows the OPM employee calendar. Weekends do not create extra-weekday matches. We projected no dates from recurring rules and imputed no missing holiday.

The review step

The headline figure counts stored evidence. Among the 360 calendars with some 2027 dates, 314 list a weekday outside the federal baseline. The Friday after Thanksgiving appears in 234 of those 360. In 24 of the 360, the only stored date is New Year's Day. That makes the first figure narrow. A stored date does not establish that a court has issued a complete annual calendar.

We compared ten randomly drawn calendars against their sources. The ten completed comparisons found no date mismatch. Three drawn sources were unavailable and remain unverified in the CSV. The completed sample includes repeated statewide sources and county-office calendars. It shows agreement for those sampled dates, not for every local adoption or every holiday name. If more than one comparison had failed, we would repair the method and draw ten new rows. The completed checks did not cross that threshold. No database record was corrected.

What the numbers mean

A missing date means we retrieved no matching row. It does not prove that the court has no holiday, has not published dates or will be open. The database reflects extracted source material and can be incomplete. The 360 calendars refer to 116 distinct source URLs and 117 source-and-date-set groups. Those groups are not independent court decisions or verified unique publications. A single source URL can carry different stored subsets.

The public CSV keeps the flags, source URLs, dates, source grouping and a quotation capped at 300 characters. The per-court counts are on the data page at https://www.courtrules.app/data/holidays-2027?utm_source=devto&utm_medium=post&utm_campaign=brand