---
license: mit
language:
  - zh
pretty_name: Chinese Calendar Data
tags:
  - astronomy
  - calendar
  - chinese-calendar
  - lunar-calendar
  - almanac
  - ephemeris
configs:
  - config_name: days
    data_files: data/v1/days/*.parquet
  - config_name: calendar_months
    data_files: data/v1/calendar_months/*.parquet
  - config_name: calendar_events
    data_files: data/v1/calendar_events/*.parquet
  - config_name: astro_months
    data_files: data/v1/astro_months/*.parquet
  - config_name: astro_events
    data_files: data/v1/astro_events/*.parquet
  - config_name: eclipse_years
    data_files: data/v1/eclipse_years/*.parquet
  - config_name: solar_eclipses
    data_files: data/v1/solar_eclipses/*.parquet
  - config_name: lunar_eclipses
    data_files: data/v1/lunar_eclipses/*.parquet
---

# Chinese Calendar Data

A lossless, versioned Parquet publication of Chinese calendar, Huangli, lunar
phase, solar-term, eclipse, and astronomical-event data generated by
CalcChineseCalendar (CCC).

The dataset title intentionally does not contain a year range.  Release `v1`
covers civil years **1950 through 17000**, inclusive.  Civil dates and times are
stored as numeric columns and never require JavaScript `Date`.

## Release facts

- Civil timezone: UTC+08:00
- Lunar day boundary timezone: UTC+08:00
- Huangli longitude: 120 degrees east
- Ephemeris: JPL DE441, with exact source hashes in `metadata/v1/manifest.json`
- Huangli profiles: `folk`, `ziping`, `purple`, `xieji`
- Far-future civil dates use CCC's documented delta-T extrapolation
- Source web-data validation: 15,051/15,051 years passed, zero failed years

## Tables

| Config | Row meaning | Important contents |
|---|---|---|
| `days` | one civil day | Gregorian/lunar date, festivals, daily Moon state, four Huangli profiles, event references |
| `calendar_months` | one civil month | month metadata and source event/day counts |
| `calendar_events` | one solar term or exact four-phase event | civil second, JD TDB/UTC, original time strings |
| `astro_months` | one civil month | observer metadata and source event count |
| `astro_events` | one astronomy event | typed common event coordinates plus lossless event-specific payload |
| `eclipse_years` | one civil year | date-assignment and Greatest Eclipse metadata |
| `solar_eclipses` | one solar eclipse | assignment event plus lossless full technical data |
| `lunar_eclipses` | one lunar eclipse | assignment event plus lossless full technical data |

Small release metadata remains JSON/Markdown because it is configuration and
documentation, not tabular data.  All high-volume records are Parquet.

Complex objects that do not yet have one stable cross-kind Arrow schema are
stored as canonical UTF-8 JSON in `binary` columns.  This is deliberate: it
preserves missing versus null, list order, double precision, and all
source-defined fields.  `metadata/v1/PARQUET_FORMAT.md` documents the exact
round-trip rules.

## Provenance and validation

The authoritative source manifest, schema, dictionaries, SHA-256 list, and
validation report are published under `metadata/v1/`. Every published Parquet
file was read back and checked for file hash, schema, row count, canonical
payload parsing, and lossless source-object reconstruction.

## License

MIT.  See `LICENSE`.  JPL DE441 is identified as scientific input provenance;
the BSP kernels themselves are not redistributed in this Dataset repository.