yuzu-data

Native I/O layer for citrusquant: reads price and fundamental files into yuzu-core panels.

Formats

The file format is detected from content, not the object key, so existing data keeps working and mirrors can migrate incrementally:

gzip CSV (.csv.gz) — the default write format.
plain CSV (.csv).
Apache Parquet (.parquet) — columns matched by name (day, adj_close, pe, …; one column per symbol for combined panels); the day column may be an integer YYYYMMDD, a YYYY-MM-DD string, or a logical date. Reading only — writing/rebuild output stays gzip CSV.

All three work out of the box; the loaders probe .csv.gz, then .parquet, then .csv for each symbol/panel.

Parquet pulls the arrow stack. Consumers that only need CSV/gzip and want to avoid that dependency can opt out:

yuzu-data = { version = "0.1", default-features = false }

On-disk / object-key contract (directory tree, CSV columns, series-name map): Input data layout.

Engine semantics (panels, NAV, Report): backtest-engine.md.

MIT