Catalog¶

The thatch.catalog module merges per-object photometry into a unified catalog in Parquet format.

Building the Catalog¶

from thatch.catalog import build_catalog, save_catalog, catalog_summary

catalog = build_catalog("data/")
catalog_summary(catalog)
save_catalog(catalog, "thatch_catalog.parquet")

Output:

THATCH Catalog Summary
  Total measurements: 584
  Objects: 12
  Valid detections: 449

  By object:
    AT2017gfo           :   18 total,    6 detections, 3 filters
    SN1987A             :  160 total,  143 detections, 10 filters
    SN2011fe            :   37 total,   11 detections, 6 filters
    SN2014J             :   25 total,   20 detections, 11 filters
    ...

Schema¶

The catalog uses a standardized column set:

Column	Type	Description
`object`	str	Transient name
`mjd`	float	Modified Julian Date
`delta_t_days`	float	Days from reference epoch
`filter`	str	HST filter name (e.g., F110W)
`instrument`	str	Instrument/detector (e.g., WFC3/IR)
`ab_mag`	float	Calibrated AB magnitude
`ab_mag_err`	float	Magnitude uncertainty
`count_rate`	float	Aperture-corrected count rate (e-/s)
`photflam`	float	Inverse sensitivity (erg/cm²/Å/e-)

Data Formats¶

Primary: Parquet (fast columnar reads, HuggingFace compatible)
Secondary: HDF5 for cutouts and spectra (array data)

Loading from HuggingFace¶

from thatch.data import load_catalog

catalog = load_catalog()  # Downloads from applecider-ml/thatch