.gitattributes

*.zip filter=lfs diff=lfs merge=lfs -text

# LFS is for files too big to live in git, NOT for "it is a CSV".
#
# The blanket `*.csv` rule this replaces was set in the repo's first commit and swept up
# 24 files, of which 22 are small and 9 are CURATED dbt seeds — occurrence_synonyms.csv,
# curated_taxon_ids.csv, inat_query_taxa.csv and friends. Those are hand-edited files
# whose entire value is the reviewable diff and the provenance in their `source`/`note`
# column, and LFS renders every change to them as an opaque oid+size swap. Adding the
# Bombus lapponicus synonymy, a real curatorial decision with a citation, showed up in
# git as two changed hashes and nothing else.
#
# The 22 small files total ~559 KB against a 67 MB .git, so this is not a size decision —
# storing them as ordinary blobs is free and they delta-compress well. It also removes a
# trap: pushing a commit that adds a new .csv needs git-lfs-authenticate on the remote,
# which is not on maderas's non-login PATH, so `git push maderas main` failed the first
# time anyone added a CSV in a while.
#
# Only genuinely large data files are listed, by path. Add a line here when a new one
# appears; do not restore a glob.
data/raw/inat_expert_obs.csv filter=lfs diff=lfs merge=lfs -text
data/checklists/checklist_records_full.csv filter=lfs diff=lfs merge=lfs -text