# Normalize line endings; keep data files byte-for-byte.
* text=auto eol=lf
*.csv -text
*.tsv -text
*.gz binary
# Scripts must keep LF endings to run.
*.sh text eol=lf
# Gitea (and GitHub): collapse generated files in diffs, keep pixi.lock out of language stats.
pixi.lock merge=binary linguist-language=YAML linguist-generated=true -diff
# paperspec-specific: generated data whose exact bytes matter (the broken packages include CRLF and
# empty files on purpose), diffed by line.
vectors/** -text diff
fixtures/** -text diff
fixtures/**/*.png binary
