v3.2.0

What's new

• Smaller pages — the frontend ships gzipped and unpacks itself, cutting about 45% off every file. An empty page goes 264KB → 146KB, a 1000-row one 366KB → 248KB.
• --split writes a directory of separately cacheable files instead of one page. Filenames carry a content hash, so an unchanged rerun keeps its URL and the cache, while changed rows get a new URL that can't be served stale. Still opens from disk, no fetch needed.
• --serve is a real HTTP server rather than a file:// URL, and takes an optional [HOST]:PORT:

csvtotable data.csv --serve                  # a random loopback port
csvtotable data.csv --serve :8080            # port 8080 on loopback
csvtotable data.csv --serve 0.0.0.0:8080     # every interface
csvtotable data.csv --serve --split          # each asset served separately

An empty host binds loopback, so exposing data on the network takes writing the host out, and warns when you do.

Flags

New: --split, --no-compress. Changed: --serve takes an optional [HOST]:PORT; a bare --serve behaves as before.

Compatibility

Pages need DecompressionStream (Chrome 103+, Firefox 113+, Safari 16.4+). Older browsers get a message rather than a blank page, and --no-compress inlines readable source for them. --serve now opens http://127.0.0.1:PORT instead of a file path.

Full Changelog: v3.1.0...v3.2.0