Unclear how to set parquet compression

[scottyhq]

The example notebook appears incorrect in that there is a comment that it writes 'compressed stac-geoparquet' by default
https://stac-utils.github.io/stacrs/v0.5.9-beta.0/example/

import stacrs

items = await stacrs.search(
    "https://stac.eoapi.dev",
    collections="openaerialmap",
    bbox=[-125, 25, -67, 49],  # CONUS
    sortby="-properties.datetime",
    max_items=1000,
)
await stacrs.write("items.parquet", items)  # compressed stac-geoparquet

But as far as I can tell no compression is used and I'm unsure how to change it (gpq describe items.parquet --format json)

For example geopandas.to_parquet gives the following options compression = {‘snappy’, ‘gzip’, ‘brotli’, None}, default ‘snappy’

[gadomski]

@scottyhq thanks for the report, really appreciate you creating these — so far this project has been created more-or-less in a vacuum so you're taking the brunt of a "first user" 🙇🏼

I've opened #75, lmk if that helps. You're correct that the example wasn't writing compressed parquet, I've fixed it. Question for ya -- should we be writing compressed by default? Curious what the "best practice" is here...

Also TIL on gpq, many thanks!

[scottyhq]

All good. Happy to kick the tires! Thanks for iterating so quickly on things.

Question for ya -- should we be writing compressed by default

Yes in my opinion, b/c it's a main 'advantage' of parquet over other formats, and other established libraries do it by default (https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html#pyarrow-parquet-write-table).

[gadomski]

Cool ... funny that they default to snappy when the Rust parquet crate recommends zstd default: https://docs.rs/parquet/53.4.0/src/parquet/basic.rs.html#376-380. I think I'll go snappy 🙇🏼