module 'pyarrow.fs' has no attribute 'S3FileSystem'

[szdrasiak-apptension]

pyarrow version = 0.17.1

I have latest version installed but I cant import S3FileSystem from pyarrow.fs module.

I'm trying read parquet file from S3 using read_table with argument use_legacy_dataset=False to filter data without partitions, I tried to use file system from s3fs package but I'm getting filesystem` argument must be a FileSystem instance or a valid file system URI. I tried pass full s3 url as source without filesystem parameter but I'm getting Got S3 URI but Arrow compiled without S3 support

Why I'm missing S3FileSystem and is my approach to filter data on reading good?

[wesm]

S3FileSystem is not yet available in the wheel packages from PyPI. Use conda-forge to install instead. If you run into an issue other than this please open a JIRA issue

[kkozmic]

S3FileSystem is not yet available in the wheel packages from PyPI

Is addressing that on the roadmap? Using Conda isn't a viable option for me, so that issue is a bit of a blocker

[wesm]

See d04f54b.