Currently, we expect users to write a Conduit to read data from external sources. This is quite easy, however it would be even better to provide some combinators to use common formats and storage systems; eg. JSON, CSV, gzip, parquet and HDFS, S3, HTTP.
Almost all of them already have libraries on Hackage providing Conduit's we can directly use, however it is not desirable increase our dependency footprint a lot. So, maybe we should hide them behind a flag, or create many small libraries (distributed-dataset-json, distributed-dataset-gzip e.g).
Currently, we expect users to write a
Conduit
to read data from external sources. This is quite easy, however it would be even better to provide some combinators to use common formats and storage systems; eg. JSON, CSV, gzip, parquet and HDFS, S3, HTTP.Almost all of them already have libraries on Hackage providing
Conduit
's we can directly use, however it is not desirable increase our dependency footprint a lot. So, maybe we should hide them behind a flag, or create many small libraries (distributed-dataset-json
,distributed-dataset-gzip
e.g).Relevant: #8