parquetjs
parquetjs copied to clipboard
ironSource
Add the ability to fetch remote files (s3 and http[s])
Adapters included for S3 files and files available over http(s). Only the parts if interest are fetched over the wire, eliminating the need to download the complete files. The performance of the adapters when fetching full rows is drastically improved with https://github.com/ironSource/parquetjs/pull/33
No tests so far, but if you are happy with the approach we could add tests using a simple localhost server and an S3 compatible local service, if that makes sense?
Added the ability to read directly from buffer containing the whole parquet file
@ZJONSSON let's add tests if you have the time, if not let us know and we'll try to do it
I agree. To keep tests as unit tests we would have to add a couple of things to the devDependencies
- server that accepts range requests, such as expressjs
- and an s3 mock adapter such as mock-aws-s3
A test for the buffer reader does not require any additional dependencies
@Kosta-Github it's not abandoned but we're (mainly me) are having problem allocating time for this project, since the initial effort of building it. If you want to contribute, let's discuss :-)