readCSV fails for *.zip

[koperagen]

This file extension is treated in a special way: there's a isCompressed method, and depending on it readCSV wraps InputStream. But it doesn't work for *.zip because InputStream is wrapped in a GZIPInputStream. Apparently it's also not enough to just wrap an InputStream, because ZIP has more complex structure and you need to call methods of ZipInputStream:

val zipInputStream = ZipInputStream(
    File("data.csv.zip").inputStream(),
    Charsets.UTF_8
)
zipInputStream.nextEntry
val df1 = DataFrame.readCSV(zipInputStream)
zipInputStream.closeEntry()

[koperagen]

Another issue is that file ending with *.gz can be *.tar.gz, and we cannot read it properly without some special handling. So, i suggest to either support it or at least provide an exception message that file should be just an archive and not a *.tar

[koperagen]

After the fix it needs to be mentioned in the docs

[Jolanrensen]

There's actually a lot of places where DataFrame assumes a type based on the file extension, but we should avoid that, as file extensions can be changed while the contents of the file are not.

[Jolanrensen]

Will be solved in the new CSV implementation: "dataframe-csv". I will probably also migrate its new Compression class to the :core module in the future to solve reading zips from other read functions too.