Support for compressed CSV files

[Enchufa2]

Most CSV readers (notably, base R readers) support compressed (gz, bzip2...) CSV files transparently. It would be a nice addition to MonetDBLite::monetdb.read.csv, because big CSVs are commonly gzip'ed.

[hannes]

Yes we had removed this feature because it adds a dependency, which is a pain especially on Windows.

[Enchufa2]

And what about allowing an external command to be specified? May this be possible? Example: data.table::fread. You can read a compressed CSV as follows:

data.table::fread("zcat somefile.csv.gz")

zcat is invoked and its output feeds the reader.

[hannes]

That sounds pretty good. I am unlikely to implement this at the moment. Happy to review a PR though.

[Enchufa2]

And another quick and portable option would be to rely on the R.utils package, which implements gunzip and the like based on base R (efficiently copying from a gzfile, bzfile... connection to a file connection). But this would mean adding R.utils as a dependency (or porting to MonetDBLite just the relevant code). What do you think?