Why isn't zip compression included for read_csv?

[stoffprof]

Is there a reason why the compression options for read_csv and other IO functions don't include zip?

[jreback]

well you can simply use compression='gzip' to read a zip file, or see here.

I suppose in theory this could be implemented, but easy enough to wrap it.

I'll mark this an an enhancement issue if you'd like to work on it.

[stoffprof]

I thought compression='gzip' reads .gz files, but not .zip. If I understand your suggestion, you're saying just to incorporate the z.open('some_file_name.csv') into pd.read_csv to add a compression='zip' option. I'm fairly new to GitHub (and python), but I could give it a shot.

I suppose this approach would require there to be only one file in the .zip archive, unless a new parameter were to be added to pd.read_csv that allowed the user to specify which file to read from the archive.

[jreback]

IIRC compresssion='gzip' should read a zip file

you can always do

with open(....) as fh:
    pd.read_csv(fh...)

to do it anyhow.

http://pandas.pydata.org/pandas-docs/stable/contributing.html are the contributing docs

[bashtage]

A zip file can contains many csv files, folders, and other files- and is closer to a file system with compression than a compressed file. IMO it isn't practical to add support for zip files in any generic way. The simplest way is for the user to pass a buffer-like object from the zip to read csv.

One could add the special case where a zip contains a single csv file in the root of the zip, but this is far from zip "support".

[jzf2101]

Right now we can read .zip but not write .zip files, any plans to include write to .zip?