yes! i will look into that. i just started using something different than emacs and the command line. so, yes, it wasn't very thoughtful to put such settings.

you can’t think of everything! thanks for your work

I removed all this automatic setting of backends etc.

Currently "is_interactive" is only used to choose different progress bars (tqdm behaves very differently on the command line, in jupyter and then, unfortunately again differently in Rodeo) and to decide on whether a total wall time should be output when leaving the session.

It's now left to the user to choose the matplotlib backend. If she/he logs in via ssh without setting an -X tunnel, the default interactive backend will simply fail. But that's left to the user now, no longer output of, which seemed to annoy you (I can understand that)

... WARNING: did not find DISPLAY variable needed for interactive plotting
--> try ssh with `-X` or `-Y`
    setting `sett.savefigs = True`

here’s code to detect if we’re in a notebook or not.

we could use that to decide if we want to use tqdm_notebook or tqdm

Ah, looks great! Better than what I do right now. matplotlibs is_interactive() is not useful for this.

The solution I have now does the job when comparing jupyter and running a script. But it will fail in the regular Python shell.

Btw: I plan to release version 0.1 later today. Together with benchmarks and many examples for the 10x datasets. Any objections?

protip: when wanting to link to lines, press the y key to change the url from …/master/… to …/<sha1sum>/…. that way the copied URL will stay valid and not break once the file is modified on the master branch.

no objections to the release!