search

nteract / jupyter-paths

:city_sunrise: Pure JavaScript implementation of jupyter --paths --json
BSD 3-Clause "New" or "Revised" License
9 stars 11 forks source link

add inexpensive guess for sysPrefix when withSysPrefix: false #15

Closed minrk closed 7 years ago

minrk commented 8 years ago

effectively does which jupyter without instantiating any subprocesses, and resolves realpath(jupyter)/../.. for the prefix.

I also removed the sort calls from the tests, which hid a bug where the ordering of system data dirs was not correct

Tests pass on OS X with conda Python, as well as virtualenvs.

Another option is to check on python, rather than jupyter. The realpath in that case will give the right answer for homebrew Python, which this doesn't (no change from master). The disadvantage is that it will include Python's prefix whether the Jupyter Python package is installed or not, and there isn't a way to identify which Python without reading the shebang line of the Jupyter script, and I wanted to make sure that no files were opened or processes executed.

If Jupyter is installed with --user, this will add the user prefix, which is not on the Python implementation's path, so will generally turn up no results.

rgbkrk commented 8 years ago
screen shot 2016-08-02 at 3 52 16 pm

Is this the case where it works for virtual envs yet not --user?

minrk commented 8 years ago

Is this the case where it works for virtual envs yet not --user?

Yes, precisely. Because of things like this, I'm not 100% sure it's a good idea to do it this way, however with the increasing use of things like conda to install kernels, I think specs are increasingly going to show up on sys.prefix, so I thought I would take a run at it.

I switched it to check for python instead of jupyter. This should get the right answer, as long as which python returns the Python for which Jupyter is installed. In fact, it should do just as good a job as sys-prefix-promise, which also relies on which python returning the right thing, so maybe that option should be dropped.

rgbkrk commented 7 years ago

Would you want to tackle this one again @minrk or should we close this?

minrk commented 7 years ago

Sorry, totally forgot this one. Should be fixed.

rgbkrk commented 7 years ago

CI not passing, I'll dig on this one later.

lgeiger commented 7 years ago

The CI failure should be fixed by #19

minrk commented 7 years ago

rebased after #19, all green.

rgbkrk commented 7 years ago

Ah, sweet