driveanon installation
jlandmann opened this issue · 3 comments
Hi guys!
Sorry for bursting in...
I was just trying to install the geoprofile environment from the yml file to maybe suggest some additions then. While installing, I wondered why you want to install driveanon in editable mode. To my understanding, it can breaks things, since then driveanon can be updated anytime, not even only when the environment is built new.
I was also confused that driveanon was installed right into the src directory, which usually doesn't happen.
I fear it's me, but is there an opinion on that?
Hi Johannes!
driveanon does not exist on PyPi (kind of just a quick side project) so we need to install it from the git repo directly. Making it editable is optional, but shouldn't be a problem if left unmodified.
The alternative is to add instructions to the README.md to install this dependency manually. This would allow you to put the repo outside of the geoprofile directory, instead of inside the src folder, which gets automatically created the way things are set up now in the environment.yml file.
Or - we just get rid of driveanon and find a better solution to pull and/or host test data :)
One option might be Git LFS:
Just realized a few instances of "dpet" weren't changed to "geoprofile" when I transferred the repo to GlacioHack and renamed it. Sorry about that! This should be fixed now with the most recent quick commits.
Thanks for your comprehensive reply!
I think the point for me was that git/PyCharm got confused by the 'package in the package',
For me it also worked to just change in environments.yml:
- "--editable=git+https://github.com/friedrichknuth/driveanon.git#egg=driveanon"
to
- git+https://github.com/friedrichknuth/driveanon.git#egg=driveanon
so that it's installed into your environment directory (e.g. Anaconda3\envs\geoprofile\Lib\site-packages for Anaconda on Windows).
Unless you would really change your data and paths really often, I would leave it there, but that's maybe a matter of taste.
Regarding the data storage: the way OGGM/salem (my favorites ;-)) solve it is to have a data directory and load the sample data from there into a dot-directory in the home folder. Maybe an overkill if the amount of data stays small and constant, but probably the cleanest solution.