Add the drizzle library to requirements.txt. It's on pypi, but triggered some compilation when I installed it, so it might not work on all platforms. An alternative would be to avoid the drizzle import until runtime, but users will need it to simulate images in our default configuration.
Add minimum dependency versions to requirement.txt, and load requirements.txt in setup.py. Maybe you'll eventually want two package lists (a minimum one and one for automated test builds).
Avoid runtime tensorflow imports, by making manada.Analysis unavailable if tensorflow is missing.
If you don't have tensorflow, you'll get a notification manada.Analysis disabled since tensorflow is missing when importing manada.
If you later try manada.Analysis (which works fine if you have tensorflow) you'll get an AttributeError. (Since this is rather uninformative, I thought the notification was warranted.)
import manada.Analysis, from manada import Analysis, etc. will still trigger the tensorflow ModuleNotFound error if you don't have tensorflow.
This aims to make manada easier to install / use:
drizzlelibrary torequirements.txt. It's on pypi, but triggered some compilation when I installed it, so it might not work on all platforms. An alternative would be to avoid the drizzle import until runtime, but users will need it to simulate images in our default configuration.requirement.txt, and loadrequirements.txtinsetup.py. Maybe you'll eventually want two package lists (a minimum one and one for automated test builds).manada.Analysisunavailable if tensorflow is missing.manada.Analysis disabled since tensorflow is missingwhen importing manada.manada.Analysis(which works fine if you have tensorflow) you'll get anAttributeError. (Since this is rather uninformative, I thought the notification was warranted.)import manada.Analysis,from manada import Analysis, etc. will still trigger the tensorflowModuleNotFounderror if you don't have tensorflow.