search

SyneRBI / SIRF-Exercises

SIRF Training and demonstration material
http://www.ccpsynerbi.ac.uk
Apache License 2.0
18 stars 21 forks source link

Fix #97: only-binary in requirements.txt #127

Closed ashgillman closed 1 year ago

ashgillman commented 3 years ago

Tested on Docker sirf:latest (not service), installed successfully without --only-binary

KrisThielemans commented 3 years ago

question is of course if it installed everything that we need, as you've removed some stuff. Can you confirm that the other packages are actually there?

KrisThielemans commented 3 years ago

or just list them anyway (might be easiest)

ashgillman commented 3 years ago

The other packages that I removed are not explicit dependencies - we never import them, so we only want them if the other dependencies need them. I included them only to make sure they were binary only.

No matter, I've checked the pip install output and can confirm that with the current pip they are still installed

KrisThielemans commented 3 years ago

As this isn't going to fix the conda as it doesn't have brainweb, I suggest we leave this for 3.2

ashgillman commented 3 years ago

I may be confused here, but usually conda doesn't interface directly with a requirement.txt? You would use pip from within conda, so it should be available via pip?

KrisThielemans commented 3 years ago

sure. It's just that the docker script uses requirements.txt with conda install -f as a first attempt, and then falls back to pip. the latter works fine.

anyway, let's forget about this for now

ashgillman commented 3 years ago

Yes I agree, above comment was for later :)

paskino commented 3 years ago

@KrisThielemans @ashgillman in the creation of the jupyterhub image for the training I remove the --only-binary during installation, so conda may succeed.

For the requirements of @gschramm's pyapetnet, I install from requirements-pyapetnet.txt all that it's possible via conda, and only what's not possible via pip, by reading the file line by line.

KrisThielemans commented 3 years ago

@KrisThielemans @ashgillman in the creation of the jupyterhub image for the training I remove the --only-binary during installation, so conda may succeed.

When I checked, it doesn't as brainweb isn't in conda. Anyway...

paskino commented 3 years ago

Yes, I should do the line-by-line also there.

KrisThielemans commented 3 years ago

Yes, I should do the line-by-line also there.

except that conda recommends to do all installs together such that it can resolve versions better...

package management isn't easy, which is something that we know by now!

KrisThielemans commented 1 year ago

fixed in #187