add if for run-ext.txt

[Tieqiong]

closes #55

[bobleesj]

closes #55

I thought of this, but don't you also have to apply if then for pip.txt?

[bobleesj]

Also 2 other yml files also need "if then" if we decide to implement this

[bobleesj]

@Tieqiong @sbillinge

What if we use a single command to install all conda packages at once so that the dependency sorting can be handled at once?

conda install \                          
    --file requirements/run.txt \                  
    --file requirements/run-ext.txt \                  
    --file requirements/test.txt \
    --file requirements/pip.txt \
    --file requirements/build.txt

[sbillinge]

Can you say some more words about what problem this solves? I am not sure what you mean by "dependency solving"?

[bobleesj]

@sbillinge

Can you say some more words about what problem this solves? I am not sure what you mean by "dependency solving"?

When multiple packages listed in files like text.txt, build.txt, and run.txt are installed together in a single command, Conda can more effectively analyze and resolve the dependencies of all packages in one unified process.

This means there won't be instances where downgrades or upgrades are needed between separate Conda install commands.

[sbillinge]

I see, that makes sense then. However it presumably makes more difficult, though not impossible, conditional installs?

[sbillinge]

I wonder if it is better to not have the run-ext.txt in the cookiecutter

[sbillinge]

sorry, clicked the wrong thing there by mistake.

[bobleesj]

what if we include run-ext.txt dependencies under run.txt? This was a special case for diffpy.srfit

[sbillinge]

what if we include run-ext.txt dependencies under run.txt? This was a special case for diffpy.srfit

sorry, not sure what you mean here. Normally I would say that edge cases can deviate from the cookie-cutter if they are rare. We like to keep such cases to a minimum. But because we have the release scripts centralized, we would have to have a separate yaml presumably for each edge case, either residing in the local directory of the edge case, or here centrally, but the local yml updated to call a differently named yaml from here? Or we can find a way to catch all the edge cases with conditionals (which also adds unwanted complexity).

[sbillinge]

Closing on basis that we have change this to conda/pip