Uh oh!
There was an error while loading. Please reload this page.
add dynamic dependency in pyproject.toml - #98
Conversation
bobleesj
commented
Sep 3, 2024
@Tieqiong Noticed the same bug with I checked cookiecutter`s pyproject.toml that it doesn't have the dependency install section. |
sbillinge
commented
Sep 3, 2024
Thanks @Tieqiong I will test |
sbillinge
commented
Sep 3, 2024
Very slow internet here so not sure if I can test befoer I have to board plane. But in answer to the other questions, I agree that adding a |
sbillinge
commented
Sep 3, 2024
OK, I tested on local and it worked. I will push out a release bump to pypi if I can (internet!) |
sbillinge
commented
Sep 3, 2024
@Tieqiong@bobleesj please can you test by pip installing I couldn't build the pycifrw dependency on my windows computer. I get: (not the full trace) which is pretty annoying. |
sbillinge
commented
Sep 3, 2024
I guess this is because pycifrw only deploys their code to pypi, as we do for pdffit2, and so a pip install of it tries to build locally and I don't have the right compiler. |
bobleesj
commented
Sep 3, 2024
It works for me. Mac M1 Testing the import |
Tieqiong
commented
Sep 3, 2024
@sbillinge It'll work with the Microsoft c++ compiler installed. |
sbillinge
commented
Sep 3, 2024
I don't think this is testing what we want. You have to do an install in a new clean empty environment. |
bobleesj
commented
Sep 3, 2024
@sbillinge Okay. pip install test passed from a new env using the recent 3.2.1 version. |
@sbillinge I think in fact we can't include ![]() so I guess instead of skipping This is specific to ![]() |
sbillinge
commented
Sep 4, 2024
so is everything fixed if we just remove python from build.txt? This seems like a good solution? |
Tieqiong
commented
Sep 4, 2024
as far as I know yes I think this will get rid of the error. However I'm not exactly sure if we want to remove it because I have no idea why it was included in the first place... |
sbillinge
commented
Sep 4, 2024
Yah, if we don't understand what is going on we could just be introducing more problems |
@sbillinge So I searched around online and asked Andrew, it seems like there's nothing special with this |
sbillinge
commented
Sep 5, 2024
What do you mean by "nothing special"? |
Tieqiong
commented
Sep 5, 2024
I didn't find any documentation that have it as convention, and according to Andrew:
So by "nothing special" I mean there's no hidden mechanism that require it to be there |
Tieqiong
commented
Sep 5, 2024
I guess a better way to go is to figure out how to build wheel so that build.txt would not be needed for pip install for sure. |
sbillinge
commented
Sep 6, 2024
Do you mean that it doesn't have to be there? |
sbillinge
commented
Sep 6, 2024
@Tieqiong please can you propose a solution for what you want the yml to look like? We will work on building the wheels but we also want the source bundle to be right either way. |
Tieqiong
commented
Sep 6, 2024
Yes it doesn't have to be there, but whether it should be there depends on how we are using the file (I think we are talking about build.txt not yml file right?). If we use the file as a source of information during pip installation, then we need to remove python because it'll cause error for pip. |
sbillinge
commented
Sep 6, 2024
No human will read it, so let's get rid of the python in there. We want pip to be able to build from sources even if we distribute wheels |



closes#97
@sbillinge Please check and better test it on your setup.
We probably want this in the cookiecutter too, maybe also change the worked packages.
Another thing that might worth notice is do we want to include
requirements/build.txt? This is specific todiffpy.pdffit2'sgsldependency. Howeversetuptoolandpythonare included, which in theory should already be present if a user is installing this package to a conda environment with python installed.