I am reaching out to seek your assistance with building a package in Conda.
I have been using Anaconda for managing my python environments and dependencies. I am currently facing difficulties in creating package using Conda build, and I’m in need of your guidance to resolve the issue.
While attempting to build the package, I encountered the following error message:
conda_build.exceptions.DependencyNeedsBuildingError: “Unsatisfied dependencies for platform win-64:(pip, python=3.10)”
In my Conda Build recipe, I have included explicit specifications for pip and python as follows:
requirements:
build:
-Python
-pip
run:
-python
I would greatly appreciate it if you could guide me through the steps to resolve this issue.
Its hard to know exactly whats happening here without seeing the full meta.yaml and the conda build command, but you may want to double check that your channels include defaults so that you get the full Anaconda respository. I can confirm that pip==23.1.2 is available on main for win-64.
I’m new here so please inform me if this is bad etiquette, but I’m having a related issue. I wanted to use conda build to build a recipe for galgebra, which is on pypi., and then upload that to Anaconda(dot)org. I was successful with the conda build in that galgebra is now available through conda install on my own computer, but I cannot figure out how to upload!?! At first I couldn’t even log in with anaconda client, but then I actually registered for an account (with anaconda(dot)org instead of anaconda(dot)cloud) and was able to login via client. But when I try to upload I get an error message “trouble reading metadata, is this a valid installer package.” My path address is C:\Users\Gateway\anaconda3\conda-bld\galgebra_1690223189722_h_env, which doesn’t have the tar.bz2 tag, which is what I think the problem is? Is this so? I’m just trying to make this public so anyone can use it if they so desire without pip install. It works on my computer. I conda installed it in an environment just fine and it works fine.
Hi Wes! Here’s some instructions on how to upload built packages to anaconda.org, hopefully that helps!
Getting started — Anaconda Documentation
Well, sure, I saw that documentation when I first registered with anaconda(dot)org, but when I use the conda build --output command I get an error message saying a Recipe Path is required. I’m pretty damn certain my recipe path is the C:\Users\ . . . .tar.gz because it includes conda-skeleton which was used to build the recipe. But this gives an error saying no such directory exists. It’s really weird. I activated this environment, my_conda_build_env, which is what I used to perform this simple build. Conda skeleton generated the tar.gz which was the recipe I used to do the actual conda build, but then everything just disappeared? I mean, the entire “my_conda_build_env” just flat disappeared!?! Of course, the package I was after, galgebra, is there in my root environment and I can conda install it in any other environment and it works fine. But the entire reason I embarked on this little adventure was to attempt to solve this apparent issue that arises when using pip install and conda install together. Apparently certain users of galgebra have had difficulties, using Ubuntu. I was interested in seeing if I could somehow conda install galgebra myself, which led me to the conda build. When that turned out successfully, I thought I could upload that rather easily to a channel so anyone could conda install, circumventing any issues. But I haven’t had any luck.
In Lee Vaughn’s book, he has a note on page 42 about doing what I am trying to do, but he’s a bit cryptic about it, i. e. nothing but the one-sentence note. I have also seen in my Navigator that they have apps specially dedicated to building and uploading packages, but I don’t know anything about those apps. I’m also curious about whether or not I can use the “modular” approach Vaughn discusses in the context of Spyder and Notebook, i.e. using a galgebra kernel to link other environments to the galgebra in my root.
Anyway, it’s not a huge deal; I just thought it would be nice to offer an easy solution to the galgebra community, the Pythonic Geometric Algebra Enthusiasts group, of which I am not even a member!?! Of course, anyone can do what I did and it was pretty much trouble free, probably because Alan Bromborsky has written some pretty clean code . . .