Fix linear_transformation bug #40905
Conversation
…orialFreeModule bug
|
Great! Could you also provide a doctest? I am not very familiar with this part of sage, but I would have expected from the documentation that something like should work, but it doesn't. Am I doing something wrong? |
|
PS: unfortunately, I cannot approve workflow runs. |
|
Well there are a few issues... First, you'll want to use sage: V = VectorSpace(QQ, IntegerRange(3))For the images, if you map all basis elements to V.term(0), you'd get a transformation that collapses everything to a 1-dimensional space. For an identity-like transformation, you'd want: sage: im = [V.term(0), V.term(1), V.term(2)]Finally, without specifying the category, you'd get the wrong type of morphism. Here's what should work now: sage: V = VectorSpace(QQ, IntegerRange(3))
sage: H = Hom(V, V, category=VectorSpaces(QQ))
sage: im = [V.term(0), V.term(1), V.term(2)] # identity mapping
sage: lt = H(im)
sage: type(lt) # Should be VectorSpaceMorphism
sage: lt(V.an_element()) # This used to crash, now worksI'll provide a doctest too :) |
Yes, I was sloppy.
Yes, but that should be OK, it's a good linear transformation. In any case, also the identity transformation doesn't work, if provided as a list of images. |
|
I think it should work now, I ran some tests. Thanks for catching it out |
Co-authored-by: Martin Rubey <axiomize@yahoo.de>
Co-authored-by: Martin Rubey <axiomize@yahoo.de>
|
I think everything should be fine now. I fixed all the indentations and error messages to conform to the codebase. |
|
|
|
@fchapoton, could you approve the workflow here? |
|
Thank you very much for the quick fix! That aside, I am a bit confused why the use of the meathod Furthermore, is there a shorter way to make the matrix sparse other than doing the matrix construction before calling Thank you! |
|
Um so maybe I did mess up for the matrix orientation But c.coordinates does work because in Sage, both VectorSpace and CombinatorialFreeModule provide a .coordinates() method. So this usage is supported for both. And on the making the matrix sparse You can pass sparse=True directly to the matrix constructor. Example: matrix(D.base_ring(), D.dimension(), C.dimension(), ..., sparse=True) |
|
Weirdly, i get án error when trying to call saying that And concerning sparsity, I was referring to the case when one passes a list of images to |
|
Yes The problem was that CombinatorialFreeModule (which is what I think it's solved |
I get many test failures now (using |
|
Awesome, Meanwhile, do you think that it may be better for performance to use the |
|
I feel ignored. @DolphinTechCodes, @breatn, could you please respond? |
|
Sorry, I do not have any sage development environment set up, so i cannot tell you anything about tests (I just looked at the modified file) |
OK, although I don't quite understand. How are you testing your code then? In any case, it does not work at all, currently. Could you please confirm that you are not a LLM? (with very few exceptions, we develop with our real name here) |
|
Tbh, I just looked at the modified parts of the code and tested snippets in a jupyter notebook. I do not have a sage development environment set up so I do not actively make changes to the codebase, rather just 'consulting' bratn (who thankfully took on the effort to fix this issue). Aside from that, I hereby confirm that I am not an LLM (however I dont know if that is sufficient proof for you). Is it really a requirement to have a github account linked to my name? |
|
Documentation preview for this PR (built with commit f124c66; changes) is ready! 🎉 |
OK. I don't see how a human could reasonably develop (or review other peoples code) without development environment. Thus, please refrain from praise (except specifically for the fact that someone is trying to fix something), it is misleading. What is keeping you from compiling sage on your computer?
I haven't met an LLM yet that would lie (without being instructed to) about itself. No, it's not a requirement, although I personally tend to trust accounts with real name more easily. |
|
Im trying to download sage but there are so many things to download and the compilation took hours and hours. |
Then, most likely, something went wrong. What operating system? |
|
macOS. But its an old intel computer |
|
OK. Note that if you only modify |
|
i've fixed the errors and it compiles on my machine |
|
*and runs |
|
I don't exactly know what you did, but it doesn't even get started here (and since this is python only, probably nowhere). If you look at the diff ("Files changed" tab on github), you will see that several blank lines were removed that shouldn't be removed. What editor are you using? (I recomment emacs + sage-shell-mode) Please run the testsuite before claiming that it works! (I am very grateful that you are working on the issue, I do realize that the road is very bumpy at the beginning. I hope I do not sound too harsh.) |
|
Sorry. This might be obvious but why is there a segmentation error in the executable: ./sage0 signals.cpython-312-darwin.so 0x0000000105756850 sigdie + 92
|
|
I'm sorry, I have no idea. I don't have a mac. Did it work before you changed the source code? |
|
This is the base source code. Even chatGPT and Claude can't help :( |
|
sage-support, most likely. Or linux. |
This fix enables consistent behavior for linear transformations across all types of vector spaces in SageMath, resolving evaluation failures for vector spaces with combinatorial bases while maintaining full backward compatibility.
This fixes this bug: #40847
📝 Checklist
⌛ Dependencies