Meson: compile without embed-positions

[tobiasdiez]

Fixes #39735.

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[user202729]

This is obviously a bad idea (for one when you type function?? into the command-line for certain functions you will no longer be able to view the source code, I think). Rather you should probably modify the code somehow to gracefully fail when the source code cannot be found.

Side note: Assume #39279 is merged, I guess.

[github-actions]

Documentation preview for this PR (built with commit 540883d; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[tobiasdiez]

This is obviously a bad idea

I'm still trying to understand what this option actually does (or better: where the file path is actually used in sage). Since you worked on this recently, can you please expand why you think it is a bad idea.

(for one when you type function?? into the command-line for certain functions you will no longer be able to view the source code, I think). Rather you should probably modify the code somehow to gracefully fail when the source code cannot be found.

Since the pyx files are not distributed in the wheel, this info wouldn't be available anyway in a "normal" installation. It only makes sense for editable installs, where one could probably fall-back to SAGE_SRC anyway.

[user202729]

expand why you think it is a bad idea

Looks like you already got the point, that ?? depends on this feature because some ipython hook modifies the imports in ipython to point to functions in sageinspect which parses this docstring.

p/s, no offense intended. Sorry if it came across as such.

Since the pyx files are not distributed in the wheel, this info wouldn't be available anyway in a "normal" installation

Yes I got that, but removing the embedding would also break ?? in cases where pyx files are available right?

[tornaria]

Could we just ship the .pyx files (and make sure there are no absolute paths)? We definitely want ?? working in any case, don't we?

[tobiasdiez]

expand why you think it is a bad idea

Looks like you already got the point, that ?? depends on this feature because some ipython hook modifies the imports in ipython to point to functions in sageinspect which parses this docstring.

p/s, no offense intended. Sorry if it came across as such.

No offense taken! Just wanted to make sure that only the feature ?? is affected.

Since the pyx files are not distributed in the wheel, this info wouldn't be available anyway in a "normal" installation

Yes I got that, but removing the embedding would also break ?? in cases where pyx files are available right?

I thought that your recent PR for the special handling of the editable meson install would work here (obviously, the broken CI proofs that my assumption is wrong). So the idea would be: if it's an editable install, then use the module name relative to SAGE_SRC as the source file. Would that work?

Could we just ship the .pyx files (and make sure there are no absolute paths)? We definitely want ?? working in any case, don't we?

I just tried numpy.random.random_sample?? and it doesn't show any source code info (while numpy.amax?? does). So even if ?? doesn't work for cython method, it at least wouldn't be a worse user experience than for numpy.
But yes, probably we could ship .pyx and use importlib to find them.

[tornaria]

Could we just ship the .pyx files (and make sure there are no absolute paths)? We definitely want ?? working in any case, don't we?

I just tried numpy.random.random_sample?? and it doesn't show any source code info (while numpy.amax?? does). So even if ?? doesn't work for cython method, it at least wouldn't be a worse user experience than for numpy. But yes, probably we could ship .pyx and use importlib to find them.

I'm talking about sage. Things like:

sage: prime_pi??
Type:           PrimePi
String form:    prime_pi
File:           /usr/lib/python3.13/site-packages/sage/functions/prime_pi.pyx
Source:        
cdef class PrimePi(BuiltinFunction):
[...]

[user202729]

Just wanted to make sure that only the feature ?? is affected

More like it's the only feature that I have come across that make use of this. There might be more that I'm not aware of.

I just tried numpy.random.random_sample?? and it doesn't show any source code info (while numpy.amax?? does).

The problem is because numpy does not embed the signature or the source location in its docstring, so we have no choice.

https://github.com/numpy/numpy/blob/main/numpy/random/mtrand.pyx#L389

sage: print(np.random.random_sample.__doc__)

        random_sample(size=None)

        Return random floats in the half-open interval [0.0, 1.0).

Compare:

sage: print(prime_pi.__call__.__doc__)
File: /…/sage/src/sage/functions/prime_pi.pyx (starting at line 91)

        EXAMPLES::

            sage: # needs sage.symbolic

By the way, #39279 will show more information (signature and call signature) in ??.

[tobiasdiez]

Okay, let's close it for now. Once mesonbuild/meson-python#256 is implemented, we could disable the embedding of the filename for non-editable installs.