Make Sage work with maxima 5.47 #35707

antonio-rojas · 2023-06-01T22:17:56Z

Maxima 5.47 has been released. The main change that affects Sage is https://sourceforge.net/p/maxima/code/ci/f5ca00582c24cfedca53a664335f6c13e1e40cf9, which makes the maxima ecl library crash because the e-val variable is unassigned. We fix this by running the initialize-runtime-globals function on load.

The remaining issues are a few harmless test failures caused by numerical noise and output format changes (such as additional parentheses near minus signs)

…s with maxima 5.47

src/doc/de/tutorial/interfaces.rst

src/doc/de/tutorial/tour_algebra.rst

src/sage/calculus/calculus.py

src/sage/functions/bessel.py

tornaria · 2023-06-02T00:08:45Z

Just FTR, doctests pass for me with 5.47.0 (normal and long) with the current PR. My suggestions are merely cosmetic, and I wouldn't mind keeping it as it is (all these testing is horrible anyway).

Once we settle on something, I could try to run tests on 5.47.0 / 5.46.0 / 5.45.1, just to make sure it's ok.

kiwifb · 2023-06-02T01:31:33Z

If the appending of .sage() fix most of the brackets, appearing and disappearing then it is a much better solution. It makes the answers much clearer.

Another solution would be subtracting the know answer and simplify it to 0. That can only work for exact results of course.

src/doc/en/constructions/linear_algebra.rst

tornaria · 2023-06-02T21:41:55Z

src/sage/functions/orthogonal_polys.py

@@ -974,7 +974,7 @@ def __init__(self):
            sage: chebyshev_U(x, x)._sympy_()
            chebyshevu(x, x)
            sage: maxima(chebyshev_U(2,x, hold=True))
-            3*((-(8*(1-_SAGE_VAR_x))/3)+(4*(1-_SAGE_VAR_x)^2)/3+1)
+            3*(...-...(8*(1-_SAGE_VAR_x))/3)+(4*(1-_SAGE_VAR_x)^2)/3+1)


sage: maxima(chebyshev_U(2,x, hold=True)).sage() 4*(x - 1)^2 + 8*x - 5

I don't know what this is trying to acomplish... I guess that maxima knows how to handle chebyshev_U(2, x).

I guess it's testing conversion to maxima, so I'm not sure converting back to sage the result makes much sense

Yeah, it's supposed to be an example of the __init__ method. Why send it to maxima at all?

src/sage/functions/special.py

src/sage/interfaces/interface.py

tornaria · 2023-06-02T22:29:45Z

Summary: don't waste too much time in this... Maybe just put the "new correct" answer in the elliptic_e test (with the # not tested b/c old maxima is broken tag) just so that the thing that shows up in the documentation is correct (and the thing in the comment seems to be incorrect (???).

src/doc/en/tutorial/interfaces.rst

src/sage/functions/bessel.py

src/sage/functions/other.py

src/sage/functions/special.py

src/sage/interfaces/maxima.py

src/sage/interfaces/maxima_abstract.py

src/sage/modules/free_module_element.pyx

src/sage/symbolic/relation.py

orlitzky · 2023-06-03T00:44:27Z

src/doc/en/tutorial/interfaces.rst

-    sage: A.eigenvectors()
-    [[[0,4],[3,1]],[[[1,0,0,-4],[0,1,0,-2],[0,0,1,-4/3]],[[1,2,3,4]]]]
+    sage: A.eigenvectors().sage()
+    [[[0, 4], [3, 1]], [[[1, 0, 0, -4], [0, 1, 0, -2], [0, 0, 1, -4/3]], [[1, 2, 3, 4]]]]


This is difficult to test correctly because the eigenvalues, eigenvectors, and multiplicities are split across three lists and need to be matched up. The eigenvectors I would not expect to be unique anyway.

Lazy fix: switch it to a matrix with only nonnegative eigenvalues and a set of eigenvectors whose entries can be chosen integral and nonnegative.

src/sage/functions/bessel.py

orlitzky · 2023-06-08T20:28:41Z

src/doc/en/constructions/linear_algebra.rst

@@ -277,8 +277,8 @@ Another approach is to use the interface with Maxima:

    sage: A = maxima("matrix ([1, -4], [1, -1])")
    sage: eig = A.eigenvectors()
-    sage: eig
-    [[[-sqrt(3)*%i,sqrt(3)*%i],[1,1]], [[[1,(sqrt(3)*%i+1)/4]],[[1,-(sqrt(3)*%i-1)/4]]]]


I thought I commented on this one already, but don't see it. I don't know why we're telling people to use the low-level maxima interface to compute the eigenvalues of a 2x2 matrix in a linear algebra tutorial. Personally I would just delete it (and the surrounding explanation).

Then the gap section immediately after that should be dropped too. I think we should leave any further doc cleanup for a follow up ticket.

src/doc/en/tutorial/tour_algebra.rst

orlitzky

Ok, we shouldn't hold this up any longer. The first commit is pretty important. Thanks.

tornaria · 2023-06-25T02:31:05Z

src/sage/interfaces/maxima_lib.py

@@ -133,6 +133,7 @@
    ecl_eval("(require 'maxima \"{}\")".format(MAXIMA_FAS))
 else:
    ecl_eval("(require 'maxima)")
+ecl_eval("(maxima::initialize-runtime-globals)")


It seems this is unfixing #26968, as seen here: https://github.com/void-linux/void-packages/actions/runs/5366826557/jobs/9736459766?pr=44624#step:7:43488

Presumably, calling initialize-runtime-globals will run set-pathnames and fail without the workaround that is in the try block... Can this be moved after that?

My suggestion:

--- a/src/sage/interfaces/maxima_lib.py +++ b/src/sage/interfaces/maxima_lib.py @@ -133,12 +133,11 @@ if MAXIMA_FAS: ecl_eval("(require 'maxima \"{}\")".format(MAXIMA_FAS)) else: ecl_eval("(require 'maxima)") -ecl_eval("(maxima::initialize-runtime-globals)") ecl_eval("(in-package :maxima)") -ecl_eval("(setq $nolabels t))") -ecl_eval("(defvar *MAXIMA-LANG-SUBDIR* NIL)") ecl_eval("(set-locale-subdir)") +# this has to happen early so the workaround works +# do not add other initialization before this block try: ecl_eval("(set-pathnames)") except RuntimeError: @@ -155,6 +154,8 @@ except RuntimeError: # Call `(set-pathnames)` again to complete its job. ecl_eval("(set-pathnames)") +ecl_eval("(initialize-runtime-globals)") +ecl_eval("(setq $nolabels t))") ecl_eval("(defun add-lineinfo (x) x)") ecl_eval('(defun principal nil (cond ($noprincipal (diverg)) ((not pcprntd) (merror "Divergent Integral"))))') ecl_eval("(remprop 'mfactorial 'grind)") # don't use ! for factorials (#11539)

Calling initialize-runtime-globals will run set-pathnames and be subject to the issue described in sagemath#26968. Thus the workaround introduced in sagemath#35195 has to be done before anything that may call set-pathnames (e.g. initialize-runtime-globals).

tornaria · 2023-06-25T16:44:47Z

Please merge antonio-rojas#2 into this PR before merging this PR in main.

Make sure sagemath#26968 is not unfixed after sagemath#35707.

github-actions · 2023-06-25T23:50:50Z

Documentation preview for this PR (built with commit 808ab47; changes) is ready! 🎉

followup to sagemath#35707 (make sage maxima 5.47-compatible) fixes sagemath#38599

followup to sagemath#35707 (make sage maxima 5.47-compatible) fixes sagemath#38599 ### 📝 Checklist  - [x] The title is concise and informative. - [x] The description explains in detail what this PR is about. - [x] 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    URL: sagemath#38601 Reported by: Dima Pasechnik Reviewer(s): Kwankyu Lee

antonio-rojas added 3 commits June 1, 2023 22:34

Run initialize-runtime-globals on maxima initialization to fix crashe…

7b35369

…s with maxima 5.47

Make tests pass with maxima 5.47

2b57a08

Fix one more test with maxima 5.47

ffff27d

antonio-rojas requested review from kiwifb, dimpase, orlitzky and tornaria June 1, 2023 22:18

antonio-rojas added c: packages: standard s: needs review labels Jun 1, 2023

tornaria reviewed Jun 2, 2023

View reviewed changes

Nicer fixes for tests

ee0928e

tornaria reviewed Jun 2, 2023

View reviewed changes

src/doc/en/constructions/linear_algebra.rst Outdated Show resolved Hide resolved

tornaria reviewed Jun 2, 2023

View reviewed changes

src/sage/functions/special.py Outdated Show resolved Hide resolved

tornaria reviewed Jun 2, 2023

View reviewed changes

src/sage/interfaces/interface.py Outdated Show resolved Hide resolved