maxima interface gives precedence to function dictionary instead of local variables

[burcin]

From the sage-devel thread:

​http://groups.google.com/group/sage-devel/t/c89582242c83a349

On Fri, 11 Dec 2009 13:46:31 +0100
Nathann Cohen <nathann.cohen@gmail.com> wrote:

> sage: var('delta k')
> sage: m1=2*delta**2 + 2**2*delta*k
> sage: n=delta*k+2
> sage: m2=(2*delta)**2+(k-1)*4
> sage: m=(delta+delta*k-(delta-1))
> sage: ((m1/n)-(m2/n)).expand().simplify()

On 4.3.rc0, I get this:

TypeError: unsupported operand parent(s) for '*': 'Symbolic Ring' and
'<class 'sage.functions.generalized.FunctionDiracDelta'>'

The Maxima interface seems to give precedence to the global function dictionary instead of the local variables when converting Maxima output back to Sage expressions.

sage: dirac_delta(x)
dirac_delta(x)
sage: maxima(dirac_delta(x))
delta(x)

[was]

People run into this all the time, evidently:

[15:21] --> SageWWW has joined this channel (~SageWWW@64.241.37.140).
[15:23] <SageWWW> hey guys.  what do you think about http://pastebin.ca/1772520
[15:24] <SageWWW> d = var('delta'), so now d is a reference to a sage.symbolic.expression.Expression
[15:25] <SageWWW> but when we try to add it to something else, it thinks its a sage.functions.generalized.FunctionDiracDelta
[15:27] <wstein> http://trac.sagemath.org/sage_trac/ticket/7661

[eigenlambda]

sage: d = var('delta') sage: e = d._maxima_() sage: sage.calculus.calculus.symbolic_expression_from_maxima_element(e) dirac_delta

somewhere in symbolic_expression_from_maxima_element(), the string from maxima is looked up in sage.calculus.calculus._syms, which by default has 'delta': dirac_delta . So this is what's happening, next, SR() barfs on trying to turn dirac_delta into a symbolic expression, at which point people who just wanted their variable 'delta' back get confused and frustrated.

sage: del sage.calculus.calculus._symsdelta? sage: sage.calculus.calculus.symbolic_expression_from_maxima_element(e) delta

That may not be such a good idea, however, since what sage calls dirac_delta, maxima refers to as delta. Nevertheless, since reset('delta') appears to remove delta from that dictionary, perhaps var('delta') should also do so?

Of course, what happens when someone does a Laplace transform with delta as a sage variable will then come out confusing and wrong. At least the current behavior is merely broken.

[burcin]

attachment:trac_7661-maxima_convert_back.patch​ fixes the problem reported above, and a similar problem with function conversions back from maxima reported in comment:2:ticket:8459.

[burcin]

I updated attachment:trac_7661-maxima_convert_back.patch​ to remove a doctest fix broken by a previous patch in my queue (#6949, symbol... line in sage/symbolic/ring.pyx).

This patch depends on #7748.

[robert.marik]

Thanks for working onthis. Is #7748 the only prerequisity? I installed three patches as described at #7748 and got the following error

patching file sage/calculus/calculus.py
Hunk #3 succeeded at 1414 with fuzz 1 (offset -4 lines).
Hunk #5 FAILED at 1455
1 out of 14 hunks FAILED -- saving rejects to file sage/calculus/calculus.py.rej
abort: patch failed to apply

[burcin]

AFAICT, #8237 also changes that code. Can you try with #8237 applied?

I'm sorry for the dependency hell we get into with these patches for every release. I don't know any way to automatically get a list of dependencies for a patch in my queue.

Thanks for your time Robert.

[robert.marik]

Hello Burcin

I think that two lines should be removed from the patch

global _syms 
_syms = sage.symbolic.pynac.symbol_table.get('maxima', {}) 

I updated your patch, it is now ​http://user.mendelu.cz/marik/sage/trac_7661-maxima_convert_back2.patch

If everything will work, I'll return in few (several) hours with positive review (tests in functions, interfaces, symbolics and calculus passed, now running all the test).

Robert

[burcin]

OK. That is one approach to solving this problem. Now we need to rebase the patch at #8237 so that it applies on top of these. Removing the offending hunk from calculus.py should be enough for that.

Note that your updated patch shows you as the author. I'd appreciate it if you changed that back.

Thanks.

[robert.marik]

Sure, it was intended as temporary patch and from this reason I did not upload to trac server unless tested. I got some doctest failures in three files. See ​http://boxen.math.washington.edu/home/marik/ and the files a, b and c.

I think that b is simple to fix, but have no idea about a and c.

[burcin]

apply only this patch

[burcin]

Thanks a lot for the quick feedback.

• a is because you have the pynac package from #8644 installed, but not the corresponding patch from #8565.
• b is a simple import problem, fixed by the updated attachment:trac_7661-maxima_convert_back.take2.patch​
• I can't reproduce c here. It doesn't seem to be related to the changes in ticket. Do you have any other patches applied?

[robert.marik]

I tested it on a fresh install and seems that all a,b,c are resolved. I am running all tests again, to be sure :)

[robert.marik]

Tests passed, postive review, thanks for fixing - very very usefull ticket.

Release manager: apply only trac_7661-maxima_convert_back.take2.patch

[jhpalmieri]

Merged "trac_7661-maxima_convert_back.take2.patch" in 4.4.alpha0