maxima interface severely broken for large input, etc.

[williamstein]

This doesn't work

~/tmp$ sage
│ Sage Version 6.4, Release Date: 2014-11-14                         │
...
sage: maxima.eval("""RedGauss(A,variable):=block([q,var,listeD,listeQ,i,a,l,D],
                            quad:expand(transpose(variable).A.variable),
                            q:quad,var:variable,listeD:[],listeQ:[],D:zeromatrix(length(A),length(A)),
                            while var#[] do (
                            a:diff(q,var[1],2)/2,l:subst(0,var[1],diff(q,var[1])),
    if a=0 and l=0 then var:rest(var)
    elseif a#0 then (
                   listeD:append(listeD,[a]),listeQ:append(listeQ,[var[1]+l/(2*a)=0]),
                   q:subst(0,var[1],q)-l^2/(4*a),var:rest(var))
    else (
        i:2,while subst(0,var[i],diff(l,var[i]))=0 do i:i+1,
        a:diff(q,var[i],2)/2,
        if a#0 then (
        listeD:append(liste,[a]),listeQ:append(listeQ,[var[i]+l/(2*a)=0]),q:subst(0,var[i],q)-l^2/(4*a),
        var:append(rest(var,i-1-length(var)),rest(var,i)))
    else (
        a:diff(q,var[1],1,var[i],1),
        l1:subst([var[1]=0,var[i]=0],diff(q,var[1])),li:subst([var[1]=0,var[i]=0],diff(q,var[i])),
        listeD:append(listeD,[a/4,-a/4]),
        listeQ:append(listeQ,[var[1]+var[i]+(l1+li)/a=0,var[1]-var[i]+(li-l1)/a=0]),
        q:subst([var[1]=0,var[i]=0],q)-l1*li/a,
        var:append(rest(rest(var,i-1-length(var))),rest(var,i))))),
    for k:1 thru length(listeD) do D[k,k]:listeD[k],
            return([D,transpose(coefmatrix(listeQ,variable))]))$""") 

But I guess it should:

Maxima 5.34.1 http://maxima.sourceforge.net
using Lisp ECL 13.5.1
Distributed under the GNU Public License. See the file COPYING.
Dedicated to the memory of William Schelter.
The function bug_report() provides bug reporting information.
(%i1) RedGauss(A,variable):=block([q,var,listeD,listeQ,i,a,l,D],
                            quad:expand(transpose(variable).A.variable),
                            q:quad,var:variable,listeD:[],listeQ:[],D:zeromatrix(length(A),length(A)),
                            while var#[] do (
                            a:diff(q,var[1],2)/2,l:subst(0,var[1],diff(q,var[1])),
    if a=0 and l=0 then var:rest(var)
    elseif a#0 then (
                   listeD:append(listeD,[a]),listeQ:append(listeQ,[var[1]+l/(2*a)=0]),
                   q:subst(0,var[1],q)-l^2/(4*a),var:rest(var))
    else (
        i:2,while subst(0,var[i],diff(l,var[i]))=0 do i:i+1,
        a:diff(q,var[i],2)/2,
        if a#0 then (
        listeD:append(liste,[a]),listeQ:append(listeQ,[var[i]+l/(2*a)=0]),q:subst(0,var[i],q)-l^2/(4*a),
        var:append(rest(var,i-1-length(var)),rest(var,i)))
    else (
        a:diff(q,var[1],1,var[i],1),
        l1:subst([var[1]=0,var[i]=0],diff(q,var[1])),li:subst([var[1]=0,var[i]=0],diff(q,var[i])),
        listeD:append(listeD,[a/4,-a/4]),
        listeQ:append(listeQ,[var[1]+var[i]+(l1+li)/a=0,var[1]-var[i]+(li-l1)/a=0]),
        q:subst([var[1]=0,var[i]=0],q)-l1*li/a,
        var:append(rest(rest(var,i-1-length(var))),rest(var,i))))),
    for k:1 thru length(listeD) do D[k,k]:listeD[k],
            return([D,transpose(coefmatrix(listeQ,variable))]))$
 
(%i2)

Issues: maxima._read_in_file_command isn't implemented.

sage: maxima._read_in_file_command('x')
...
NotImplementedError:

But even if you don't use that (instead hard coding the eval cutoff to something large), things are still broken.

This might be related to #1660. This was first reported here:

http://ask.sagemath.org/question/24955/maxima-on-sagecloud/

since people are starting to use SageMathCloud as a Maxima notebook, which it would be quite good for, if only the maxima interface actually worked for large input. (It's very robust for small input -- just not large input.)

Component: interfaces

Issue created by migration from https://trac.sagemath.org/ticket/17388

[nbruin]

comment:1

Adding

    def _read_in_file_command(self, filename):
        return 'batchload("%s");'%filename

into maxima_abstract seems to do the trick. It returns the temporary filename, so perhaps something should be consumed from the output. I don't know what "load" is supposed to consume.

Also, perhaps it's better to put it in maxima.py, because maxima_lib.py has very little business using this (indeed, the example mentioned in the ticket seems to work for maxima_lib out of the box--I don't think maxima_lib ever reverts to using files). Of course, we do so many non-standard things to maxima_lib that using that to interface with "maxima" is probably not a good idea.

Finally, someone should probably doctest this properly when implementing it.

[nbruin]

comment:2

Incidentally, if you want to use SMC as a maxima notebook, it might be nice to run maxima/SBCL (or some other lisp that performs better than ECL) as the backend. SBCL doesn't fit the sage distribution/dependency model, so for shipping sage it's better to use ECL (also because it allows us to have a library interface), but for a web service it should be straightforward to drive an expect interface to maxima/SBCL.