Alejandro Jakubi

MaplePrimes Activity


These are replies submitted by Alejandro Jakubi

@H-R 

Alex has been very diplomatic in describing the situation of the latex command... My assesment is that it has been a disaster for many years. And actually, additional regression bugs have been reported for several recent versions (including Maple 18). 

@H-R 

For linear combination factors you may try a wrapper procedure along these lines:

> collect2:=proc(ex,f)
> local u,ex1:=algsubs(f=u,ex),ex2;
> ex2:=collect(ex1,u);
> subs(u=f,ex2);
> end proc:
> 
> F:=(a*x+b*y)*m*n+(a*x+b*y)*p*q:
> collect2(F,a*x+b*y);
                            (m n + p q) (a x + b y)

@H-R 

A parametrized rule, as I have already shown to you, may help where collect fails:

> f:=(m+n)*b*x+(m+n)*b*y:
> applyrule(fac(n+m),f);
                              (m + n) (b x + b y)

You may also write a wrapper procedure for collect that allows collecting on expressions like m + n (I have done such things many years ago).

@Joe Riel 

I think that it would be much better that such information (which package provided the extension to diff that allowed that) were provided in ?diff , rather than relying on the memory of the user...

@Axel Vogt 

No, something "normal". Actually, frequently requested. See e.g. the links in the section "Formal linear algebra" here. Moreover, I have seen many requests for such a facility in the forums of several other CAS for years.

@H-R 

It depends on what exactly you mean by "each kind of factorization". May be that a single procedure (or a few of them) may deal with all those transformations that you have in mind.

About the first question, I have the previous question on whether you want to have that very lengthy term shown explicitly in the result. Or, perhaps, a symbol representing it, say, could be better. In the latter case, see e.g. ?LargeExpressions,Veil .

@H-R 

For controled factorization of symbols, I would suggest trying a parametrized transformation rule like:

> fac:=proc(p)
> A::identical(p)*B::algebraic+A::identical(p)*C::algebraic=A*(B+C):
> end proc:
> 
> f:=a*b*x+a*b*y:
> applyrule(fac(a),f);
                                 a (b x + b y)

@apm 

The question is how many transformations should be done automatically by the limit code on the expression. Trying more transformations may increase the chances of finding the limit, but at the expense of additional computational cost. Two different approaches occur in this simpler example:

> limit(sin(a)^2+cos(a)^2,a=infinity);
                                     0 .. 2
> MultiSeries:-limit(sin(a)^2+cos(a)^2,a=infinity);
                                       1

You see that, limit does not use the trigonometric identity to reduce the expression to 1 and takes quickly each summand individually, while MultiSeries:-limit looks at the expression more carefully, but takes much longer (that there are two limit commands with different problems instead of a single good one is yet another story...).

As, probably, this dichotomy cannot be resolved automatically, my stand is that the limit command of the future should include an option so that the user could request harder simplification if time is not a concern. Currently, it could be done by writing a wrapper procedure.

@Carl Love 

Actually, the form returned by limit is wrong because the value of signum(0) depends on the value of the environment variable _Envsignum0, see ?signum. Meaning that this result in terms of the single argument signum function is computationally indeterminate, missing this environment information. For instance:

f:=sqrt(-2*cos(alpha)*cos(alpha+d)+2-2*sin(alpha+d)*sin(alpha))/d:
> _Envsignum0:=1:
> limit(f, d = 0, right):
> combine(%);
                                   infinity I
> _Envsignum0:=0:
> limit(f, d = 0, right):
> combine(%);
                                   undefined

A computationally well defined result, would set this value to 0, using e.g. the three argument form signum(0,...,0).

@Joe Riel 

I do not see any difference in this more convoluted example of a set of lists. For n>100, the user needing to assign should use Arrays instead of lists. So, the cost would be O(m^2*n) or whatever it goes for mutable structures anyway. 

@Carl Love 

But then, in Maple, comparing two Arrays, being mutable, is also O(n). So, if one needs assigning and comparing, the O(n) cost for n>100 (where it matters most) is unavoidable anyway, isn't it?

@Mac Dude 

As I understand the question, the OP wants to run somehow the Classic GUI on a Mac, possibly using a virtual machine (rather than openning mws worksheets in the Standard GUI).

@Joe Riel 

Yes, rtable_scanblock is poorly documented. But even much more important is that a command for this task is missing in the library. A command designed with the normal user in mind, like Mathematica's Position, shown by nm. Purely computational efficiency (whether a command runs in a few tens of seconds more or less) is generically irrelevant when it takes hours to a normal user to make it work. Efficiency of the combined system (computer+user) is what matters.

@Axel Vogt 

There are several situations where a web version of the help documentation is a better or unique alternative. For instance, if I post here about some statement in ?solve,details , it becomes much more convenient to have it as a hyperlink to the web version of that help page, so that the reader just clicks and opens it, than have to copy it, open Maple, and paste in the help browser (or at the worksheet prompt) and execute. Note, by the way, that for some time they were automatically converted to hyperlinks, a facility later broken in one of the recent "improvements".

Then, a search for some subjects about Maple on the web may provide additional information available in diverse sites, including this one, where web help pages add to the pool of information. Besides, web search engines are more sophisticated than the help browser search engine. And indeed, web help pages is a formidable marketing tool (if well done), as they can be seen by every potential user without access to the system. In particular, several CAS have them (Mathematica, Maxima, etc). 

@Mac Dude 

The choice of 2-D input format for the web version of the help pages is a practical problem. As it is presented by means of images only, they cannot be copy&pasted, as complained many times, now again by nm (previously here),  And as these images are frequently of low quality, their content can barely understood. The same for the 2-D content posted here (the main reason why I typically ignore it). As Laurent agreed in 2011 (in the linked thread above) that this is a problem to be fixed, I can only expect a fix soon... 

First 14 15 16 17 18 19 20 Last Page 16 of 109