a bug or a feature.
11-12-2015, 03:05 PM
Post: #1
 ji3m Member Posts: 149 Joined: Sep 2015
a bug or a feature.
I am having great difficulty doing cas matrix operations in a program.

The test case below shows one of the problems

EXPORT hsbat2(M)
BEGIN
LOCAL I,P,Q;
DEBUG;
I := inv(M); <- the inverse seems ok. Both M and I are lists
P := I*M; <- a problem here since '*' does silly list multiply
Q := simplify(P); <- since P is wrong this doesnt do much.
{M,I,P,Q}
END;

From the cas console by hand the matrices are sugar coated with
[[..]] and the parser knows that they are matrices (hidden type) so
it does the right thing.

Not so in a program where '*' sees things as lists so it multiplies element
by element, the wrong answer for matrices and vectors.

If there is a multiplyMat(M1,M2) like function, i cant find it.

Apparenty most of the cas functions expecting vectors or matrices treat
lists appropriatly. Not so for lexical operators, i guess.

Another problem i have is determing the return type when matrices or vectors
are involve since everything is a list.

Do i need to write my own matrix /vector multiply routines for program use[/u]?

I work with android Prime only for now on my
Galaxy Tab Pro/Note 2
11-12-2015, 03:14 PM
Post: #2
 Marcus von Cube Senior Member Posts: 760 Joined: Dec 2013
RE: a bug or a feature.
(11-12-2015 03:05 PM)ji3m Wrote:  EXPORT hsbat2(M)
BEGIN
LOCAL I,P,Q;
...
{M,I,P,Q}
END;

Have you tried to create a #CAS program using the pragmas (#CAS, #END if I'm right)? Your program is a Home program which does a lot of unnecessary conversions between the environments for each call to a CAS function. These conversions may lose information such as the CAS subtype of a list.

Marcus von Cube
Wehrheim, Germany
http://www.mvcsys.de
http://wp34s.sf.net
http://mvcsys.de/doc/basic-compare.html
11-12-2015, 03:29 PM (This post was last modified: 11-12-2015 03:31 PM by Han.)
Post: #3
 Han Senior Member Posts: 1,817 Joined: Dec 2013
RE: a bug or a feature.
(11-12-2015 03:05 PM)ji3m Wrote:  I am having great difficulty doing cas matrix operations in a program.

The test case below shows one of the problems

EXPORT hsbat2(M)
BEGIN
LOCAL I,P,Q;
DEBUG;
I := inv(M); <- the inverse seems ok. Both M and I are lists
P := I*M; <- a problem here since '*' does silly list multiply
Q := simplify(P); <- since P is wrong this doesnt do much.
{M,I,P,Q}
END;

How are you calling the program? It runs just fine to me.
M5:=[[1,2,3],[−1,0,1],[0,0,1]];
hsbat2(M5) returns
Code:
{   [[1,2,3],    [−1,0,1],    [0,0,1]],   [[0,−1,1],    [0.5,0.5,−2],    [0,0,1]],   [[1,0,0],    [0,1,0],    [0,0,1]],   [[1,0,0],    [0,1,0],    [0,0,1]] }

Why are you using lists? Can you show us your example that fails?

Graph 3D | QPI | SolveSys
11-12-2015, 03:45 PM
Post: #4
 ji3m Member Posts: 149 Joined: Sep 2015
RE: a bug or a feature.
Sure enough, converted test case to #cas and it produced the very correct results.

So i can make a matmpy(m1,m2) in #cas.

Ive been trying to avoid #cas pgms since they are deleted when a restart is done.
The purge () function doesnt let a program purge selected vars only.(no way to get a list of vars from a program env)

BTW, the general help menu shows operators like ",*" , ",+" to do element by element
operations on lists, etc. which implied that the normal "*" should not do this. But alas, it doesnt know the subtypes as you point out.

I work with android Prime only for now on my
Galaxy Tab Pro/Note 2
11-12-2015, 03:51 PM
Post: #5
 ji3m Member Posts: 149 Joined: Sep 2015
RE: a bug or a feature.
Han:

Everything is fine with numeric matrices. They even have their own TYPE.

Symbolic (cas) Matrices and vectors are all typed as lists so the home-mode program language is lost in space when symbolics cross into its domain.

I work with android Prime only for now on my
Galaxy Tab Pro/Note 2
11-12-2015, 04:16 PM
Post: #6
 Han Senior Member Posts: 1,817 Joined: Dec 2013
RE: a bug or a feature.
(11-12-2015 03:51 PM)ji3m Wrote:  Han:

Everything is fine with numeric matrices. They even have their own TYPE.

Symbolic (cas) Matrices and vectors are all typed as lists so the home-mode program language is lost in space when symbolics cross into its domain.

I see now. Yes, unfortunately a #cas program is necessary to avoid these cross-parsings of inputs (and outputs, too).

Graph 3D | QPI | SolveSys
 « Next Oldest | Next Newest »

User(s) browsing this thread: 1 Guest(s)