Let's discuss it here on the mailing list.  Probably Mateusz will have
better input about it than I will.  He will be better to answer
questions about implementation.  I could answer questions about my
potential use case.

Aaron Meurer

On Sat, Jun 11, 2011 at 2:49 AM, SherjilOzair <[email protected]> wrote:
> I meant the Frac class.
>
> On Jun 11, 1:48 pm, SherjilOzair <[email protected]> wrote:
>> Aaron, it would be wonderful if we could discuss the implementation of
>> the Frac function.
>>
>> On Jun 2, 10:37 pm, Aaron Meurer <[email protected]> wrote:
>>
>>
>>
>>
>>
>>
>>
>> > I use Matrices in the Risch algorithm.  
>> > Seehttps://github.com/asmeurer/sympy/blob/integration3/sympy/integrals/p....
>> >  The function in the test 
>> > athttps://github.com/asmeurer/sympy/blob/integration3/sympy/integrals/t...
>> > should give you an idea of a typical usage.
>>
>> > The matrices are rational functions, with possible symbolic
>> > coefficients, though the computability problems for symbolic
>> > coefficients is something we know we will have to deal with (see the
>> > comment at the top of constant_system()).  At the moment, it doesn't
>> > get very large with what is implemented, but it could when more things
>> > are implemented.  The main things that I need to do are rref(), with
>> > correctness assured with rational functions, and the ability to
>> > compute null spaces (mainly with rational entries, but I suppose they
>> > could be any symbolic entries).  This is the only part of the Risch
>> > algorithm code that uses Expr instead of Poly, since Matrix doesn't
>> > work with Poly (we would need a Frac class for that).  I don't like
>> > how I have to manually make sure rref calls cancel to assure
>> > correctness (actually, if we had Frac, I could remove a ton of calls
>> > to Poly.cancel in my code).
>>
>> > Like Mateusz pointed out, heurisch() solves a huge linear system.  The
>> > sizes he gives are a little misleading, since those are only for the
>> > integrals that run fast enough to be in the tests.  If you try to run
>> > an integral like the one from issue 1441, it hangs because of a sparse
>> > system of about 600 equations in about 450 variables (put a print
>> > statement in the code).
>>
>> > Aaron Meurer
>>
>> > On Tue, May 31, 2011 at 9:51 PM, Brian Granger <[email protected]> wrote:
>> > > Hi,
>>
>> > > In sympy.physics.quantum we use sympy Matrix instances all over the
>> > > place.  These can be quite large (100x100 up to many 1000x1000.  In
>> > > the future we could get even bigger) and always have symbolic entries.
>> > >  At times we do like to convert them to numerical numpy arrays, but in
>> > > many cases we really want the symbolic forms.
>>
>> > > On Sat, May 28, 2011 at 6:56 AM, SherjilOzair <[email protected]> 
>> > > wrote:
>> > >> I would like to know how and where Sympy's matrices are used.
>> > >> Is Sympy matrices used for numeric computing anywhere ?
>> > >> Are Sympy Matrices expected to offer any advantage that matrices in
>> > >> numpy/scipy or other libraries cannot offer ?
>>
>> > >> Is its use limited to symbolic ? What size of Matrices with symbolic
>> > >> content is used ?
>> > >> Operations on Expr are way costlier than operations on numerics. So,
>> > >> knowing the size of the symbolic matrices that are required would help
>> > >> me in optimization when writing algorithms for sparse matrices, and
>> > >> also when refactoring Matrix.
>>
>> > >> I expect that one cannot use too large symbolic matrices, as solving/
>> > >> inversing/etc. would result in expression blowup.
>>
>> > >> I would be glad if you could also tell what running time you would
>> > >> expect from the matrices that you use.
>>
>> > > instant ;)
>>
>> > > When we are dealing with large symbolic matrices, we are typically
>> > > just doing matrix/vector multplies.  But for small matrices we do
>> > > other things like linear solves, decompositions and eigenvalue
>> > > problems.  symbolic eigenvalues are great, but expressions quickly get
>> > > out of hand as the matrix size increases.
>>
>> > > Cheers,
>>
>> > > Brian
>>
>> > >> --
>> > >> You received this message because you are subscribed to the Google 
>> > >> Groups "sympy" group.
>> > >> To post to this group, send email to [email protected].
>> > >> To unsubscribe from this group, send email to 
>> > >> [email protected].
>> > >> For more options, visit this group 
>> > >> athttp://groups.google.com/group/sympy?hl=en.
>>
>> > > --
>> > > Brian E. Granger
>> > > Cal Poly State University, San Luis Obispo
>> > > [email protected] and [email protected]
>>
>> > > --
>> > > You received this message because you are subscribed to the Google 
>> > > Groups "sympy" group.
>> > > To post to this group, send email to [email protected].
>> > > To unsubscribe from this group, send email to 
>> > > [email protected].
>> > > For more options, visit this group 
>> > > athttp://groups.google.com/group/sympy?hl=en.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sympy" group.
> To post to this group, send email to [email protected].
> To unsubscribe from this group, send email to 
> [email protected].
> For more options, visit this group at 
> http://groups.google.com/group/sympy?hl=en.
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sympy" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/sympy?hl=en.

Reply via email to