Am 08.03.2015 um 07:18 schrieb AMiT Kumar:
I think, building a robust framework, which felicitates further
development, worth much more than adding new solvers.
+1
I can't comment much on how such a framework would need to look like
because I know too little about what solving methods exist and what
kinds of services they'd need, but given the sheer number of solvers
it's clear that making a framework will pay off.
solveset already is such a framework.
Building a subframework on top of solveset, specialized for the needs of
ODE solving, would certainly be helpful.
Note that to build a framework (or subframework), experience with or
knowledge about existing solvers will be very important; it's the guide
to selecting what kinds of services you want (and do not want) in the
framework. It's not a good idea to add a service "just in case", because
(a) it requires that coder for new solvers will have to look at it and
decide whether it's useful to them or not, (b) each additional feature
adds design constraints.
So without knowledge about solving methods, there is a very real danger
that you end up coding features that nobody needs and that make it
harder to code or use the features that are really needed.
Actually this happens a lot even with people with lots of experience in
the field. It is the main reason why the first incarnation of most
frameworks is stillborn (but still essential because it is where
experience is collected).
IOW if you do a framework project, it will be immensely more worth it if
you can do it for a few years and help it mature.
--
You received this message because you are subscribed to the Google Groups
"sympy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sympy.
To view this discussion on the web visit
https://groups.google.com/d/msgid/sympy/54FC0F81.3010303%40durchholz.org.
For more options, visit https://groups.google.com/d/optout.
