Hi
I recommend using profiling tools such as pyinstrument and snakeviz to see
if there are any obvious ways we can improve the performance.
Aaron Meurer
On Sun, Jul 23, 2017 at 11:08 AM Arihant Parsoya
wrote:
> The slowness of tests is primarily due to:
>
> 1. We need to simplify, expand or di
The slowness of tests is primarily due to:
1. We need to simplify, expand or differentiate the integrated expression
in order to equate them. I used numerical evaluation at random points for
few tests and it gave faster results. I haven't tried using numerical
evaluation for the whole tests yet.
That's pretty bad. Is it because of the number of tests or the performance
of the integrator? If the performance is bad, that's something that should
be looked into.
Aaron Meurer
On Sat, Jul 22, 2017 at 7:40 AM, Arihant Parsoya
wrote:
> Rubi test suit take hours to complete. Are we going to add
Rubi test suit take hours to complete. Are we going to add the test suit in
sympy code after the rubi module is complete? If we do, then all the travis
test will take longer time to pass.
On Sunday, July 16, 2017 at 8:07:21 PM UTC+5:30, Aaron Meurer wrote:
>
> That will be nice if Rubi updates i
That will be nice if Rubi updates its rules. But the other use-case I
was thinking of is if a SymPy user has an integral that SymPy cannot
do, but they know how to integrate it, they can add a rule to the
integrator. In that case, they will most likely want to write the rule
in some Python/SymPy sy
Hi,
I think we should also include parser which converts rules from Mathematica
Downvalues to MatchPy. If we want to add new rules, we can write it in
Mathematica, and parse them using the parser.
Arihant
On Sunday, July 16, 2017 at 5:55:14 PM UTC+5:30, Ondřej Čertík wrote:
>
> On Sat, Jul 15
On Sat, Jul 15, 2017 at 11:29 AM, Francesco Bonazzi
wrote:
>
>
> On Saturday, 15 July 2017 10:29:23 UTC-4, Ondřej Čertík wrote:
>>
>> Hi Arihant,
>>
>> On Thu, Jul 13, 2017 at 10:52 AM, Arihant Parsoya
>>
>> If so, Then one option is to get your setup into Travis for Python 2,
>> and that way you
OK, shall we add some new Travis tests for this module only? How would you do
that?
--
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 sympy+unsubscr...@googlegroups.com.
I would be fine with a 3.6-only dependency just for adding new rules.
That is, if we can commit a pre-compiled decision tree file that works
on 2.7-3.6 and doesn't depend on MatchPy. Then MatchPy is only needed
if the rules get updated, or if someone wants to add a custom rule at
runtime.
Aaron Me
On Saturday, 15 July 2017 10:29:23 UTC-4, Ondřej Čertík wrote:
>
> Hi Arihant,
>
> On Thu, Jul 13, 2017 at 10:52 AM, Arihant Parsoya
>
> If so, Then one option is to get your setup into Travis for Python 2,
> and that way you have something that passes all SymPy tests in Python
> 2 and Python
Hi Arihant,
On Thu, Jul 13, 2017 at 10:52 AM, Arihant Parsoya
wrote:
> Hi Aaron and Francesco,
>
> I converted most code of MatchPy to python 2. However, MatchPy has
> dependency on few other python3.6 libraries(such as typing). Backport
> version of those libraries is available online. I found t
Hi Aaron and Francesco,
I converted most code of MatchPy to python 2. However, MatchPy has
dependency on few other python3.6 libraries(such as typing). Backport
version of those libraries is available online. I found that 1-2 matchpy
test cases were failing in python2.7 but all were running in
I think that the current priority is to make the rules work. RUBI has a lot of
utility functions that need to be translated.
So far, it looks like the rules that have been ported are much slower than on
Mathematica, but this is to be expected, as Mathematica is better optimized
than Python.
--
That's great. I asked Manuel if he had heard of 3to2 and he hadn't.
That's great if it already works out of the box. I think there is
also pasteurize
http://python-future.org/automatic_conversion.html#pasteurize-py3-to-py2-3,
which is probably a better option if the goal is long term support.
He
On Thursday, 13 July 2017 11:36:47 UTC-4, Francesco Bonazzi wrote:
>
> As a temporary workaround, we are generating
>
As a temporary workaround, we are generating the power set of all optional
wildcards in a single rule, replacing them with the default matching value
each time.
--
You rece
Hi Aaron,
Arihant was able to use the 3to2 tool to port MatchPy to Python 2.7, and it
looked like it was working. We are actually waiting for some extensions to
MatchPy (support for optional wildcards), as RUBI uses them extensively. As
a temporary workaround, we are generating
I also believe
Hi.
Manuel Krebber is here at SciPy. I've talked with him a little bit
already, and will likely talk more on Saturday and Sunday during the
sprints. I'm hoping we can get something that works in all versions of
Python we support (this is important).
I discussed the possibility of a pre-compiled P
Hi,
As part of the Rubi GSoC project, the two students are importing the code from
Mathematica using the MatchPy library, which provides support for Mathematica's
associative and commutative pattern matchings.
So far the code hasn't been merged into SymPy, given the dependence on MatchPy,
and
18 matches
Mail list logo
