• fyi, Rubi in Sympy added to CAS independent integration tests. Init

    From nma%12000.org@gtempaccount.com@21:1/5 to nma%12...@gtempaccount.com on Wed Dec 23 17:58:15 2020

    Just installed sympy 1.7.1 to try to see if now rubi is working again or not.

    It is still not working:

    (base) >python
    Python 3.7.9 (default, Aug 31 2020, 12:42:55)
    [GCC 7.3.0] :: Anaconda, Inc. on linux
    Type "help", "copyright", "credits" or "license" for more information.
    import sympy
    Use exit() or Ctrl-D (i.e. EOF) to exit
    (base) >python
    Python 3.7.9 (default, Aug 31 2020, 12:42:55)
    [GCC 7.3.0] :: Anaconda, Inc. on linux
    Type "help", "copyright", "credits" or "license" for more information.
    from sympy import *
    from sympy.integrals.rubi.rubimain import rubi_integrate
    x = symbols('x')
    rubi_integrate(x, x)
    Integral(0, x)


    It seems the port of rubi to Sympy have been abandoned. Too bad. It was working in version Sympy 1.4 but due to some bug it stopped working after that, and no one has fixed this bug ever since.


    On Sunday, November 25, 2018 at 5:56:11 AM UTC-6, nma%12...@gtempaccount.com wrote:

    Added Rubi in Sympy to the CAS integration tests. Used Sympy 1.3.


    This is an initial report, Rubi on Sympy was run on the first 12 files of Albert Rich's test suite.

    The number of integrals is 1,867. This report only added Rubi in Sympy and combined earlier results for the other CAS systems since the same input files were used for all tests.

    In the future will try to run the full test suite again.

    The following is the % passed integrals ========================================
    1. 98.82 %, Rubi 4.15.2 in Mathematica 11.3
    2. 98.55 %, Mathematica 11.3
    3. 93.09 %, Maple 2018.1
    4. 92.61 %, Fricas 1.3.4
    5. 83.29 %, XCAS/Giac 1.4.9
    6. 82.22 %, Maxima 5.41 Using Lisp ECL 16.1.2
    7. 78.95 %, Rubi in Sympy 1.3
    8. 69.42 %, Sympy 1.1.1

    So Rubi in Sympy did better than Sympy integrate. If they fix the few bugs still there, it should do much better than the above. In theory, it should give same result as Rubi itself I would think.

    But it was a little slower than Sympy integrate. It had 5.4 second mean time for each integral while Sympy integrate had 2.91 seconds.

    Another issue is the initial long load time to read the rules into Python.

    from sympy.integrals.rubi.rubi import rubi_integrate

    It takes about 15-20 minutes (depending on the PC) for the above to complete.

    So each time one starts a new Python session, they have to wait this time to use the command.

    This time should be reduced, otherwise some might not bother waiting this long to use it.

    ps I entered a bug on Sympy tracking system for Rubi-Sympy not able to integrate x. This should get fixed by Sympy 1.4 hopefully.


    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)