Static code analysis to get source of objects used in functions/methods

Issue #18 resolved
Pierre Augier created an issue

Such analysis of the code of the functions/methods would be very useful!

  • It would allow us to solve many issues, namely #1, #6, #7, #10.

  • It would help a lot for issues #3 and #8 (pythran_def in classes).

  • It would allow us to make @cachedjit much nicer (without # pythran import and @used_by_cachedjit).

But I think this one is too difficult for me (@paugier)...

Note: it has to be done without Pythran because we don't want to depend on Pythran (even if yes, it is to produce Pythran files).

Comments (3)

  1. Ashwin Vishnu

    An easy way to achieve this is to "vendor" Pythran's python modules within fluidpythran. This is common practice for package managers like "pip"

  2. Pierre Augier reporter

    Update after FluidPythran 0.1.5:

    Such analysis would be useful to

    • no need for # pythran import (issue #6)
    • no need for @include (for boost) and @include(used_by="func") (for cachedjit) (issue #7)
    • use global variables in pythranized functions (issue #30)
    • nicer syntax for blocks (see issue #29 and #1)
    • detect unsupported assignment self.attr = 1
  3. Log in to comment