On Fri, Jun 5, 2009 at 9:59 AM, Ondrej Certik<ond...@certik.cz> wrote:
> Hi Toon!
>
> On Fri, Jun 5, 2009 at 8:09 AM, Toon
> Verstraelen<toon.verstrae...@ugent.be> wrote:
>> Hello,
>>
>> I'm writing a very initial version of a code generator. I would like to add
>> proper tests for the generated c code. The test will involve a compilation
>> with
>> gcc and linking against a small c program that tests the output of the
>> function
>> for several inputs. If some outputs are wrong, the negative return code can
>> be
>> captured with a proper assert statement.
>
> Excellent, thanks for looking into this, this is needed.
>
>>
>> I could not find existing tests that generate output files. What is the
>> recommended directory to write these files? I was thinking of
>> tempfile.mkdtemp,
>> but that makes it annoying to check the generated code manually. A temporary
>> test_output directory at the root level of the source tree with
>> subdirectories
>> created by the tests seems convenient to me, but this involves changes in
>> setup.py, which runs the tests. './setup.py test' should create this
>> directory
>> and './setup.py clean' should remove it enterily. Is this OK?
>
> I think we should test it using statements like:
>
> assert code_generate(<whatever>) == """\n
> int some_function(...)
> {
> return a*b-23;
> }
> """
>
> That way you don't need a compiler and it will run even if gcc is not
> installed or not working properly, or if you are not allowed to create
> any files. Remember, that all tests need to run even if you install
> sympy and then do:
>
> import sympy
> sympy.test()
>
>
> Besides that, it's a good idea to also have tests, that actually
> compile things using gcc, as you suggested. For that, I suggest to
> create a new test sympy/test_external, for example
> "test_code_generation.py", that will work something like this:
>
> try:
> <try to write a file (if it works) and find a gcc executable (I am
> not sure if we should also test if gcc works here)>
> except <anything from above fails>:
> disabled = True
>
> def test_code1():
> <generate C file>
> <compile it>
> <run it?>
> assert <all is fine>
>
> ...
>
>
> That way, if gcc is not installed, or you can't generate output files,
> the test will be skipped. It's the same philosophy as we use for
> test_numpy and test_sage.
And to answer your question, I would just create a temporary files
using tempfile.mkdtemp (and you can always leave it there if the test
fails -- suggest there is some variable to control that (maybe env
variable, see sympy/core/cache.py:315 for usage)). That's because the
sympy root directory can be in /usr/lib/python...., which might be
read only, and the current directory can also be read only.
Ondrej
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"sympy" group.
To post to this group, send email to sympy@googlegroups.com
To unsubscribe from this group, send email to sympy+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sympy?hl=en
-~----------~----~----~----~------~----~------~--~---
