[pygsl-discuss] building PyGSL-0.3.2 with NumPy-0.9.6

[Krishna M. G. <gk...@gm...>]

Hello,

Is there any easy way to build PyGSL with NumPy. I noticed that PyGSL
requires either Numeric or numarray and generates appropriate code.
But NumPy is different from Numeric. Is it compatible with NumPy at
all?

I faked the presence of Numeric library and used convertcode.py that
ships with NumPy on the PyGSL installation. I also changed
"_numobj.py" "_mlab.py" appropriately. That did the trick for imports.

But looks like thats not enough. When I tried to test the library I
got the following errors. Is this because the test file has bugs or my
installation and further modifications are not enough?

[krishna@master local]$ python inst/pygsl-0.3.2/tests/blas_test.py

.......................EE.....EEE.EEE......E.E
======================================================================
ERROR: testZaxpy (__main__.BlasTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "inst/pygsl-0.3.2/tests/blas_test.py", line 70, in testZaxpy
    3)
  File "/home/krishna/local/inst/pygsl-0.3.2/tests/gsl_test.py", line
35, in arrayCompare
    result = result and fpcompare(a[i], l[i], digits)
  File "/home/krishna/local/inst/pygsl-0.3.2/tests/gsl_test.py", line
24, in fpcompare
    res1 = (fcmp(a, b.real, eps) == 0)
TypeError: argument number 1: a 'double' is expected,
'complex128scalar((-120.5+17.29j))' is received

======================================================================
ERROR: testZaxpy_cr (__main__.BlasTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "inst/pygsl-0.3.2/tests/blas_test.py", line 78, in testZaxpy_cr
    3)
  File "/home/krishna/local/inst/pygsl-0.3.2/tests/gsl_test.py", line
35, in arrayCompare
    result = result and fpcompare(a[i], l[i], digits)
  File "/home/krishna/local/inst/pygsl-0.3.2/tests/gsl_test.py", line
24, in fpcompare
    res1 = (fcmp(a, b.real, eps) == 0)
TypeError: argument number 1: a 'double' is expected,
'complex128scalar((-120.5+17.29j))' is received

.
.
.

----------------------------------------------------------------------
Ran 46 tests in 0.242s

FAILED (errors=10)