Bio.Cluster not checking arguments under Python 3

[peterjc]

I found this while investigating what is probably a big-endian issue in the C code which checks the single-letter arguments (not fixed, see mailing list discussion):

> On powerpc and s390x:
>
> | ======================================================================
> | ERROR: test_clusterdistance (test_Cluster.TestCluster)
> | ----------------------------------------------------------------------
> | Traceback (most recent call last):
> |   File "/«BUILDDIR»/python-biopython-1.64+dfsg/.pybuild/pythonX.Y_3.4/build/Tests/test_Cluster.py", line 212, in test_clusterdistance
> |     method='a', transpose=0)
> | ValueError: method should be a single character
> |
> | ======================================================================
> | ERROR: test_kcluster (test_Cluster.TestCluster)
> | ----------------------------------------------------------------------
> | Traceback (most recent call last):
> |   File "/«BUILDDIR»/python-biopython-1.64+dfsg/.pybuild/pythonX.Y_3.4/build/Tests/test_Cluster.py", line 141, in test_kcluster
> |     method='a', dist='e')
> | ValueError: method should be a single character
> |
> | ======================================================================
> | ERROR: test_somcluster (test_Cluster.TestCluster)
> | ----------------------------------------------------------------------
> | Traceback (most recent call last):
> |   File "/«BUILDDIR»/python-biopython-1.64+dfsg/.pybuild/pythonX.Y_3.4/build/Tests/test_Cluster.py", line 557, in test_somcluster
> |     inittau=0.02, niter=100, dist='e')
> | ValueError: distance should be a single character
> |
> | ======================================================================
> | ERROR: test_treecluster (test_Cluster.TestCluster)
> | ----------------------------------------------------------------------
> | Traceback (most recent call last):
> |   File "/«BUILDDIR»/python-biopython-1.64+dfsg/.pybuild/pythonX.Y_3.4/build/Tests/test_Cluster.py", line 290, in test_treecluster
> |     transpose=0, method='a', dist='e')
> | ValueError: method should be a single character
> |
> | ----------------------------------------------------------------------
> | Ran 210 tests in 337.138 seconds
> |
> | FAILED (failures = 1)
> | Skipping any tests requiring internet access
> | Python version: 3.4.1 (default, Jun  7 2014, 17:37:50)

Relevant C code:

 if (PyUnicode_Check(object))
     data = PyUnicode_AS_DATA(object);
 # ...
 if (strlen(data)!=1)
 { PyErr_Format(PyExc_ValueError, "method should be a single character (not '%s')", data);
   return 0;
 }

Jakob worte: But PyUnicode_AS_DATA() returns a pointer to UTF-32 data. On big-endian architectures (such as powerpc or s390x), the first byte is always 0, so strlen() returns 0.

[peterjc]

Is the new unit test helpful though? i.e. 5b99854

[peterjc]

@mdehoon How should the C code be updated?

[mdehoon]

I think this will not be super complicated. Just somebody needs to figure out how to get the length of a unicode string in C. Unfortunately I don't have the time now to go after this.

[peterjc]

@mdehoon How does this look (on a test branch): peterjc@6741204

The (new) test passes: https://travis-ci.org/peterjc/biopython/builds/32624110

[peterjc]

Fixed by 2a140f4

Tests added in c7811df

Thanks all :)