X.flat returns an indexable 1-D iterator (mostly similar to an array but always 1-d)

long(<>) --> pylong(<>) if from Numeric * was changed to from scipy.base import *

.typecode() --> .dtypechar

.iscontiguous() --> .flags['CONTIGUOUS']

.byteswapped() -> .byteswap()

.itemsize() -> .itemsize

If you used typecode characters:

'c' -> 'S1'

'b' -> 'B'

'1' -> 'b'

's' -> 'h'

'w' -> 'H'

'u' -> 'I'

C -level

some API calls that used to take PyObject * now take PyArrayObject * (this should only cause warnings during compile and not actual problems).

PyArray_Take

These commands now return a buffer that must be freed once it is used

using PyMemData_FREE(ptr) or PyMemData_XFREE(ptr);

a->descr->zero --> PyArray_Zero(a)

a->descr->one --> PyArray_One(a)

Numeric/arrayobject.h --> scipy/arrayobject.h

# These will actually work and are defines for PyArray_BYTE,

# but you really should change it in your code

PyArray_CHAR --> PyArray_BYTE (or PyArray_STRING which is more flexible)

PyArray_SBYTE --> PyArray_BYTE

Any uses of character codes will need adjusting....

use PyArray_XXXLTR where XXX is the name of the type.

If you used function pointers directly (why did you do that?),

the arguments have changed.

a->descr->cast[i](fromdata, fromstep, todata, tostep, n)

a->descr->cast[i](fromdata, todata, n, PyArrayObject *in, PyArrayObject *out)