PyQt4 with Cython = Bad Idea?

Question

Someone at my job wrote a few python modules using PyQt4. I wrote a Jython application that was supposed to allow Python modules access to the Java binding of our API but I did not realize Jython couldn't use ctypes thus the entire application is basically useless.

I'm scrambling to find a solution on how I can expose our API to his Python modules. Our API is C and has a C++ binding (as well as Java and Perl) so I'm thinking the best solution is to just expose one of the C/C++ APIs to his Python modules.

Instead of doing (what I anticipate would be a lot of work) a C/C++ wrapper for Python using ctypes, sig, swig, etc. I thought maybe I could just have him use Cython. From what I've been reading you don't have to do much work to make your Python code into Cython so I could create a few test examples for him demonstrating how to convert his Python modules into Cython and show him how you would access our C API.

But is it even possible to use Cython and PyQt4 at the same time. It doesn't seem to make much sense, Qt is C or C++ and PyQt is a binding or wrapper to use Qt with Python then we're changing the Python code into C code using Cython. But despite it not making sense and possibly slowing everything down is it even possible? If it's possible I'm going to offer this as a temporary solution.

Answer 1

I did this as a quick and dirty solution to a problem, passing a QImage to the following function:

cpdef void export_array(object image, int[:] colorList, int[:] dim, int[:, :] arr, int color_offset):
    """
    Updates the image with the values from an entire array.

    :param image:           A QImage object to be written to
    :param colorList:       List of colors, as integers
    :param dim:             Dimensions of the array
    :param array:           Array to use as new image
    :param color_offset:    Use primary colors (0) or secondary colors (2)?
    :return:                None
    """
    cdef int color
    cdef Py_ssize_t i, j
    cdef int a = dim[0]
    cdef int b = dim[1]
    for i in range(a):
        for j in range(b):
            color = colorList[arr[i][j] + color_offset]
            image.setPixel(i, j, color)

It's an order of magnitude faster:

%timeit cf.export_array(image, colorList, imageDim, arr, 2)
%timeit pf.export_array(image, colorList, arr, 2)
507 ms ± 730 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)
3.38 s ± 22.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

Though another function that I tried it with slowed down by about 10%. In this case the nested for loops make it a good candidate.

(I assume your problems have evolved considerably since you posted this question, but maybe this answer can help other contemporary googlers.)