I'm editing some Python code containing the line:
arr = (c_double * len(self.static_mapped[0, :]))(*self.static_mapped[0, :])
where field self.static_mapped is created using numpy.zeros and where c_double is imported from ctypes.
Based on the context, I'm assuming this is typecasting from numpy doubles to C-type doubles, but this syntax doesn't appear in any of the Python guides I've consulted, and even my IDE is complaining that "'int' object is not callable".
Am I correct in my interpretation of its meaning? What sequence of function calls is going on under the hood? Is this a well-known code idiom in Python or is there a more standard/straightforward way of accomplishing this?
Strictly from a syntactic perspective, your code should be equivalent to
t = c_double * len(self.static_mapped[0, :])
arr = t(*self.static_mapped[0, :])
This implies that c_double.__mul__ is overridden to produce some other type object that gets called to produce an array. No type-casting, just dynamic type generation.
Specifically, multiplying type by an integer n creates an array of n values:
>>> c_double * 2
<class '__main__.c_double_Array_2'>
>>> c_double * 3
<class '__main__.c_double_Array_3'>
>>> c_double * 7
<class '__main__.c_double_Array_7'>
Note that "type multiplication" is not associative:
# An array of arrays
>>> c_double * 7 * 2
<class '__main__.c_double_Array_7_Array_2'>
# A single array
>>> c_double * (7 * 2)
<class '__main__.c_double_Array_14'>