Using SWIG
Numpy
Homework
10. Using C++ in Python
Ján Dugáček
November 21, 2018
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Table of Contents
1 Using SWIG
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
2 Numpy
A note about C arrays
A function
SWIG file
Result
Exercise
More info
3 Homework
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Motivation
You will run into a lot of code written in Python in your life
You might need to integrate C++ into it because cycles are
not exactly fast in Python
Using C++ is the only way to add new builtin types into
Python
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Some class
#i f n d e f PSEUDOFUNC_H
#d e f i n e PSEUDOFUNC_H
#i n c l u d e <map>
c l a s s pseudofunc {
std : : map<f l o a t , f l o a t > data ;
p u b l i c :
void add ( f l o a t x , f l o a t fx ) ;
f l o a t at ( f l o a t x ) ;
};
#e n d i f
The parts with #ifndef and others are preventing the file
from being included more than once, all you have to do is to
give it a unique name
Name is supposed to be pseudofunc.h
Declarations of methods will be in another file
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Some class #2
#i n c l u d e " pseudofunc . h"
void pseudofunc : : add ( f l o a t x , f l o a t fx ) {
data [ x ] = fx ;
}
f l o a t pseudofunc : : at ( f l o a t x ) {
i f ( data . empty ( ) ) r e t u r n 0;
i f ( x<data . begin()−> f i r s t ) r e t u r n data . begin()−>second ;
i f ( x>data . r b e g i n ()−> f i r s t ) r e t u r n data . r b e g i n ()−>second ;
auto more = data . upper_bound ( x ) ;
auto l e s s = std : : prev ( more ) ;
f l o a t part =(x − l e s s −>f i r s t )/( more−>f i r s t − l e s s −>f i r s t ) ;
r e t u r n part ∗ more−>second + (1 − part ) ∗ l e s s −>second ;
}
Now we define the methods, in a file that should be named
pseudofunc.cpp
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Configuration file for SWIG
%module pseudofunc
%{
#i n c l u d e " pseudofunc . h"
%}
%i n c l u d e " pseudofunc . h"
It tells SWIG how to compile the file, it allows a lot of tricks
Make sure you don’t mistake the % for the #
If using some other python-bound types, %include them as
std_string.i
These slides assume this file is named pseudofunc.i
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Using SWIG
swig −c++ −python pseudofunc . i
You need to have SWIG installed to do this in the command
line
You have to specify the code is in C++ (and not in C) and
that you want to generate python code (and not for other
programming languages)
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Using distutils to have Python compile it for itself
#!/ usr / bin /env python3
from d i s t u t i l s . core import setup , Extension
pseudofunc_module = Extension ( ’ _pseudofunc ’ ,
s o u r c e s =[ ’ pseudofunc_wrap . cxx ’ , ’ pseudofunc . cpp ’ ] , )
setup (name = ’ pseudofunc ’ , v e r s i o n = ’ 1 ’ , author = "Dugi" ,
d e s c r i p t i o n = """A pseudofunction """ ,
ext_modules = [ pseudofunc_module ] ,
py_modules = [ " pseudofunc " ] , )
If it’s called makePseudofunc.py, call
python3 makePseudofunc.py build in command line
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Usage
>>> import pseudofunc
>>> p = pseudofunc . pseudofunc ()
>>> p . add (3 ,15)
>>> p . add (4 ,20)
>>> p . add (5 ,30)
>>> p . at ( 4 . 1 )
21.0
If you have the compiled file and the generated python file
available (for example in the same folder), you can import it
and use the class in Python
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
Exercise
1 Use C++ to create a mumpy Python module that will tell a
random yer mum joke (out of at least 3)
2 Use C++ to create a simulacrum of std::vector<double>
in Pyton (or easy::vector<double>
3 Use C++ to create some sort of 2D vector in Python
4 Use C++ to create a function in Python that computes the
derivative of data in a form that fits you
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Motivation
Some class
Configuration file for SWIG
Using SWIG
Using distutils to have Python compile it for itself
Usage
Exercise
More info
More info
A short guide to using SWIG is at
https://www.cs.ubc.ca/~gberseth/blog/
using-swig-to-wrap-c-for-python.html
A very detailed description of this functionality is at
http://www.swig.org/Doc1.3/Python.html
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
A note about C arrays
void arrayDemoC ( i n t ∗ a r r a y ) {
i n t a = ∗ a r r a y ;
i n t b = a r r a y [ 0 ] ;
i n t c = a r r a y [ 1 ] ;
}
Arrays in C are actually naked pointers to the first element in
an array
Naked pointers, the only kind of pointers in C, are thus used in
the same way as arrays
The array’s size must be in another variable
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
A function
#i n c l u d e " unnormalise . h"
void unnormalise ( double ∗ in , i n t inS , double ∗ out , i n t outS )
{
i f ( inS == 0 | | inS != outS ) r e t u r n ;
double min = i n [ 0 ] ;
f o r ( i n t i = 0; i < inS ; i++)
i f ( i n [ i ] < min ) min = i n [ i ] ;
min = 1 / min ;
f o r ( i n t i = 0; i < inS ; i++)
out [ i ] = i n [ i ] ∗ min ;
}
Numpy arrays are accessed as 2 variables, data and size
Can you guess what this does?
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
A function #2
#i f n d e f UNNORMALISE_H
#d e f i n e UUNORMALISE_H
void unnormalise ( double ∗ in , i n t inS , double ∗ out , i n t outS ) ;
#e n d i f
We make it a header
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
SWIG file
%module unnormalise
%{
#d e f i n e SWIG_FILE_WITH_INIT
#i n c l u d e " unnormalise . h"
%}
%i n c l u d e "numpy . i "
%i n i t %{
import_array ( ) ;
%}
%apply ( double ∗ IN_ARRAY1, i n t DIM1) {( double ∗ in , i n t inS )}
%apply ( double ∗ INPLACE_ARRAY1, i n t DIM1) {( double ∗ out , i n t
%i n c l u d e " unnormalise . h"
We need to enable the swig initialisation function import_array()
with SWIG_FILE_WITH_INIT
We need to map the input types to the argument names we are
using
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
Result
>>> import unnormalise
>>> r e s u l t = numpy . empty_like ( x )
>>> unnormalise . unnormalise ( x , y )
It’s compiled like before
It’s not possible to return an array (numpy’s SWIG interface is
not great)
It’s faster without returning, as new arrays aren’t created
You can add a Python function that creates the result array
and returns it
To bind more values, use:
%apply ( double ∗ IN_ARRAY1, i n t DIM1) {( double ∗ in1 ,
i n t inS1 ) , ( double ∗ in2 , i n t inS2 )}
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
Exercise
1 Write a C++ function that can find the median of a numpy
array
2 Write a C++ function that can find the highest common
denominator of all numbers in a numpy array
3 Write a C++ function that interpolates a two-variable function
given by a 2D numpy array (assuming the indexes are x and y)
4 Write a C++ function that computes the divergence of a
function given by a 3D numpy array
5 Write a C++ function that computes the curl of a function
given by a 3D numpy array
%apply ( double ∗ IN_ARRAY2, i n t DIM1 , i n t DIM2)
{( double ∗ in , i n t in1 , i n t in2 )}
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
Exercise for those who don’t know numpy
1 Use C++ to add a function to Python that will sort a vector
so that the highest number is the first
2 Use C++ to add a function to Python that prints a smiley of a
given size made of spaces and some letter
3 Use C++ to create a Python class vector3D that supports
addition and vector product
4 Use C++ to create a Python function to compute the mean
root-mean-square deviation of given data
5 Use C++ to create a Python function to compute the
covariance of two sets of given data
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
A note about C arrays
A function
SWIG file
Result
Exercise
More info
More info
A description of using numpy with C++ is at
https://www.scipy-lectures.org/advanced/
interfacing_with_c/interfacing_with_c.html#swig
In-depth information can be found here
https://docs.scipy.org/doc/numpy/reference/swig.
interface-file.html
Ján Dugáček 10. Using C++ in Python
Using SWIG
Numpy
Homework
Homework
Use C++ to create a Python class that represents analytical
functions composed of addition, subtraction, multiplication,
division, power and logarithm
You have two weeks to do it
It’s recommended to compose it in a tree structure, starting
with function f (x) = 1 and function f (x) = x and using
operators to compose it, then operator() to call it
For the ambitious: Give it a method to compute its derivative
(still in analytical form)
For lunatics: Give it a method to compute its primitive
function (still in analytical form)
Ján Dugáček 10. Using C++ in Python