py-libseek-thermal-master

##Description

NOTE: Work in progress.

This project provides a python interface wrapper to the libseek-thermal c++ user space driver with a python interface.

Many different options exist for doing this, many poorly documented and/or have complicated custom dependencies/build systems that make them less palatable:

1. Extending python manually:
   • Advantages: Simple
   • Disadvantages: It gets tricky when trying to export/deal with more complex datastructs, and technically I was rather confused with the documentation on what the true goal of this all was (they talk about rebuilding Python, and technically I don't want to rebuild an extension of python -- I want to wrapper a c++ library with a python module.
2. SIP:
   • Advantages: Supposedly seemless integration/wrappering of c++. A part of the Qt package interface to provide the PyQt wrappers.
   • Disadvantages: Ugly .sip interface files, confusing syntax for more complex situations, and failed to seemlessly build module when I attempted to include the libseek headers (couldn't find uint16_t, despite the fact that I gave the correct include reference in my .sip file), and I didn't want to try to figure out why it was failing to build.
3. ctypes
   • Advantages: Very simple, native interface in python to ctypes.
   • Disadvantages: Technically made as a c-interface to python, not c++, so I'd have to wrapper C around C++ to wrap python. Moreover, dealing with complex opencv Mat class types was something I didn't want to spend a lot of time trying to reverse engineer to feed into python.
4. boost.python
   • Advantages: One of the simpler syntax/easier wrapper frameworks for c++ to python. Supposedly handles c++ classes, reference counts, etc. seemlessly.
   • Disadvantages: Uses the bjam build system, a boost-specific build system that didn't work out of the box for their examples, and their documentation was convoluted enough that I didn't want to spend hours trying to debug & solve it.
5. CLIF - Wrapper generator c++ for several languages using LLVM
   • Link
   • Advantages: Once you learn, supposedly scales to many different languages.
   • Disadvantages: Uses a compiler I'm not familiar with, I don't want to have to learn about for this simple task, and not sure of support on Macs. It also has a large number of annoying dependencies.

In the end, this implementation uses libseedoost.python, as it seems to be one of the better wrapper frameworks for python -> C++. Given that the bjam build framework was avoided, this interface worked great.

##Setup This python SeekThermal wrapper is currently built and tested on a Raspberry Pi Raspian OS. It currently only supports the SeekThermal Compact camera.

###Dependencies

*libboost *libseek-thermal-master *python (>=2.7) *libopencv *boost.python

###Building and Installing

####Macports presetup

• I had some linker issues that may be tied to the python version that MacPorts boost was built under (+python27), given that I'm using python 3.6. I uninstalled and reinstalled as follows:

>sudo port install boost +python36

• Rudimentary compilation attempt:

sudo clang++ -v -shared -L/opt/local/lib -L/opt/local/Library/Frameworks/Python.framework/Versions/3.6/lib -lpython3.6m -
I/opt/local/include -I "/opt/local/Library/Frameworks/Python.framework/Versions/3.6/include/python3.6m" hello.cpp -lboost_python3-mt -fpic -o libhello.so

This appears to have generated libhello.so without a hitch, but now I may want to expand to cmake (as non-bjam people seem to use this build framework) to build what I need (I have the library now, but I obviously need to generate other files to correctly have the python script import.

After running cmake, I realized that I don't need to use cmake to do this (but I will anyways, as libseek-thermal already uses cmake for it's build system). In fact, the output *.so file is what is imported by the 'import' command in python.

####CMake build and Install

#mkdir build
#cd build
#cmake ../
#make

Copy the pylibseek_ext.so to the appropriate location and import as in python.

###TODO

*Fix CMakeLists.txt to be more universal. *Allow to be installed as system-wide python package for import (currently only a local package). *Include example.