jhb / graphagus Goto Github PK

Hey,

Hope you are all well !

I am currently developing a small utility to extract dependencies of some various CMake driven project and I wanted to create a global graph, so I found graphagus as a potential reference to help me understand how to create all the nodes and edges. But I had some issues that I was wondering you could help me to solve as you wrote this interesting repo. ^^

My first question is related to some errors related to my current python version, v3.6. I had some errors with cPickle and some syntax from python 2.x, I tried to fix them it but was wondering if you have already thought about it and could share some important list of points to check for this version migration. In a nutshell, how would it be complicated to make it compatible with Python 2.x and 3.x versions ?

And, the second question is much more related to my little utility. I would like to store in a persistent way a global graph of all my projects dependencies with some specific attributes. (package version, project build type and the toolchain used)

For example, I would like to be able to find/save/search, from my various projects, all find_package/find_library function calls, their build schemes, their requested version and probably the toolchain I used to build the project.

To summarize, how is it possible to store persistently all the info for a specific project into a global graph and to create a root path specific for each of my projects, ideally the project vcs uri as an index) ?

eg:

|- Project-URI (eg. github.com/username-1/project-name-1)
   |- Project-Name (eg. dlib-for-arm)
      |- Project-Version (eg. user defined version or a git dirty hash)
          |- Toolchain-ID (ios-universal or android-ndk-r12b-api-19-armeabi-v7a-neon)
              |- Build-Type (Debug
              |- Options 
                 |- (eg) BUILD_TEST=OFF 
              |- Packages 
                 |- Name: Boost  
                 |- Input:
                    |- Arguments: REQUIRED  
                    |- Version: Boost  
                    |- Components:  filesystem system serialization  
                 |- Output
                    |- Results:  BOOST_FOUND=val1,  BOOST_INCLUDE_DIRS=val2, ... 

To extract most of my required informations, I found some solutions like CMake-AST python package but for the graph creation/import I am still trying to figure out the best approach to register those dependencies. So, I have already have a list of nodes to register because I already extract all the function calls in the CMake list (eg. https://github.com/chrisvm/pymake/blob/master/pymake/cmake_list.py)

I just want to create a structured graph logging and allowing me to search for similarities between my projects in order to override some future/new projects with some CMake rules based on machine learning.

Let me know if my context introduction sounds a little bit unclear.

Ps. I know the Graphviz feature available with CMake but I want to deal with this graph but as mentioned above i want to do some machine learning processing with python (package recommendation engine, package ovverides suggestions per toolchain) on all these accumulated/aggregated/extracted data.

Have an awesome day !

Cheers,
Richard