cpp-includes-graph

Small tool to visualize how are made inclusions in your project and help you to find easily unecessary ones to reduce compilation time and improve IDE reactivity.

This project is similar to https://github.com/tomtom-international/cpp-dependencies. Some other tools like https://include-what-you-use.org/ seems to me much more compilcated to set up and slower (at it relies on a real cpp compiler).

What it does

• Parse quickly given source folders and find includes directive
• Output a dot file that is used to generate an image of the graph
• It assume that the given code is correct
• Pretty simple to use

What it doesn't do

• It doesn't help you to know which header file is slow to build
• It doesn't help you to know if an inclusion can be avoided (forward declaration sufficient,...)

Use cases

Improving compile time and refactoring

I originally made this tool to help me to reduce compile time of a project before doing a more in depth refactoring. So it have to be robust, fast and clear to be useful on a big code base. Sadly I discover that the point when the graph is too big for dot being able to generate the image come pretty soon. Gracefully with the configuration file it is pretty easy to split a code base into multiple sub-projects to reduce graph size. If dot failed to generate the image simply reduce the number of input sources by selecting sub-folders.

Monitoring evolution of a new project

I think that it also can be useful to check regulary if everything evolves in the right way as your project will grow.

Dependencies

• https://www.graphviz.org binaries (should be in PATH environment variable)
• cpp17 (actually only visual studio 2019 is supported)

Example

The result looks like:

TODO

Configuration

• Add an ignore list that can contains folders or source file path
• Let the user choosing the image output file format
• Subgraph per folder?

Graph

• Improve link color, to be able to show most used headers (depending of a computationnal ratio)
• Make the user able to choose a color for a specific header (can be usefull to find it quickly in a big graph)
• Push orphan headers in the graph (not linked to a source file)
• Add sub-graph per directories (optionnal)
• HTML output format with active links to files?

Implemenation

• Fix unique_name of nodes generation
• Resolve macro language conditions (only if the user set some defines)
• Parallelize per project
• Configuration file: Support empty string list
• Stats: Add the total execution time
• Do we need to factorize parsers? It seems to be possible to make a generic tokenizer that takes languages definitions as parameters,...
• Investigate on cases that makes dot crash

Release

• Add dot binary?