BoostDepGraph

A tool to visualize the boost dependency graph.

WARNING: This project is still very much in flux (alpha version), so expect the Readme to be out of sync with the code and occasionally broken commits.

Basic Usage

• First, you need a recursive copy of the boost project https://github.com/boostorg/boost (i.e. the libs sub-folder contains a copy of all boost libraries)

  git clone --recursive https://github.com/boostorg/boost
  

• After starting the program, you'll be asked for the boost root directory you want to analyze

• Next you will be asked if you are only interested in a particular module ¹⁾. Leave as "none" if you want to view the whole dependency graph.

• After analysis is finshed (which takes a couple of seconds on my system) it will show you the dependency graph between the boost libraries:

  • A library BoostA depends on library BoostB, if any of the header or source files of BoostA (not the test files) includes any of the headers of library BoostB.
  • The analysis does not consider any form of conditional inclusion, so if e.g. Boost.Array is only included when compiled in pre-c++11 mode, it will still show a dependency.

• The color coding of the nodes refers to their cmake status:

  • Green: The library has a CMakeLists.txt file in its root directory.
  • Yellow: There is no CMakeLists.txt file in the root, but all dependencies have one.
  • Red: No CMakeLists.txt file present and at least one dependency (direct or indirect) also doesn't have a cmake file.

1) This will do a include-based filtering of the graph. Only libraries whose files are actually included directly or indirectly by the files of the given root module will be displayed. E.g. If a file a.h of the root module includes a header d1.h of Dep1 and a different header d2.h of Dep1 includes a file from serialization serialization, but d2.h isn't included by any file in the root module, serialization will NOT be shown in the dependency graph

Controls:

• Left clicking on a node will select it and highlight all connected edges.
• Right clkicking on a node will toggle its cmake status and update the rest of the graph accordingly.
• You can move nodes around via drag and drop.
• You can zoom in and out via mouse wheel.
• Hit space to pause/continue the auto movement of the nodes.
• Hit enter to rerun the analysis.

Supported Platforms and Dependencies:

• The code should be portable c++ code, but so far, development and testing is only happening on VS2017 and VS2019.
• The code is written in iso c++17 and makes heavy use of std::filesystem.
• It uses Qt5 widgets for visualization - development happens against Qt 5.12 on windows.
• Some gcc and clang versions require the explicit linking with the c++17 filesystem library (e.g. via -lstdc++fs or -lc++fs ). The current cmake script doesn't do that.

Acknowledgements:

• The code extracting the dependency information from the boost repository is loosely based on boostdep from Peter Dimov (https://github.com/boostorg/boostdep).
• The visualization code was initially based on the Elastic Nodes Example on the qt homepage (https://doc.qt.io/qt-5/qtwidgets-graphicsview-elasticnodes-example.html).

However, both of those parts have been completely rewritten by now.