joergbrech / cpp_include_walker

my first rust project

C++ Include Walker

Read the documentation

The idea is to parse all of the #include statements in a set of .cpp, .cxx, .c, .h, .hpp, .hxx files, build a depencency forest and provide some functionality, such as

• finding circular dependencies (either using Johnson's alg, or during topo sort by Kahn's method),
• finding superfluous #include statements (superfluous in the sense of "include only what you need")
• sort header files topologically.

In the long run, it would be awesome if this could be used to automatically generate .i files for swig.

Note

Other tools use a compiler to obtain the AST.

• Con: Additional dependency, how would you do this platform-independently?
• Pro: This would be useful to catch platform dependent includes and not make assumptions about include guards.

cpp_include_walker's Issues

assumption on include directories

Currently, the two dependencies

#include<vector>
#include "myImpl/vector.hxx"

get the same key, because it is implicitly assumed that all subdirectories of the root directory are passed to the compiler as include directories.

It makes more sense to assume that these point to different files.

implement topo sort

so far, the nodes are only sorted by used_by.len(), which is the first step.

keyify "uses" vector of Dependency node

make paths relative to source dir

rethink SimpleGraph

SimpleGraph is really lean, because it makes no assumptions on memory management and the generic data type. This makes some things hard or inefficient, such as determining the index of a child node, since there is also no assumption on ordering:

cpp_include_walker/src/simple_graph.rs

Line 65 in 85af696

let idx = nodes.iter().position(|&x| x==child).unwrap();

unit test for simple_graph

There are no unit tests for SimpleGraph, because implementing is not easy. One way to levitate this is to allow values and references into the nodes and children functions of the trait maybe...

make it a CLI