Replies: 4 comments 3 replies
-
Many dependencies are not real dependencies. For example This CMake list is automatically generated, and therefore it picks up these dependencies. Having said that, we could manually update this CMake list, of course, and we probably should. |
Beta Was this translation helpful? Give feedback.
-
I am not against the manual update of |
Beta Was this translation helpful? Give feedback.
-
Apparently, it disturbs some users (the reason why this discussion is started). |
Beta Was this translation helpful? Give feedback.
-
My intention was not to question how CMakeLists.txt is generated (I was pointing to it as a concise list of other Boost header libraries I suspect are required), but rather try to understand if there is any effort underway in making Boost.Geometry more self-contained. In my experience, it is sometimes hard to convince others to adopt Boost.Geometry because it tends to pull a sizable number of other Boost headers with it. I have attempted to pare down the requirements for some projects, but my manual approach is not as thorough as I'd like. It's also a process I have to repeat every time I want to update to a newer version of the library. Maybe there are utilities these days that can help identify and remove unused headers? If so, I haven't been able to find any. So, what I'm really after is finding the minimal subset of Boost needed just for my specific use of the Geometry library. My ideal would be everything in a single repository that I can submodule and update easily when a new version is released. |
Beta Was this translation helpful? Give feedback.
-
Forgive me if this has already been discussed before, but one of my biggest challenges with the geometry library is that it depends on a lot of accessory boost libraries. I am not very familiar with all these dependencies, but several of them appear related to "recent" (e.g. at least since C++14) additions to the C++ Standard, and I imagine they could be replaced with standard library functionality.
I imagine many of them are needed to support code bases that are stuck on earlier versions of the Standard, but I'm curious if anyone (maybe on a separate fork?) has looked into removing the other boost dependencies and making geometry a more self-contained library.
Beta Was this translation helpful? Give feedback.
All reactions