Code Conventions
This page details the coding conventions used within the CMake project. We specifically focus on the conventions to use when writing CMake/CMakePP code as Python code should adhere to PEP-8.
Module Setup
CMake/CMakePP code is written in modules. A typical CMake module file should look like:
#
# 1. The license boilerplate header
#
include_guard() # 2
include(a_submodule) # 3
include(b_submodule)
#[[[
#
# 4. Documentation comment for class/function/variable
#
#]]
function(my_fxn)
# Defined the function
endfunction()
For the most part this follows the typical conventions in Python and C/C++. The main pieces are:
All module files must start with the boilerplate license of the Apache 2.0 license. To maintain uniformity please/copy paste this verbatim from another file.
The first actual command in a CMake module should always be
include_guard()
include_guard()
was introduced in CMake 3.10 and serves the same purpose as#pragma once
in C/C++, namely to avoid including the same module more than once, even if multiple modules attempt to include it.
You should include all CMake modules required to use your module.
Modules should be listed in alphabetic order
It should be possible to use your module simply by including it, i.e., users should not have to include additional modules in order for your module to work
If your module depends on a module A, which in turn depends on B, you should directly include B in your module only if your module uses B too.
Using B to interface with A does not count as your module using B.
After the top matter, you should include your code. All code should have API documentation immediately proceeding it. The API documentation should follow CMakePP’s documentation standards.
You can have multiple functions, classes, etc. in the same CMake module if you want.
Generally speaking CMake/CMakePP are very verbose languages, which can require a great deal of comments to understand quickly. For this reason it is often beneficial for readability to only include a single function or class in a module.
Additional considerations:
Please limit your code to 80 characters per line.
There are plenty of StackOverflow and blog posts on why this is a good idea. To summarize them:
Ease of reading. The human eye gets tired quicker if it has to scroll horizontally too much.
Many people often have more than one source file open at a time. Using an IDE with a reasonable font size, two sources file side-by-side basically takes up an entire monitor if they are around 80 characters per line.
Easier to embed code in documents, emails, etc. without additional formatting.
The choice of 80 versus say 85 is somewhat arbitrary and is primarily made to adhere to usual conventions.
Another motivation for 80 is that in high-performance computing it is often convenient to edit a file from a terminal. For better or for worse, for legacy reasons terminals tend to be based on 80 characters.