A build tool for C++.


Keywords
cpp, build-tool, poetry, python, linux, windows
License
MPL-2.0
Install
pip install aim-build==0.1.33

Documentation

GitHub release (latest SemVer including pre-releases) GitHub commits since latest release (by SemVer including pre-releases) Python package PyPI - Python Version PyPI - Downloads GitHub contributors GitHub

Aim

A command line tool for building C++ projects.

Introduction

Aim is an attempt to make building C++ projects from source as simple as possible while encouraging a modular approach to software development.

Aim only requires a target.toml file which is used to specify the builds of your project. Each build specifies a component of your project, like a static library, dynamic library, or an executable.

Aim supports:

  • Windows with msvc frontend.
  • Linux with gcc frontend.
  • It should also be possible to use the gcc frontend on Windows when using GCC-like compilers but this hasn't be tested.

Getting Started

Prerequisites

Aim requires the following dependencies:

Installation

Aim is a python project and is installed using pip.

pip install --user aim-build

Using

There are 3 main commands:

  • init - initialises a directory with an empty project structure
  • list --target=path/to/target_toml_dir - displays the builds for the target
  • build --target=path/to/target_toml_dir <build name> - executes a build

For more information run:

aim <command> --help

The easiest way to get started is to use:

aim init --demo-files

aim init can be used to generate an empty project structure and the --demo-files flags will copy a small test application into the current directory for demonstration purposes.

You can then list the available builds of a target by specifying:

aim list --target=builds/linux-clang++-debug

And to build:

aim build --target=builds/linux-clang++-debug <build name>

Target files

A target.toml file describes a project and its build components.

Begin by specifying projectRoot which is the path from the target file to your source files. All relative paths will be relative to this directory.

The compiler frontend informs Aim how to construct the arguments for the compiler. Use gcc for GCC-like compilers and msvc for Microsoft cl-like compilers. Next specify the compiler, archiver, flags and any defines.

projectRoot = "../.."

compilerFrontend="gcc"
compiler = "clang++"
archiver = "ar"

flags = [
    "-std=c++17",
    "-O3",
    "-g",
    "-Wall",
]

# defines = [...] # Defines do not need the -D prefix.

Next specify your builds. For each build you must specify the name and buildRule. Valid build rules are staticLibrary, dynamicLibrary, executable, headerOnly or libraryReference. A build typically looks like:

[[builds]]
    name = "calculatorApp"
    buildRule = "executable"
    requires = ["calculatorDynamic"] # A list of dependencies for this build.
    outputName = "CalculatorApp"     # The output name. Aim will manage any prefixes or suffixes required.
    sourceFiles = ["src/*.cpp"]      # A list of path globs.
    includePaths = ["include"]       # A list of include paths.
    # The libraryPaths and libraries fields can be used to specify additional
    # libraries and paths to the build. This allows for linking against third
    # party libraries.
    #libraryPaths = []
    #libraries = []

Other notes:

  • The requires field is important as it is how you specify the dependencies for a build. For example, if you create a static library named "myAwesomeLibrary", this can be used in other builds simply by specifying requires=["myAwesomeLibrary"].

  • A headerOnly build does not have an outputName or sourceFiles as it is not built. The headerOnly rule is not essential and is mostly for convenience. If you have a header only library, repeating the include paths across several builds can be become repetitive. Instead, create a headerOnly build to capture the include paths and use it in other builds by adding the rule to the builds requires field.

  • A libraryReference does not have sourceFiles as it is not built. Like the headerOnly rule it is mostly for convience to reduce duplication. The primary use case is for capturing the includePaths, libraryPaths and libraries of a third party library that you need to use in a build. A libraryReference can then be used by other builds by adding it to a builds requires field.

  • The fields compiler, flags and defines are normally written at the top of the target file before the builds section. By default, all builds will use these fields i.e. they are global, but they can also be overridden by specifying them again in a build. Note that when these fields are specified specifically for a build, they completely replace the global definition; any flags or defines that you specify must be written out in full as they will not share any values with the global definition.

Supporting Multiple Targets

Aim treats any build variation as its own unique build target with its own unique target.toml.

A build target is some combination of things that affects the output binary such as:

  • operating system (Windows, OSX, Gnu Linux)
  • compiler (MSVC, GCC, Clang)
  • build type (Release, Debug, Sanitized)
  • etc.

Each build target and corresponding target.toml file must have its own directory ideally named using a unique identifier that comprises the 'parts' that make up the build. For example, builds/linux-clang++-release/target.toml indicates that the target file describes a project that is a release build, uses the clang++ compiler and is for the linux operating system.

As an example, if you were developing an application for both Windows and Linux, you may end up with a build directory structure like the following:

  • builds/linux-clang++-release/target.toml
  • builds/linux-clang++-debug/target.toml
  • builds/windows-clangcl-release/target.toml
  • builds/windows-clangcl-debug/target.toml

Note: each target.toml file must be written out in full for each target that you need to support. There is no way for target files to share information or to depend on another. While this leads to duplication between target files, it makes them very explicit and makes debugging builds much easier.

Advice Structuring Projects

If you structure your project/libraries as individual repositories then it may seem logical to nest dependencies inside one another. For example, if library B depends on library A, then B needs a copy of A in order for it to be built. So you may choose to nest the source of A inside B, perhaps using a git submodule.

The problem comes when your dependency hierarchy becomes more complex. If library C also depends on A, and an application D depends on B and C, you'll end up with multiple copies of library A which can become difficult to manage.

You may need to use this approach, as it is useful to be able to build a library in isolation, however you should do so in such a way where pulling the source for the dependencies is optional.

The approach the author uses is to use a non-project-specific directory that includes all your projects directly below it i.e. a "flat" structure. So rather than nesting dependencies you have:

 + MyProjects
 + - LibA
 + - LibB
 + - LibC
 + - ApplicationD
 + - SomeOtherApp
 + - builds
 + - - linux-clang++-debug
 + - - - target.toml

The flat structure has a single build directory and a single target file for each build target you need to support. This eliminates any duplication and is easy to manage. Aim is flexible enough that you can add additional levels to the project structure should you need to. For example, you may want to group all libraries under a libraries sub-directory. But the take-away message is that you should not enforce nested dependencies as this leads to duplication.

Developing Aim

Aim is a Python project and uses the poetry dependency manager. See poetry installation for instructions.

Once you have cloned the project, the virtual environment and dependencies can be installed by executing:

poetry install

Dev Install

Unfortunately, unlike setuptools, there is no means to do a 'dev install' using poetry. A dev install effectively generates an application that internally references the active source files under development. This allows developers to test the application without having to re-install the application after each change.

In order to use a development version of Aim on the command line, is it recommended creating an alias. The alias needs to:

  • add the Aim directory to PYTHONPATH to resolve import/module paths
  • execute the main Aim script using the virtualenv created by poetry

There are dev-env.bash and dev-env.fish scripts that configure this for you in the root of the Aim project directory. Note, these files must be sourced in order for them to work.