conan.tools.layout

Warning

This is an experimental feature subject to breaking changes in future releases.

Available since: 1.37.0

LayoutPackager

The LayoutPackager together with the package layouts feature, allow to automatically package the files following the declared information in the layout() method:

It will copy (being xxx => build and source (if destination is only one dir):

  • Files from self.cpp.xxx.includedirs to self.cpp.package.includedirs matching self.patterns.xxx.include patterns.

  • Files from self.cpp.xxx.libdirs to self.cpp.package.libdirs matching self.patterns.xxx.lib patterns.

  • Files from self.cpp.xxx.bindirs to self.cpp.package.bindirs matching self.patterns.xxx.bin patterns.

  • Files from self.cpp.xxx.srcdirs to self.cpp.package.srcdirs matching self.patterns.xxx.src patterns.

  • Files from self.cpp.xxx.builddirs to self.cpp.package.builddirs matching self.patterns.xxx.build patterns.

  • Files from self.cpp.xxx.resdirs to self.cpp.package.resdirs matching self.patterns.xxx.res patterns.

  • Files from self.cpp.xxx.frameworkdirs to self.cpp.package.frameworkdirs matching self.patterns.xxx.framework patterns.

Usage:

from conans import ConanFile
from conan.tools.layout import LayoutPackager

class Pkg(ConanFile):

    def layout(self):
        ...

    def package(self):
        LayoutPackager(self).package()

Predefined layouts

There are some pre-defined common layouts, ready to be simply used in recipes.

For example, this would implement the standard CMake project layout:

from conan.tools.layout import cmake_layout

def layout(self):
    cmake_layout(self)

If you want to try it, use the conan new hello/0.1 -m v2_cmake template.

It is very important to note that this cmake_layout() is just calling the folders and cpp attributes described before:

def cmake_layout(conanfile, generator=None):
    gen = conanfile.conf["tools.cmake.cmaketoolchain:generator"] or generator
    if gen:
        multi = "Visual" in gen or "Xcode" in gen or "Multi-Config" in gen
    elif conanfile.settings.compiler == "Visual Studio" or conanfile.settings.compiler == "msvc":
        multi = True
    else:
        multi = False

    conanfile.folders.source = "src"
    if multi:
        conanfile.folders.build = "build"
        conanfile.folders.generators = "build/conan"
    else:
        build_type = str(conanfile.settings.build_type).lower()
        conanfile.folders.build = "cmake-build-{}".format(build_type)
        conanfile.folders.generators = os.path.join(conanfile.folders.build, "conan")

    conanfile.cpp.source.includedirs = ["."]
    if multi:
        conanfile.cpp.build.libdirs = ["{}".format(conanfile.settings.build_type)]
    else:
        conanfile.cpp.build.libdirs = ["."]

First, it is important to notice that the layout depends on the CMake generator that will be used. So if defined from [conf], that value will be used. If defined in recipe, then the recipe should pass it as cmake_layout(self, cmake_generator).

The definitions of the folders is different if it is a multi-config generator (like Visual Studio or Xcode), or a single-config generator (like Unix Makefiles). In the first case, the folder is the same irrespective of the build type, and the build system will manage the different build types inside that folder. But single-config generators like Unix Makefiles, must use a different folder for each different configuration (as a different build_type Release/Debug).

Finally, the location where the libraries are created also depends. For multi-config, the respective libraries will be located in a dedicated folder inside the build folder, while for single-config, the libraries will be located directly in the build folder.

This helper defines a few things, for example that the source folder is called "src". This could be customized without fully changing the layout:

def layout(self):

cmake_layout(self) self.folders.source = “mysrcfolder”

Even if this pre-defined layout doesn’t suit your specific projects layout, it is a good example how you could implement your own logic (and probably put it in a common python_require if you are going to use it in multiple packages).