Filesystem Layout

This document describes the filesystem hierarchy of BASIS projects, which is based on the Filesystem Hierarchy Standard of Linux. It has a goal of supporting:

  • Unix and Windows
  • Installation of multiple versions of each package on a single system
  • Seamless integration of BASIS software packages
  • A superproject, or super-build, concept based on a bundle build

Please note that the variable names used below are defined by BASIS using CMake, and will often refer to particular directories of a software project. These variables should be used where possible, so that directories can be renamed without breaking the build system.

The Project Template provides a reference implementation of this standard. See the Create/Modify a Project How-to Guide for details on how to make use of this template to create a new project which conforms with the filesystem hierarchy standard detailed in this section.

Legend

  • <project> (<package>) is a placeholder for the lowercase project (or package) name
  • <Project> is the case-sensitive project name.
  • <major> is the major release number
  • <minor> is the minor update number
  • <patch> is the patch number
  • <version> is a placeholder for the project version string <major>.<minor>.<patch>
  • <source> is the root directory of a particular project source tree
  • <build> is the root directory of the project’s build or binary tree

Source Code Repository

Git

BASIS recommends that Git distributed version control users follow the nvie git-flow branching model. The Atlassian Gitflow Workflow Tutorial is another excellent source for this information.

Mercurial

BASIS recommends that Mercurial (hg) distributed version control users follow the hg-flow branching model. This is identical to the git-flow branching model explained in Source Code Repository, but uses mercurial as the version control system. The hg-flow extension is useful for assisting with development, but not required.

Subversion

Each Subversion (SVN) repository contains the top-level directories trunk/, branches/, and tags/. No other directories may be located next to these three top-level directories.

The root directory of a development branch, typically the trunk (see Subversion), is denoted by <tag> and considered relative to the base URL of the project repository. The base URL is referred to as <url>.

Repository Path Description
trunk/ The current development version of the project. Most development is done in this master branch.
branches/<name>/ Separate branches named <name> are developed in subdirectories under the branches directory. One reason for branching is, for example, to develop new features separate from the main development branch, i.e., the trunk, and merging the desired changes back to the trunk once the new feature is implemented and tested.
branches/<project>-<major>.<minor>/ This particular branch is used prior to releasing a new version of the project. This branch is commonly referred to as release candidate of version <major>.<minor> of the project. It is used to adjust the project files prior to tagging a particular release. For example, to set the correct version number in the project files. This branch is further be used to apply bug fixes to a previous release of this version, in which case the patch number has to be increased before tagging a new release of this software version. See the Branch and Release guide for further details.
tags/<project>-<version>/ Tagged release version of the project. The reason for including the project name in the name of the tagged branch is, that SVN uses the last URL part as name for the directory to which the URL’s content is checked out or exported to if no name for this directory is specified.

See the Branch and Release guide for details on how to create new branches and the process of releasing a new version of a software project.

Below the trunk and the release branches a version of the entire source tree should be present. Other branches below the branches/ directory may contain a subset of the trunk such as the source code of the software without the examples and tests.

Source Code Tree

The Soruce Code Tree refers to the filesystem directory structure of all source code that is managed by version control. The build and intallation trees are separate entities created and populated from the source tree, so the source tree is essentially the “beating heart” of a software project.

Source Categories Source files can fall under the categories of software, build, configuration, documentation, or testing. Any files essential to the execution of the software are also considered to be part of the software source. Examples of essential files include a pre-computed lookup table and a medical image atlas.

Documentation Examples within a software project are considered to be part of both documentation and testing.

Testing The testing category can be divided into system testing and unit testing. It is important to note the difference of system tests and unit tests. As testing can often require a huge amount of image data, these datasets may be stored and managed outside the source tree. Please refer to the Managing Test Data guide for details on this topic.

  • System Tests System tests are usually implemented in a scripting language such as Python, Perl, or BASH. System tests simply run the built executables with different test input data and compare the output to the expected results. Therefore, system tests can also be performed on a target system using the installed software where both the software and system tests are distributed as separate binary distribution packages. Large data sets, such as medical image data sets in their entirety, should only be required for system tests and downsampled to a very low resolution for practical reasons whenever possible.
  • Unit Tests Unit tests, provide a specialized test of a single software module such as a C++ class or Python module. Generally, the size and amount of additional data required for unit tests is kept reasonably small. The unit tests are compiled into separate executable files called test drivers. These executable files are not essential for the functioning of the software and are solely build for the purpose of testing.

Source Code Filesystem Heirarchy The filesystem hierarchy of a software project’s source tree is defined below. The names of the CMake variables defined by BASIS are on the left, while the actual names of the directories are listed on the right:

- PROJECT_SOURCE_DIR              - <source>/
    + PROJECT_CODE_DIR                + src/
    + PROJECT_CONFIG_DIR              + config/
    + PROJECT_DATA_DIR                + data/
    + PROJECT_DOC_DIR                 + doc/
    + PROJECT_EXAMPLE_DIR             + example/
    + PROJECT_MODULES_DIR             + modules/
    + PROJECT_TESTING_DIR             + test/
    + PROJECT_SUBDIRS                 + <multiple additonal subdirs>

Here are CMake variables defined in place of the default name for each of the following directories:

Directory Variable Description
PROJECT_SOURCE_DIR Root directory of source tree.
PROJECT_CODE_DIR All source code files.
PROJECT_CONFIG_DIR BASIS configuration files.
PROJECT_DATA_DIR Software configuration files including auxiliary data such as medical atlases.
PROJECT_DOC_DIR Software documentation.
PROJECT_EXAMPLE_DIR Example application of software.
PROJECT_MODULES_DIR Project Modules, each residing in its own subdirectory.
PROJECT_TESTING_DIR Implementation of tests and test data.
PROJECT_SUBDIRS List of additional directories for source code files.

Build Tree

CMake supports but recommends against in-source builds. Therefore, BASIS requires that the build tree be outside the source tree. Only the files in the source tree are considered to be important.

Directories in the build tree are separate from the source tree, and they are created and populated when CMake configuration and the build step are run.

- PROJECT_BINARY_DIR              - <build>/
    + RUNTIME_OUTPUT_DIRECTORY        + bin/
    + LIBRARY_OUTPUT_DIRECTORY        + lib/
    + ARCHIVE_OUTPUT_DIRECTORY        + lib/
    + TESTING_RUNTIME_DIR             + Testing/bin/
    + TESTING_LIBRARY_DIR             + Testing/lib/
    + TESTING_OUTPUT_DIR              + Testing/Temporary/

Here are CMake variables defined in place of the default name for each of the following directories:

Directory Variable Description
RUNTIME_OUTPUT_DIRECTORY All executables and shared libraries (Windows).
LIBRARY_OUTPUT_DIRECTORY Shared libraries (Unix).
ARCHIVE_OUTPUT_DIRECTORY Static libraries and import libraries (Windows).
TESTING_RUNTIME_DIR Directory of test executables.
TESTING_LIBRARY_DIR Directory of libraries only used for testing.
TESTING_OUTPUT_DIR Directory used for test results.

Installation Tree

Installation Schemes

An installation scheme is a specific installation tree layout that is utilized based on contextual information.

BASIS automatically switches the installation scheme if you change CMAKE_INSTALL_PREFIX from /opt/... to /usr/... and vice versa.

The following directory structure is used when installing the software package, either by building the install target with “make install”, extracting a binary distribution package, or running an installer.

Different installation hierarchies are defined in order to account for different installation schemes depending on the location and target system on which the software is being installed.

The directory structures including the installation is defined in DirectoriesSettings.cmake. The default “/opt” prefix is hard coded for Unix. On Windows it is “C:/Program Files” or, if the registry value can be read, the corresponding directory in the installation language of the OS, e.g., “C:/Programme” in German.

BASIS knows about a few “installation schemes”. These distinguish between common filesystem hierarchy standards such as the one for “/opt” or “/usr” on Unix. The difference is that under “/opt”, packages are installed in their own respective subdirectories which contain then subdirectories such as “include”, “lib”, “doc”, etc. Under the “/usr” directory, however, the hierarchy is first divided by “include”, “lib”, “bin”, “doc”, and then by package name.

Possible Schemes

The first installation scheme is referred to as the usr scheme which is in compliance with the Linux Filesystem Hierarchy Standard for /usr:

- CMAKE_INSTALL_PREFIX           - <prefix>/
    + INSTALL_CONFIG_DIR             + lib/cmake/<package>/
    + INSTALL_RUNTIME_DIR            + bin/
    + INSTALL_LIBEXEC_DIR            + lib/<package>/
    + INSTALL_LIBRARY_DIR            + lib/<package>/
    + INSTALL_ARCHIVE_DIR            + lib/<package>/
    + INSTALL_INCLUDE_DIR            + include/<package>/
    + INSTALL_SHARE_DIR              + share/
        + INSTALL_DATA_DIR               + <package>/data/
        + INSTALL_DOC_DIR                + doc/<package>/
        + INSTALL_EXAMPLE_DIR            + <package>/example/
        + INSTALL_MAN_DIR                + man/
        + INSTALL_INFO_DIR               + info/

Another common installation scheme, here referred to as the opt scheme and the default used by BASIS packages, follows the Linux Filesystem Hierarchy Standard for Add-on Packages:

- CMAKE_INSTALL_PREFIX           - <prefix>/
    + INSTALL_CONFIG_DIR             + lib/cmake/<package>/
    + INSTALL_RUNTIME_DIR            + bin/
    + INSTALL_LIBEXEC_DIR            + lib/
    + INSTALL_LIBRARY_DIR            + lib/
    + INSTALL_ARCHIVE_DIR            + lib/
    + INSTALL_INCLUDE_DIR            + include/<package>/
    + INSTALL_SHARE_DIR              + share/
        + INSTALL_DATA_DIR               + data/
        + INSTALL_DOC_DIR                + doc/
        + INSTALL_EXAMPLE_DIR            + example/
        + INSTALL_MAN_DIR                + man/
        + INSTALL_INFO_DIR               + info/

The installation scheme for Windows is:

- CMAKE_INSTALL_PREFIX           - <prefix>/
    + INSTALL_CONFIG_DIR             + CMake/
    + INSTALL_RUNTIME_DIR            + Bin/
    + INSTALL_LIBEXEC_DIR            + Lib/
    + INSTALL_LIBRARY_DIR            + Lib/
    + INSTALL_ARCHIVE_DIR            + Lib/
    + INSTALL_INCLUDE_DIR            + Include/<package>/
    + INSTALL_SHARE_DIR              + Share/
    + INSTALL_DATA_DIR               + Data/
    + INSTALL_DOC_DIR                + Doc/
    + INSTALL_EXAMPLE_DIR            + Example/

In order to install different versions of a software, choose an installation prefix that includes the package name and software version, for example, /opt/<package>-<version> (Unix) or C:/Program Files/<Package>-<version> (Windows).

Note that the directory for CMake package configuration files is chosen such that CMake finds these files automatically given that the <prefix> is a system default location or the INSTALL_RUNTIME_DIR is in the PATH environment.

It is important to note that the include directory always contains the package name. This way, project header files must use an include path that avoids conflicts with other packages that use identical header names. Here is a usage example:

#include <package/header.h>

Thus, the include directory that is added to the search path must be set to the include/ directory, but not the <package> subdirectory.

Here are CMake variables defined in place of the default name for each of the following directories:

Directory Variable Description
CMAKE_INSTALL_PREFIX Common prefix (<prefix>) of installation directories. Defaults to /opt/<provider>/<package>-<version> on Unix and C:/Program Files/<Provider>/<Package>-<version> on Windows. All other directories are specified relative to this prefix.
INSTALL_CONFIG_DIR CMake package configuration files.
INSTALL_RUNTIME_DIR Main executables and shared libraries on Windows.
INSTALL_LIBEXEC_DIR Utility executables which are called by other executables only.
INSTALL_LIBRARY_DIR Shared libraries on Unix and module libraries.
INSTALL_ARCHIVE_DIR Static and import libraries on Windows.
INSTALL_INCLUDE_DIR Public header files of libraries.
INSTALL_DATA_DIR Auxiliary data files required for the execution of the software.
INSTALL_DOC_DIR Documentation files including the software manual in particular.
INSTALL_EXAMPLE_DIR All data required to follow example as described in manuals.
INSTALL_MAN_DIR Man pages.
INSTALL_MAN_DIR/man1/ Man pages of the executables in INSTALL_RUNTIME_DIR.
INSTALL_MAN_DIR/man3/ Man pages of libraries.
INSTALL_SHARE_DIR Shared package files including required auxiliary data files.

Forcing Schemes

Schemes can be selected using the CMake -DBASIS_INSTALL_SCHEME variable.

You can force BASIS to use one specific scheme using BASIS_INSTALL_SCHEME. For example, if you want to install the software in /usr/<package> using the same hierarchy typically used under “/opt”.