This toolkit is known to run on Windows 10. It has run in the
past on Windows XP and Windows 7, but we no longer have access to
either for testing. We have no access to Windows Vista or Windows
8, and don't expect to.
This toolkit depends on a number of external tools which must be
installed before you install MAT. For the Windows version, they
are:
MAT is primarily a Unix-based tool, so the documentation is written mostly with reference to Unix. As a result, you should first familiarize yourself with the Unix conventions, and make sure you know the DOS CMD equivalents of basic Unix operations like pwd, cd, etc. We'll assume throughout that you've set your MAT_PKG_HOME cmd shell variable to the directory src/MAT inside the unpacked zip file.
The documentation assumes you've added Python to your Windows
path.
MAT may be installed in
directories whose paths contain spaces. If you do this,
you'll likely have to wrap double-quotes around any references to
MAT_PKG_HOME or subdirectories thereof:
> cd "%MAT_PKG_HOME%"
> dir "%CD%"
If you don't do this, you'll likely encounter bizarre behavior
due to the path being expanded and split into command-line tokens
according to the whitespace in the path.
First, ensure that you have Java installed. You'll need version
8, 9, 10, or 11. The installer looks for any version of Java found
in C:\Program Files\Java. The installer will probe all the Java
versions and choose the newest one that meets the system
requirements.
Next, install Python. You don't need any of Mark Hammond's Python
win32 extensions. You may want to add Python to your Windows path.
None of the instructions in this documentation should require
that, but if you do find an instruction to run python.exe, use the
full executable path for Python if you haven't added it to your
path..
Next, open a Windows cmd shell, e.g., via Start -> Run -> cmd. We recommend that you ensure that the output history is longer than the default 50 lines; click left on the leftmost icon in the top window bar in the "cmd" window, select Properties, select the Layout tab, and change "Height" under "Screen Buffer Size" to, say, 500.
Next, open the unzipped directory containing this distribution in
your file viewer (e.g., navigating from "My Computer"). In the top
directory of the distribution, you should see a file "install.py".
In your cmd shell window, type "python" (if you've added Python to
your Windows path), or the full path to the executable (e.g.,
"C:\Python27\python.exe" or "C:\Python36\python.exe"), and then a
space, and then drag the "install.py" file to the cmd shell
window. Switch to the cmd shell and press <return>. This
will run the MAT installer.
The output of the installer is fairly verbose. If it does not
print out "Done" at the end, it hasn't completed appropriately.
When the installation is complete, your runtime environment will be
configured for you. If you want to do something fancy during your
installation (e.g., use your own settings file, or store the task
plugin record outside your installation), see install.py.
As of MAT 3.3, the core of MAT can also be distributed as a pip-compatible Python library. This library contains the core MAT APIs, but lacks the Web server, the jCarafe engine, the documentation, and the command-line tools. Its sole dependency, a library implementing the Kuhn-Munkres bipartite set alignment algorithm, is installed as a dependency when the pip module is installed (and is only needed for the pairer/scorer).
While it is possible to override MAT's runtime environment using
environment variables, it's not recommended. So you may have to do
something special if you want to change your installation in any
way.