-
Notifications
You must be signed in to change notification settings - Fork 5
Toolboxes class
Flywheel toolboxes for Matlab (FTM) is a class helps you share Matlab functions reproducibly with colleagues. In many cases, you will be sharing the function with your future self. The idea is to write a Matlab function that uses scitran to access Flywheel data, and uses toolboxes to download Matlab libraries from github that are needed to run these functions.
There are two steps in using toolboxes. First, create an FTM file and upload it to the project page. Second, include a call to the scitran toolbox method that reads this toolbox file and checks that the user has the toolboxes installed on their path. If the toolboxes are not found, the scitran toolbox method will try to do the install.
FTM is useful for custom software and software development. Flywheel uses the Gears concept, docker containers packaged with a manifest that allows you to control the parameter setting, for executing standard tools. The Gears system also includes job control for scheduling and logging jobs. Flywheel is well on its way towards implementing an elastic cloud system using distributed storage and kubernetes (k8s).
The toolboxes object (in this case called tbx) contains two types of information. One is a command that can be executed to test whether the toolbox is on the path (testcmd). The second is a structure that contains enough information to download a git repository.
disp(tbx)
toolboxes with properties:
testcmd: 'dtiError'
gitrepo: [1×1 struct]
The gitrepo structure contains this information
>> disp(tbx.gitrepo)
user: 'scitran-apps'
project: 'dti-error'
commit: 'master'
scitran data directory includes several examples and a script s_tbxSave, that writes out toolbox files. They are written out as JSON files with the git repo projection and a .json extension.
tbx = toolboxes('');
tbx.testcmd = 'dtiError';
tbx.gitrepo.user = 'scitran-apps';
tbx.gitrepo.project = 'dti-error';
tbx.saveinfo;
Toolboxes can be combined into a single file and uploaded to a Flywheel project page. In this example, the stToolbox() function reads a JSON file for two repositories. These are placed in an array, and written out to a new JSON file.
tbx(1) = stToolbox('dtiError.json');
tbx(2) = stToolbox('vistasoft.json');
tbxWrite('aldit-toolboxes.json',tbx);
% Subsequently, your could read the combined file
% tbx = stToolbox('aldit-toolboxes.json');
% upload to the project page
project = st.search('project','project label exact','ALDIT');
st.upload('aldit-toolboxes.json','project',project{1}.project.x_id);
The key toolboxes methods are saveinfo, install and clone. These methods add a github repository with your project's toolbox onto your Matlab path. The scitran method, 'runFunction', downloads and executes a function stored on the Flywheel site, typically on a project page, that relies on these toolboxes.
We suggest that you include the toolboxes.install or toolboxes.clone within the function you are sharing and that will be executed by scitran.runFunction. The toolboxes command first check whether the toolbox is on your path, and the function only downloads if the toolbox is not present.
Read a toolbox.json file and get a zip archive, put it on path
Clone a a git repository and add it to your path. cloneDepth allowed.