DataLad can do many things, but one does not need to know every feature in order to make good use of it. As of version 0.12, one only needs just two commands that cover the most common scenarios for turning one's own data from a pile of files into a structured dataset that tracks the evolution of its content. This makes the life of novice users easier, but even Git pros might feel tempted to get a little lazy... just sometimes. The extension is available for installation on Linux/Mac/Windows from GitHub.
1. create (create a new dataset)
This command creates a new dataset. It can be run in an empty directory or, if a name for a new dataset is given, a new directory will be created. To create a new dataset called "myproject" run:
datalad create myproject
Datasets can be created anywhere and for any purpose. One can also create datasets within other datasets to link them together. For example: input data and results, or several studies that are part of a larger project.
2. save (record dataset state)
This is THE command. Whenever there was any change in a dataset (e.g. files/datasets were added, modified, or removed) just run
in the dataset to record the new state. Each such record can be used to compare changes over time or to restore to a previous state (sometimes things go wrong and time-travel becomes a very useful trick). These are already advanced scenarios, but just running save whenever some milestone is reached is all that is needed in order to make them possible.
It is extremely useful to leave notes for one's future self, to help explain why a change was made. The message option can be used to annotate a record with such information.
datalad save --message "Computed final results for paper submission"