The git version control system enables views into the history of a repository, such as names, dates, and commits, that can be useful in tracing the flows.
I am hoping to use it and pretty format in git’s log command to play around with the git clone --bare option to clone a repository without the code or data. At some moment, I may want that, but for now the metadata is fine. As reference, I am currently using: git log --pretty=format:"%h;%an;%at;%s" to read the data. It is still problematic as commas or semi-colons may appear in the message. A tab? One for when I do the mining itself, but grep may be a useful first start.
What I am hoping to do is to explore some of the challenges raised by David Berry in the Stunlaw blog on The Bonfire of the Models. My interest is in tracing changes for the dataset that may open up the histories of datasets for training.
Update: altered it to git log --pretty=format:"%h%x09%an%x09%at%x09%s so that the file is tab separated.
No Comments