Something to think about: We should allow branching of JSON changes. For this, a commit ID needs to be stored in a DAG. Now how to efficiently rollback changes given a DAG? Is there such a thing as random access to a newline delimited JSON file? Unlikely, given that the key order can be arbitrary.
Or simply store changes in one file per commit, and name them according to commit ID. A dataset will likely not experience > 10k changes, so we should be fine.
rollback json changes (with jp)
Something to think about: We should allow branching of JSON changes. For this, a commit ID needs to be stored in a DAG. Now how to efficiently rollback changes given a DAG? Is there such a thing as random access to a newline delimited JSON file? Unlikely, given that the key order can be arbitrary.
Or simply store changes in one file per commit, and name them according to commit ID. A dataset will likely not experience > 10k changes, so we should be fine.
For now stick to linear case.