Alternatinve way of managing Blosc sources about python-blosc HOT 6 CLOSED

See PR #13 for the setup.

from python-blosc.

See PR #14 for the submodule version.

from python-blosc.

See PR #15 for the subtree version.

from python-blosc.

So, now for some discussion.

The submodule requires additional commands of those users who checkout the repository. However, the sources themselves are never actually checked into the repository. The subtree merge requires additional commands on the side of the person merging the sources from upstream. In the current instructions, I have chosen to use --squash, which means that each time a new Blosc version is merged, all the new commits are squashed into a single commit. In previous versions of the subtree merge approach, all commits in the history would be merged. This was sometimes confusing because the history of the original project would be quite polluted even with tags of the merged project. To clarify what an update would look like, the most recent commit of #15 is a merge from master. One advantage of the subtree merge is, that the users need not worry about anything, since the sources are part of this project. On the downside of course, the sources are actually checked in to the repository, but this shouldn't be too bad since Blosc sources are quite small.

Perhaps you should clone my fork into /tmp when testing these approaches, since the submodule approach may do weird things, when switching branches, esp. if the new branch doesn't use submodules.

from python-blosc.

We have decided to go with the subtree approach, see #15 and #23.

from python-blosc.

Subtree approach introduced in commit 77ca657

from python-blosc.