Issue 225 #271
Issue 225 #271
Conversation
|
Might want to take a look at #270 ; I think this is the same patch from Google Code. We're on board in principle, but I think we're leaning towards a context manager rather than keywords on all the create_* functions. |
|
@andrewcollette Is there an easy way to get the creation time information out of a group or dataset object? My first attempt at writing the keyword argument for groups has no effect on the sha1 of the resulting file, which makes me think that either a) my code is broken or b) groups don't really carry this information. The high-level patch was written by @barrywardell, he may be in a better positioning to write the test for it. |
|
Right now I believe the only way is through h5py.h5g.get_objinfo. It returns an object with an mtime attribute: http://www.h5py.org/docs/low/h5g.html There's a similar function in h5py.h5o but the time fields are unimplemented. |
…e site on issue h5py#225 with the following message: 'I have tested these patches with the latest mercurial version and can confirm that they work fine for me. I also find the attached patch for the high-level interface convenient.' Shoved into git by Thomas Caswell <tcaswell@gmail.com>
The classes for data set creation (`PropDCID`) and group creation (`PropGCID) now inherit from the object creation class (`PropOCID`), which has the code for dealing with `H5P*obj_track_times`.
|
re-based this branch (did not merge entirely cleanly with the PR #272) |
|
After a bit of testing, it seems that groups always have an |
|
Is there still a test case needed or has that been covered by fbd3b31? |
|
@barrywardell, the test seems adequate (although the docstring is wrong) for datasets. I'm still trying to figure out what's going on with group mtimes. It may be intentional for performance reasons, but I would prefer to get this feature out across all objects in the file if there's a chance they could be affected. |
|
@andrewcollette Any progress on this/is there anything I should do? I think that I fixed the docstring in eb6ed0e. |
|
@tacaswell thanks for pinging me on this. From what I can discover group mtimes are not tracked in HDF5. I'll go ahead and merge this now. |
Issue 225 (disable mtime tracking for datasets)
re-based my patches and the patch from barry.wardell@gmail.com
This re-base needs to be tested, but by eye they look ok.