Fix for time_units in tskit 0.4.1 #55
Conversation
benjeffery
force-pushed
the
tskit-0.4.0
branch
from
a1a9b37 to
3f2677b
Compare
|
Thanks for the report @rwaples! We'll get a release out ASAP! |
benjeffery
force-pushed
the
tskit-0.4.0
branch
from
3f2677b to
71dc019
Compare
|
@jeromekelleher A quick fix for now, but I think tszip should be enumerating the properties it can cope with, rather than assuming the ones it doesn't recognise are arrays. |
tszip/compression.py
Outdated
| for name in columns: | ||
| if name.endswith("metadata_schema"): | ||
| if name.endswith("metadata_schema") or name == "time_units": |
There was a problem hiding this comment.
Would if isinstance(column[name], str) work here? Is the metadata column below also effectively assuming it's a bytes value?
There was a problem hiding this comment.
I've pushed a commit that does that on compress. On decompress we can't detect, although the best idea might be to write the type to the zarr metadata. Will try that later.
Codecov Report
@@ Coverage Diff @@
## main #55 +/- ##
==========================================
+ Coverage 97.70% 97.74% +0.04%
==========================================
Files 6 6
Lines 305 311 +6
Branches 55 61 +6
==========================================
+ Hits 298 304 +6
Misses 5 5
Partials 2 2
Continue to review full report at Codecov.
|
benjeffery
force-pushed
the
tskit-0.4.0
branch
from
61f512f to
e631e0d
Compare
|
@jeromekelleher I've updated this to store type info in the zarr - this should mean that tszip copes with any additional string or bytes attributes on tables. |
|
What's the status here @benjeffery - I think we decided to keep the file format changes for another version, and just do the decode side by name? |
Yep, I think I'll open another PR with those changes, and leave these here for reference if we do do the file change. |
|
Closing in favour of #58 |
Sounds good - could we open an issue to track the proposed changes to the file format? I think these would be an improvement and more future-proof. |
Fixes #54