fix(appris/mane): allow data source to determine availability#157
fix(appris/mane): allow data source to determine availability#157
Conversation
|
@ovesh What's the process now to upload new files? I'll also need write access to the DG plugins. |
Previously, GK hardcoded the filenames to encode what upstream version was used to build APPRIS/MANE. The issue with this is that older versions of GK could not be updated to use new APPRIS/MANE (ie, a new package with updated locator code needed to be released). This change brings APPRIS/MANE in line with annotations: if they exist on the data source, then it's available to use in GK. BREAKING CHANGE: this change also fixes gencode.v41 to use the correct source (e107 now instead of e103).
s22chan
force-pushed
the
fix/allow_appris_version_overrides
branch
from
cd58d14 to
1d51fe7
Compare
For now, uploads to the open source bucket have to go through me. |
| if name.endswith(".2bit") or name.endswith(".cfg") | ||
| ) | ||
|
|
||
| names.update(get_genomes(os.listdir(self.data_dir))) |
There was a problem hiding this comment.
This is a good change (list_available_genomes also returns genomes available only in the local dir), can you explicitly add that to the PR description?
genome_kit/gk_data.py
Outdated
| DataManagerImpl = list(eps)[0].load() | ||
| except: | ||
| DataManagerImpl = DefaultDataManager | ||
| DataManagerImpl = DefaultDataManager |
There was a problem hiding this comment.
Won't this change break code that relies on implicitly loading the data manager from a plugin?
There was a problem hiding this comment.
sorry this was test code
There was a problem hiding this comment.
Note: It's hard to see on Github UI, but the file moves in the PR all just remove the appris version from the file name, e.g
appris.2018_12.v28_gencode.v29.mini.v7.pkl => appris.gencode.v29.mini.v7.pkl
Can this also be explicitly called out in the PR description?
|
I'll maybe leave you to squash it (when ready), as I'm not sure how to upload the data ahead of a release anymore. |
2 participants
Previously, GK hardcoded the filenames to encode what upstream version was used to build APPRIS/MANE. The issue with this is that older versions of GK could not be updated to use new APPRIS/MANE (ie, a new package with updated locator code needed to be released).
This change brings APPRIS/MANE in line with annotations: if they exist on the data source, then it's available to use in GK.
To achieve this, all the APPRIS/MANE filenames now only contain the annotation and the binary GK version to make things discoverable without the hardcoded mappings.
A clubbed change is also to list any local genomes for
genome.list_available_genomes: this helps with building APPRIS/MANE on any annotations that have yet to be uploaded.drive-by change: this change also fixes gencode.v41 to use the correct source (e107 now instead of e103).