Opening this as a separate issue, as suggested by @manics in #1567 (comment), to continue the discussion.
As @wildintellect mentioned in that thread:
In the R world major.minor.patch is a actually a big deal on the minor part. Minor is about once per year and often has really important changes.
The current pinned version is now close to 2 years old.
Proposed change
Upgrade the default R version to something newer in
|
r_version = version_map["4.4"] |
@manics suggested applying the same policy we use for Python:
For Python we say
Starting in 2023, the default version of Python used when Python version is unspecified will be updated more often. Python itself releases a new version every year now, and repo2docker will follow, with the default Python version generally trailing the latest stable version of Python itself by 1-2 versions.
If you choose not to specify a Python version, your repository is guaranteed to stop working, eventually. We strongly recommend specifying a Python version (in environment.yml, runtime.txt, Pipfile, etc.)
https://repo2docker.readthedocs.io/en/latest/howto/languages/#supported-versions-of-python
I think we could do the same for R? If a repo is going to break due to a change in the unpinned R version it's better for it to happen sooner when there's a chance the author is still around.
Alternative options
Don't upgrade for now and keep the same default.
Who would use this feature?
R users who haven't pinned a version.
How much effort will adding it take?
Bumping the default version is minimal. Updating the docs to match the Python approach, if we decide to go that route, is a bit more.
Opening this as a separate issue, as suggested by @manics in #1567 (comment), to continue the discussion.
As @wildintellect mentioned in that thread:
The current pinned version is now close to 2 years old.
Proposed change
Upgrade the default R version to something newer in
repo2docker/repo2docker/buildpacks/r.py
Line 78 in 808ae6b
@manics suggested applying the same policy we use for Python:
Alternative options
Don't upgrade for now and keep the same default.
Who would use this feature?
R users who haven't pinned a version.
How much effort will adding it take?
Bumping the default version is minimal. Updating the docs to match the Python approach, if we decide to go that route, is a bit more.