Simplify checks for package versions

[potiuk]

Replaces more complex package version checks with one-liners.

---

^ Add meaningful description above
Read the Pull Request Guidelines for more information.
In case of fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
In case of a new dependency, check compliance with the ASF 3rd Party License Policy.
In case of backwards incompatible changes please leave a note in a newsfragment file, named {pr_number}.significant.rst or {issue_number}.significant.rst, in newsfragments.

[potiuk]

cc: @Dev-iL

[Dev-iL]

Thanks for notifying me! I've left some questions/suggestions.

While you're at it, perhaps also modify the PENDULUM3 definitions in

• airflow/utils/timezone.py
• tests/serialization/serializers/test_serializers.py
  ?

[Dev-iL]

In #37575, @Taragolis used from packaging.version import Version instead of parse, presumably because parse is just an alias for Version,

defparse(version: str) ->"Version":
"""Parse the given version string. >>> parse('1.0.dev1') <Version('1.0.dev1')> :param version: The version string to parse. :raises InvalidVersion: When the version string is not a valid version. """returnVersion(version)

Perhaps we should agree on a uniform way to import this?

[Dev-iL]

Build failed due to the patching path not being updated in the sqlalchemy tests - should be changed something like this:

def test_is_sqlalchemy_v1(major_version, expected_result):
- with mock.patch("airflow.utils.sqlalchemy._get_lib_major_version") as mock_get_major_version:+ with mock.patch("airflow.utils.sqlalchemy.parse_version") as mock_get_major_version:

And also entirely remove test_get_lib_major_version.

[potiuk]

While you're at it, perhaps also modify the PENDULUM3 definitions in

airflow/utils/timezone.py
tests/serialization/serializers/test_serializers.py

Ah. Nice. I missed those, yes, good idea to fix them there - we had recently a long discussion about using __version__ with @Taragolis#37320 (comment) and we came to the conclusion is that we should indeed switch to using importlib/metadata instead as the only "standard" way, I even have a draft PR to change the approach we use in our providers #37335 as result of that (waited for the release of providers that just happened so I might revive it).

[potiuk]

And also entirely remove test_get_lib_major_version.

Yep. This is why we have CI to tell us so.

[potiuk]

Perhaps we should agree on a uniform way to import this?

perhaps. you seem to be good at chasing all of those. Maybe a separate PR proposing such unification. This is best way here to get things done in the way you think things shoudl be done - just doing it.

[potiuk]

Right. Shoudl be nicer. simpler, more consistent and less code everywhere

[Taragolis]

I also think about some caching mechanism, we parse all distributions which installed into the environment:

airflow/airflow/utils/entry_points.py

Lines 35 to 43 in ca3ce78

	def_get_grouped_entry_points() ->dict[str, list[EPnD]]:
	mapping: dict[str, list[EPnD]] =defaultdict(list)
	fordistinmetadata.distributions():
	try:
	foreindist.entry_points:
	mapping[e.group].append((e, dist))
	exceptExceptionase:
	log.warning("Error when retrieving package metadata (skipping it): %s, %s", dist, e)
	returnmapping

Distribution have the information about version also. So maybe it is a good point to have generalised util around importlib.metadata and packaging for retrieve and parse this information in reusable format, e.g.

from __future__ importannotationsimportsysfromtypingimportNamedTuplefrompackaging.versionimportVersionifsys.version_info>= (3, 10):
fromimportlibimportmetadataelse:
# The “selectable” entry points were introduced in importlib_metadata 3.6 and Python 3.10.# Prior to those changes, entry_points accepted no parameters and always returned a dictionary of entry points,# keyed by group. With importlib_metadata 5.0 and Python 3.12, entry_points always returns an EntryPoints object.# See backports.entry_points_selectable for compatibility options.# source: https://docs.python.org/3/library/importlib.metadata.htmlimportimportlib_metadataasmetadataclassDistVersion(NamedTuple):
distribution_name: strversion: strversion_info: Versionbase_version_tuple: tuple[int, int, int]
@classmethoddeffrom_distribution(cls, dist: metadata.Distribution) ->DistVersion:
vi=Version(dist.version)
returnDistVersion(
distribution_name=dist.name,
version=dist.version,
version_info=vi,
base_version_tuple=(vi.major, vi.minor, vi.micro)
)
airflow_dist=metadata.distribution("apache-airflow")
dist_version=DistVersion.from_distribution(airflow_dist)
print(dist_version)
print(dist_version.base_version_tuple>= (2, 8))
print(dist_version.base_version_tuple>= (2, 9))
ifdist_version.base_version_tuple< (2, ):
raiseRuntimeError("¯\_(ツ)_/¯")

That is idea for potential future improvements, not for particular this PR

[potiuk]

That is idea for potential future improvements, not for particular this PR

Yep. If it can give measurable improvements - THAT could be the reason to introduce a common uitl (just retrieving it is IMHO not enough to extract common code.

[Taragolis]

I guess it is more common in general if compare to one time usage in the Provider Manager.

• We check different versions into the core, maybe even into the providers
• Providers collected information about different providers version.
• Providers collected information about itself (for tracing purpose I guess)

However it is required more time to introduce common interface for both core and providers, because if it will allow to use into the providers, than we cant change it.

[potiuk]

However it is required more time to introduce common interface for both core and providers, because if it will allow to use into the providers, than we cant change it.

Yep. Unless/untll we introduce common.utils provider :). Which we've been toying the idea with.

[Taragolis]

The problem with airflow.utils it is not covered by Public Interface (except TaskGroups), however if Community Provider use it, it might prevent to changing it. So maybe we need to move some internal utilities to the internal sub-package all common stuff keep in the separate package, and deprecate all other stuff (something we might safely remove).

Presumably good point to future discussion.