[
https://issues.apache.org/jira/browse/FLINK-25874?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Huang Xingbo closed FLINK-25874.
--------------------------------
Resolution: Not A Problem
This is not about the incompatibility of pyflink dependencies, but that there
is no way for Python dependency management to rely on the version of shade
itself like Java dependency management. Of course, if you want to upgrade the
python-dateutil and cloudpickle versions of pyflink, you can contribute PR to
upgrade them as [~nicholasjiang] said. On the other hand, if you use pip for
installation, even if the dependency conflict will not affect your installation
of dependencies. You can install apache-flink first, and then install your
other dependency packages. Based on these, I will close the JIRA first.
> PyFlink package dependencies conflict
> -------------------------------------
>
> Key: FLINK-25874
> URL: https://issues.apache.org/jira/browse/FLINK-25874
> Project: Flink
> Issue Type: Improvement
> Reporter: Martin Liu
> Priority: Blocker
>
> I need to install PyFlink with some other libraries in a project, and I found
> there's dependency conflict with `great-expectations` because PyFlink has
> pined dependency `python-dateutil==2.8.0`.
> {code:java}
> There are incompatible versions in the resolved dependencies:
> python-dateutil==2.8.0 (from apache-flink==1.14.3->-r requirements.in (line
> 4))
> python-dateutil>=2.8.1 (from great-expectations==0.14.4->-r requirements.in
> (line 5)) {code}
> I have to use newer version of great-expectations (>=0.13.1) for some
> features, so this is blocking me.
> I found `python-dateutil` v2.8.0 is released in 2019-02-05, see
> [https://github.com/dateutil/dateutil/releases/tag/2.8.0], is there any way
> to loose the dependency, e.g. >=2.8.0 ?
> (BTW, `cloudpickle==1.2.2` is also old which released in 2019-09-10, see
> [https://github.com/cloudpipe/cloudpickle/releases/tag/v1.2.2] )
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
