Following the comments to warn users about dropping Python 2, I went ahead and added the deprecation library to Airflow in this PR: https://github.com/apache/airflow/pull/4763, and added the first @deprecated annotation.
I checked a few major packages and all use some self-made annotation for showing deprecation warnings. I didn’t want to re-invent the wheel and used a tiny package for it. More details are in the PR itself. If you know a better way to do this, let me know :-) Bas On 17 Feb 2019, at 09:00, Bas Harenslak <basharens...@godatadriven.com<mailto:basharens...@godatadriven.com>> wrote: The PR comments say drop hdfs3 and use pyarrow, however this is not implemented yet. On 17 Feb 2019, at 08:42, zhong jiajie <zhongjiajie...@hotmail.com<mailto:zhongjiajie...@hotmail.com>> wrote: I overlook the PR https://github.com/apache/airflow/pull/3560 still use **hdfs3** rather than **PyArrow**. So which package we choice finilly? [https://avatars3.githubusercontent.com/u/307739?s=400&v=4]<https://github.com/apache/airflow/pull/3560> [AIRFLOW-2697] Drop snakebite in favour of hdfs3 by jrderuiter ・ Pull Request #3560 ・ apache/airflow ・ GitHub<https://github.com/apache/airflow/pull/3560> Commit #52a2ff3 changes the HdfsSensor classes for the new HdfsHook. The biggest change is that it rewrites the HdfsSensor + HdfsSensorFolder (from contrib) sensors into a HdfsFileSensor and a HdfsFolderSensor class, which aim to retain the functionality of the previous classes. github.com<http://github.com> ________________________________ One blocker is HDFS support. Right now we rely on Snakebite which is Python2 only, so we need to update this as well. There is an open ticket: https://jira.apache.org/jira/browse/AIRFLOW-2697 Op vr 15 feb. 2019 om 07:55 schreef zhong jiajie <zhongjiajie...@hotmail.com : Agree with XD's second idea, we should warn Airflow users. ________________________________ My two cents: 1. Can we discuss on the timeline for 2.0? Another big change is FAB-UI ONLY, which is already in master branch. Personally I think it’s already worth considering 2.0, unless there is other existing expectation for 2.0. 2. Before we completely drop Python2 support, we can start adding warning/reminder mentioning that it’s stronly recommended to use py3 & py2 drooping can happen “at any time” for Airflow (we can add these warnings in 1.10.3?) XD On Thu, Feb 14, 2019 at 18:42 Ash Berlin-Taylor <a...@apache.org> wrote: We talked about this in the past < http://mail-archives.apache.org/mod_mbox/airflow-dev/201808.mbox/%3CCAK%2Be0oujpY1A0mD0G775CmmRneVDjQMTi99aqK%2BA_C07ZW8ySA%40mail.gmail.com%3E but I'd like to bring it up again with a more concrete proposal. At the time Airflow 2.0 wasn't very concrete, but now with a lot of otherwise breaking changes already on master how about we use that time to also drop support for Py2? (Reminder: Python 2 is reaching End of Life in January 1, 2020 and will receive zero updates, even security ones past this date. At least not by the Python team. I imagine RedHat will still be running Python 2.6 in 2120 ;) ) -ash
