potiuk commented on code in PR #35612:
URL: https://github.com/apache/airflow/pull/35612#discussion_r1394948598


##########
setup.cfg:
##########
@@ -163,6 +163,7 @@ install_requires =
     # See https://github.com/apache/airflow/pull/31693
     # We should also remove "licenses/LICENSE-unicodecsv.txt" file when we 
remove this dependency
     unicodecsv>=0.14.1
+    universal_pathlib>=0.1.4

Review Comment:
   > We could argue we want to pin it specifically to this version to be more 
predictable. The downside of that is that upath might pin a particular 
dependency (e.g. fsspec) in the future (it doesn't now) to something we do not 
like.
   
   I think the risk is low and I would not worry about this prematurely.
   
   They (properly like all libraries should) have >= on fsspec. Also we have to 
remember that there are two dimensions of it. If they currently not pin, but 
pin in the future, most likely `pip` will resolve it to the current version - 
not the pinned one if it caused a conflict. As long as there are no security 
fixes we have to update to, or new features - we are good with staying with the 
current version (and that one has no upper-binding on fsspec). And fsspec is 
their only non-dev dependency, so it's really nice and lean library currently.
   
   Also with small libraries like that we can **always** vendor them in - even 
temporarily, if they are holding us back., We've done that few times. Upath is 
~ 1000 lines of code + ~ 40 lines "plugin" per-each-implementation. Some of our 
PRs are bigger (some of them mine :D ). 
   
   So I would not really worry about what **might** happen - we have options to 
follow. 
   
   > But it is v0.1.4 which sounds like very fesh and instable API
   
   I don't think that fresh :) . They have 27 (!) releases, 16 contributors and 
(maybe low but steady stream of contributions - about right when you look at 
this kind of - small - library. Some of the releases at the beginning were 
0.0.* - and likely those were the "fresh". Also I think there is no "universal" 
notion of 0. being "immature". Just checked and ~150 out of 600 of our 
dependencies (25%) has 0.* version :). 



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to