JoshRosen commented on PR #37480:
URL: https://github.com/apache/spark/pull/37480#issuecomment-1215496802

   > In [SPARK-39489](https://github.com/apache/spark/pull/36885), for `Why are 
the changes needed?`, I found some of the reasons are as follows:
   > 
   > ```
   > In addition, this is a stepping-stone towards eventually being able to 
remove our Json4s dependency:
   > 
   > Today Spark uses Json4s 3.x and this causes library conflicts for end 
users who want to upgrade to 4.x; see 
https://github.com/apache/spark/pull/33630 for one example.
   > To completely remove Json4s we'll need to update several other parts of 
Spark (including code used for ML model serialization); this PR is just a first 
step towards that goal if we decide to pursue it.
   > In this PR, I continue to use Json4s in test code; I think it's fine to 
keep Json4s as a test-only dependency.
   > ```
   > 
   > I'm not sure if @JoshRosen has plans for the next step and the overall 
blueprint for this. I'm just learning this 
[SPARK-39489](https://github.com/apache/spark/pull/36885) and trying to start 
with some simple cases. Similarly, I submitted another pr: #37515
   
   @LuciferYang, my change in https://github.com/apache/spark/pull/36885 was 
primarily motivated by History Server performance. Although the unblocking of 
Json4s removal is a nice secondary benefit, I don't think that removal is a 
super high priority. There's also some burdens in terms of testing to ensure 
that the old and new JSON is fully cross-compatible (in cases where it needs to 
be). Aside from performance-sensitive places, there's limited benefit from 
_partial_ removal of Json4s: I think the big user-facing benefits would be 
achieved only when users are free to use any version of Json4s because Spark 
drops its dependency. Therefore, I think we should do some more analysis to 
confirm that it's technically possible to fully remove Json4s (and that doing 
so is safe / desirable) before we start merging these smaller piece-by-piece 
removals.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to