zookeeper-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [zookeeper] szepet opened a new pull request #959: ZOOKEEPER-3402: Add multiRead operation
Date Sun, 26 May 2019 23:59:49 GMT
szepet opened a new pull request #959: ZOOKEEPER-3402: Add multiRead operation
URL: https://github.com/apache/zookeeper/pull/959
 
 
   So, the possibility of the multi version of getChildren, getData, and a common multiRead
interface was already mentioned and briefly discussed in the conversation thread of #922.
   This patch introduces the concept of multiRead operation.
   Note: This is a 'work-in-progress', not every comment is added, and test cases should be
extended as well. Currently, the aim is to decide whether this direction is something that
the community would get behind or should fin.
   I've faced more design decision while implementing this (note: I wanted to keep the backward
compatibility at all cost), the more important ones were the following:
   1. Whether the new read operations should be part of Op or should we create new descendant
classes of `Op` like `ReadOp` and `WriteOp`. In this case, I believe it would be unnecessary
to create new classes since it would create a lot of code duplication (despite the common
ancestor) and the flow of the processing would not be as clean as in the initial state. However,
by simply adding an OpKind enum to the class enables us to ensure the type-safety behavior
(read and write operations should not mix) and still preserve the clean processing flow.
   2. At which point of request processing should we capture the mixed (transaction and read
operation) multi requests? It seems pretty clear to catch them client side and not create
extra work for the server.
   3. Do we even need a `multiRead` operation or this whole improvement could be just part
of multi? Well, the main problem is the following: currently, on server side (and in general,
everywhere in ZooKeeper) the fact that multi only contains transactions is a heavily used
and built upon it. The whole `multi` processing flow should be rewritten from zero and also
the server would get a significant extra work for that. So yeah, it seems pretty clear to
me that introducing the `multiRead` operation is beneficial.
   
   Hope these thoughts make sense! Any observations, questions are welcome!

----------------------------------------------------------------
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.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

Mime
View raw message