[ML-DataFrame] fix wire serialization issues in data frame response objects #39790
Conversation
|
Pinging @elastic/ml-core |
There was a problem hiding this comment.
LGTM
One more thing that falls into the category of serialization problems but could be left to a followup PR (or could be added to this PR) is that all action Request classes in this plugin have public constructors that take no arguments and construct an object that violates the invariants expected by serialization. They're only called by the RequestBuilders, so I think these Request() constructors should be removed and the constructors of the RequestBuilders should be changed to force the user to pass the minimum arguments required to make the Requests valid. But, like I said, this can be done in a separate PR if you prefer.
|
It's not as easy at it looks, I decreased accessibility where it was straight-forward. For other cases I run into problems. The code seems to follow common patterns, I see the same "issue" elsewhere - which does not mean it shouldn't be fixed - however it requires further thought. I am always open for simplifications, it pays off in the long run. So the rest can be fixed in a follow-up PR (e.g. I am unsure if we really need all the |
…bjects (#39790) fix wire serialization issues in data frame response objects
* elastic/master: Add pre-upgrade check to test cluster routing allocation is enabled (elastic#39340) Update logstash-management.json to use typeless template (elastic#38653) Small simplifications to mapping validation. (elastic#39777) Update distribution build instructions to reflect file names with OS/architecture classifiers. (elastic#39762) Give jspawnhelper execute permissions in bundled JDK (elastic#39787) Maintain step order for ILM trace logging (elastic#39522) [ML-DataFrame] fix wire serialization issues in data frame response objects (elastic#39790) fix index refresh in test within 20_mix_typeless_typeful (elastic#39198) Combine overriddenOps and skippedOps in translog (elastic#39771)
fixes wire serialization issues in data frame response objects uncovered on a multi node setup