[
https://issues.apache.org/jira/browse/IGNITE-23461?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-23461:
---------------------------------------
Description:
VarIntUtils should be used instead as NaiveVarInts only give a tiny advantage
in rare cases and it's not worth it to keep both implementations.
Also, the following improvements are to be done:
# Use predictable UUIDs in tests. Without this, if we use random UUIDs to
generate objects that will later be serialized and encoded as BASE64, we lose
knowledge about 'what UUIDs we used'. If, later, the binary format gets changed
in an incompatible way (which might happen before 3.0 gets released), we'll
have to recreate the BASE64 encodings in tests, and we'll have to use other
UUIDs (as we did not memorize the original ones), which will make us also
change assertions (that verify those UUIDs).
# Use UUIDs without repetitive bytes (for instance, don't use UUIDs with half
of their bytes equal to 0). This will allow to catch ser/deser errors when
deserializer swaps bytes erroneously.
was:VarIntUtils should be used instead as NaiveVarInts only give a tiny
advantage in rare cases and it's not worth it to keep both implementations
> Remove NaiveVarInts
> -------------------
>
> Key: IGNITE-23461
> URL: https://issues.apache.org/jira/browse/IGNITE-23461
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> VarIntUtils should be used instead as NaiveVarInts only give a tiny advantage
> in rare cases and it's not worth it to keep both implementations.
> Also, the following improvements are to be done:
> # Use predictable UUIDs in tests. Without this, if we use random UUIDs to
> generate objects that will later be serialized and encoded as BASE64, we lose
> knowledge about 'what UUIDs we used'. If, later, the binary format gets
> changed in an incompatible way (which might happen before 3.0 gets released),
> we'll have to recreate the BASE64 encodings in tests, and we'll have to use
> other UUIDs (as we did not memorize the original ones), which will make us
> also change assertions (that verify those UUIDs).
> # Use UUIDs without repetitive bytes (for instance, don't use UUIDs with
> half of their bytes equal to 0). This will allow to catch ser/deser errors
> when deserializer swaps bytes erroneously.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
