[ 
https://issues.apache.org/jira/browse/IGNITE-13670?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Andrey Mashenkov updated IGNITE-13670:
--------------------------------------
    Description: 
h3. Motivation.
The nullmap is currently always written to the tuple layout for all columns 
even if there are no nullable columns described in the schema.
The same is true and can be done for varlen table.

Seems, we can extend this idea to every single tuple and still have the ability 
to compare key/value content fast as byte arrays.
Apparently, this will work for rows of same schema version, but we shouldn't 
bother about the schema version,
because anyway, old row will be upgraded to the newer version before comparison 
according to the live-schema concept.

h3. Description.
Let's skip an empty nullmap and write just a flag instead.
Let's skip an empty varlen table and write just a flag instead.



  was:
The nullmap is currently always written to the tuple layout for all columns 
(even if there are no nullable columns). This data structure can be further 
used to encode default values for non-nullable columns (either a user-specified 
value or 0 for primitives).
The bits will still be left unused for non-null non-primitive types (UUID, 
String, byte[], etc).
Also, need to add support for skipping writing nullmaps (use the flags bit).


> Skip writing null-map and varlen table when possible
> ----------------------------------------------------
>
>                 Key: IGNITE-13670
>                 URL: https://issues.apache.org/jira/browse/IGNITE-13670
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Alexey Goncharuk
>            Priority: Major
>              Labels: iep-54, ignite-3
>
> h3. Motivation.
> The nullmap is currently always written to the tuple layout for all columns 
> even if there are no nullable columns described in the schema.
> The same is true and can be done for varlen table.
> Seems, we can extend this idea to every single tuple and still have the 
> ability to compare key/value content fast as byte arrays.
> Apparently, this will work for rows of same schema version, but we shouldn't 
> bother about the schema version,
> because anyway, old row will be upgraded to the newer version before 
> comparison according to the live-schema concept.
> h3. Description.
> Let's skip an empty nullmap and write just a flag instead.
> Let's skip an empty varlen table and write just a flag instead.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to