[
https://issues.apache.org/jira/browse/IGNITE-21384?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815349#comment-17815349
]
Andrey Mashenkov commented on IGNITE-21384:
-------------------------------------------
The results above says that Postgres evaluated function at some point (most
likely at ALTER command call) and applied to existed rows.
It doesn't matter if these rows were changed in-place or not (I guess PG don't
do that), because our approach with "versioned schema" allow us to postpone
such actions to achieve the same result.
We just need a fix to evaluate expression and store a value that will be
applied when "row upgrading" happens.
> Initialize all rows with new column's default value
> ---------------------------------------------------
>
> Key: IGNITE-21384
> URL: https://issues.apache.org/jira/browse/IGNITE-21384
> Project: Ignite
> Issue Type: New Feature
> Reporter: Maksim Myskov
> Priority: Major
> Labels: ignite-3
>
> In case of adding a new column with a default value specified (for example:
> ALTER TABLE ADD COLUMN col INT DEFAULT 10), all existing rows (that were
> added before "alter table") must be initialized with a new value.
> h3. Current behaviour
> The default value of a column is evaluated on read.
> h3. Motivation
> This is a pre-requisite for adding arbitrary expressions as default values.
> Let's imagine a simple case:
> {code:sql}
> ALTER TABLE ADD COLUMN modified TIMESTAMP DEFAULT CURRENT_TIMESTAMP
> {code}
> With the current behavior for all existing rows the value of "modified"
> column will be evaluated on every read.
> Many DBs use the proposed behavior:
> * PostgreSQL
> * CockroachDB
> * Yugabyte
> For example, from PostgreSQL
> [docs|https://www.postgresql.org/docs/current/sql-altertable.html#SQL-ALTERTABLE-NOTES]:
> {noformat}
> Adding a column with a volatile DEFAULT or changing the type of an existing
> column will require the entire table and its indexes to be rewritten.
> {noformat}
> h3. Downsides
> Adding a new column with a default value will be a much more expensive
> operation and can take a significant time on large tables
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
