[
https://issues.apache.org/jira/browse/BEAM-10631?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gleb Kanterov updated BEAM-10631:
---------------------------------
Description:
Schema#indexOf is broken between 2.21.0 and 2.23.0 (can't tag 2.23.0 yet).
The regression was introduced in
[apache/beam#10413|https://github.com/apache/beam/pull/10413/files#r464824602].
{code:diff}
- if (index == null) {
- throw new IllegalArgumentException(
- String.format("Cannot find field %s in schema %s", fieldName, this));
- }
+ Preconditions.checkArgument(
+ index != null, String.format("Cannot find field %s in schema %s",
fieldName, this));
{code}
`Schema.indexOf` is on a critical path, and calling `this.toString()` was
allocating a lot of small objects, that scaled badly with a number of fields
and options in schema. That wasn't significantly noticeable in benchmarks that
used schemas with few fields.
was:Schema#indexOf is broken between 2.19.0 and 2.22.0
> Performance of Schema#indexOf is broken
> ---------------------------------------
>
> Key: BEAM-10631
> URL: https://issues.apache.org/jira/browse/BEAM-10631
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-core
> Affects Versions: 2.21.0, 2.22.0
> Reporter: Gleb Kanterov
> Assignee: Gleb Kanterov
> Priority: P1
> Fix For: 2.24.0
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Schema#indexOf is broken between 2.21.0 and 2.23.0 (can't tag 2.23.0 yet).
> The regression was introduced in
> [apache/beam#10413|https://github.com/apache/beam/pull/10413/files#r464824602].
>
> {code:diff}
> - if (index == null) {
> - throw new IllegalArgumentException(
> - String.format("Cannot find field %s in schema %s", fieldName,
> this));
> - }
> + Preconditions.checkArgument(
> + index != null, String.format("Cannot find field %s in schema %s",
> fieldName, this));
> {code}
> `Schema.indexOf` is on a critical path, and calling `this.toString()` was
> allocating a lot of small objects, that scaled badly with a number of fields
> and options in schema. That wasn't significantly noticeable in benchmarks
> that used schemas with few fields.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
