[
https://issues.apache.org/jira/browse/CASSANDRA-17878?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17632392#comment-17632392
]
Stefan Miklosovic edited comment on CASSANDRA-17878 at 11/11/22 2:38 PM:
-------------------------------------------------------------------------
Simplified patch committed into cassandra-4.0 and cassandra-4.1
https://github.com/instaclustr/cassandra/commit/2638051af186eb7cee110c103ae011f853ae8bb3
was (Author: smiklosovic):
Simplified patch committed to 4.0 and 4.1
https://github.com/instaclustr/cassandra/commit/2638051af186eb7cee110c103ae011f853ae8bb3
> Harden parsing of boolean values in CQL in PropertyDefinitions
> --------------------------------------------------------------
>
> Key: CASSANDRA-17878
> URL: https://issues.apache.org/jira/browse/CASSANDRA-17878
> Project: Cassandra
> Issue Type: Improvement
> Components: CQL/Interpreter, CQL/Semantics
> Reporter: Stefan Miklosovic
> Assignee: Stefan Miklosovic
> Priority: Normal
> Fix For: 4.0.8, 4.1-beta2, 4.2
>
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
> There is currently this in PropertyDefinitions class as a pattern we use for
> testing a boolean value in cqlsh
> {code}
> private static final Pattern PATTERN_POSITIVE =
> Pattern.compile("(1|true|yes)");
> {code}
> This might be source of mistakes and typos. For example, if a user does, for
> example:
> {code}
> ALTER TABLE ks.tb WITH cdc = tru;
> {code}
> If he does not notice it, he thinks that cdc is true, but it is not.
> More to it, currently, everything which is not "1", "true", or "yes" is
> evaluated as false. We should harden this in such a way that both logical
> true and false would be parsed only on well defined values and every other
> value would be rejected and a query would fail.
> EDIT: I have checked how it behaves in cqlsh and there seems to be validation
> of this already like this:
> {code}
> cqlsh> ALTER TABLE abc.def WITH cdc = tru;
> SyntaxException: line 1:31 no viable alternative at input 'tru' (ALTER TABLE
> abc.def WITH [cdc] =...)
> {code}
> It seems that cqlsh already knows this should be a boolean and rejects such
> query.
> Nevertheless, it is still reasonable to harden this on the code level when a
> query is executed in Java, programmatically (e.g. as part of tests or
> similar). The patch also includes optimizations to not return Boolean but
> boolean on related methods (other primitives are covered as well).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
