[
https://issues.apache.org/jira/browse/CALCITE-6508?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17870012#comment-17870012
]
Mihai Budiu commented on CALCITE-6508:
--------------------------------------
The parser grammar looks like this:
{code}
<ARRAY> { s = span(); }
(
(
// nullary array function call: "array()" (Apache Spark)
LOOKAHEAD(2)
<LPAREN> <RPAREN> { args = SqlNodeList.EMPTY; }
|
args = ParenthesizedQueryOrCommaList(ExprContext.ACCEPT_ALL)
)
{
if (args.size() == 1 && args.get(0).isA(SqlKind.QUERY)) {
// Array query constructor, 'ARRAY (SELECT * FROM t)'
return SqlStdOperatorTable.ARRAY_QUERY.createCall(s.end(this),
args.get(0));
} else {
// Spark ARRAY function, 'ARRAY(1, 2)',
// equivalent to standard 'ARRAY [1, 2]'
return SqlLibraryOperators.ARRAY.createCall(s.end(this),
args.getList());
}
}
|
// by enumeration "ARRAY[e0, e1, ..., eN]"
<LBRACKET> // TODO: do trigraph as well ??( ??)
(
args = ExpressionCommaList(s, ExprContext.ACCEPT_NON_QUERY)
|
{ args = SqlNodeList.EMPTY; }
)
<RBRACKET>
{
return SqlStdOperatorTable.ARRAY_VALUE_CONSTRUCTOR.createCall(
s.end(this), args.getList());
}
{code}
So it looks like the form with () should accept queries, while the form with []
does not.
Maybe it's enough to change this in the grammar?
> Parse error when using scalar sub-query as operant to Array constructor
> function
> --------------------------------------------------------------------------------
>
> Key: CALCITE-6508
> URL: https://issues.apache.org/jira/browse/CALCITE-6508
> Project: Calcite
> Issue Type: Bug
> Reporter: Ian Bertolacci
> Priority: Major
>
> Using a scalar subquery in the {{`ARRAY[ ... ]`}} function throws "parse
> failed: Query expression encountered in illegal context"
> (Calcite version 1.36)
> For example:
> {code:sql}
> select ARRAY[ (select 1) ]
> select ARRAY[ (select max(x) from table) ]
> {code}
> Is this something that can be configured or easily altered?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
