[
https://issues.apache.org/jira/browse/CALCITE-2674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16718402#comment-16718402
]
Yuzhao Chen commented on CALCITE-2674:
--------------------------------------
[~julianhyde] Now we parse escape character through JavaCC token manager, the
only snippet we can have info about whether the identifier is escaped is in
Parser.jj#Identifier(), this method is invoked by many parse logic.
If we do not add flag in the Parser.jj#Identifier() return type and just t
return a string id, we will never know when and whether the identifier is
escaped at all.
For Parser.jj#CompoundIdentifier() i add compound id list length check:
{code:java}
if (list.size() == 1) {
return new SqlIdentifier(list, null, pos, posList, p.right);
} else {
return new SqlIdentifier(list, null, pos, posList);
}
{code}
If the list length is 1 (means that it is a simple identifier), we will mark
the SqlIdentifier as if it is escaped, else we will force make the whole
compound identifier as not escaped.
look forward to your suggestions, i will add the Oracle test cases to the UT.
> Column name with escape character should not be verified as a function when
> the column name is same with it
> -----------------------------------------------------------------------------------------------------------
>
> Key: CALCITE-2674
> URL: https://issues.apache.org/jira/browse/CALCITE-2674
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.18.0
> Reporter: Yuzhao Chen
> Assignee: Julian Hyde
> Priority: Major
> Fix For: 1.18.0
>
>
> Now If user defines table as schema below:
> {code:java}
> create table tt(
> `current_time` int,
> b int);
> {code}
> Then start a query:
> {code:java}
> select `current_time` from tt;
> {code}
> Calcite parser will parse the *current_time* as a normal *SqlIdentifier*,
> then the SqlValidator will recognize it as a builtin function with below code
> snippet:
> {code:java}
> public static SqlCall makeCall(
> SqlOperatorTable opTab,
> SqlIdentifier id) {
> if (id.names.size() == 1) {
> final List<SqlOperator> list = new ArrayList<>();
> opTab.lookupOperatorOverloads(id, null, SqlSyntax.FUNCTION, list);
> for (SqlOperator operator : list) {
> if (operator.getSyntax() == SqlSyntax.FUNCTION_ID) {
> // Even though this looks like an identifier, it is a
> // actually a call to a function. Construct a fake
> // call to this function, so we can use the regular
> // operator validation.
> return new SqlBasicCall(
> operator,
> SqlNode.EMPTY_ARRAY,
> id.getParserPosition(),
> true,
> null);
> }
> }
> }
> return null;
> }{code}
> While i tried MYSQL and such query can work properly, so is this a bug ?
> Cause if the matched function's return type is same with the column, then the
> query will output wrong results.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
