[core] Fix Identifier should parse backquote.

[LinMingQiang]

When using flink procedure, table name can not parsed correctly when use `` quote.

But Spark request use `` for SYSTEM_TABLE_SPLITTER ($) . So this is incompatible with Spark.

e.g :

Flink is OK But Spark is ERR  :  CALL sys.create_branch('default.T$branch_branch1', ......)

Flink is ERR But Spark is OK :  CALL sys.create_branch('default.`T$branch_branch1`', ......)

Flink will parse `T$branch_branch1` to table is `T  and branch is branch_branch1`.

Why :

Flink use org.apache.paimon.catalog.Identifier to parse the tableName and miss processing backquote,
spark use spark.sessionState().sqlParser().parseMultipartIdentifier to parse.

Purpose

Linked issue: close #xxx

Tests

API and Format

Documentation

[liyubin117]

I doubt whether the feature is necessary, because Flink regards `example.view` as a whole table name differ from Spark. refer to https://nightlies.apache.org/flink/flink-docs-release-1.20/docs/dev/table/common/#expanding-table-identifiers
We should follow the corresponding syntax rules.

On the other hand, the feature cannot deal with legal default.`T`$branch_branch_A.

[LinMingQiang]

I doubt whether the feature is necessary, because Flink regards example.view as a whole table name differ from Spark. refer to https://nightlies.apache.org/flink/flink-docs-release-1.20/docs/dev/table/common/#expanding-table-identifiers We should follow the corresponding syntax rules.

On the other hand, the feature cannot deal with legal default.`T`$branch_branch_A.

Thanks for your comment. But i think this has nothing to do with flink.

CALL sys.create_branch('default.`table`$branch_branch1', ......) should be executed successfully in flink.

And i will fix this case default.`T`$branch_branch_A .

[liyubin117]

Thanks for your reply, we may need to consider the following scenairos.

[LinMingQiang]

Thanks for your reply, we may need to consider the following scenairos.

In your example, you create a table called table`#branched. This is no problem in Flink-SQL parsing engine, as the backticks are handled correctly. But it is not processed in procedure, so the correct table name table`#branched cannot be parsed. This is also the purpose of this PR, here we can't just do a simple replacement, we need a parser to parse the table name like SparkUtils#catalogAndIdentifier does .

[gmdfalk]

Thanks for your reply, we may need to consider the following scenairos.

In your example, you create a table called table#branched. This is no problem in Flink-SQL parsing engine, as the backticks are handled correctly. But it is not processed in procedure, so the correct table name table#branched cannot be parsed. This is also the purpose of this PR, here we can't just do a simple replacement, we need a parser to parse the table name like SparkUtils#catalogAndIdentifier does .

I believe you also cannot do a simple replacement like currently if you want to support any entity (catalog, database or table) with special characters in it.
For instance if the database name contains a dot:

CALL sys.compact(`table` => '`my_catalog`.`my_database.containing_dot`.my_table');

[LinMingQiang]

Thanks for your reply, we may need to consider the following scenairos.

In your example, you create a table called table#branched. This is no problem in Flink-SQL parsing engine, as the backticks are handled correctly. But it is not processed in procedure, so the correct table name table#branched cannot be parsed. This is also the purpose of this PR, here we can't just do a simple replacement, we need a parser to parse the table name like SparkUtils#catalogAndIdentifier does .

I believe you also cannot do a simple replacement like currently if you want to support any entity (catalog, database or table) with special characters in it. For instance if the database name contains a dot:

CALL sys.compact(`table` => '`my_catalog`.`my_database.containing_dot`.my_table');

Sure, @JingsongLi WDYT. Do we need to implement a parser to parse the table name like spark procedure do?