[BUG] JSQLParser 5.4-SNAPSHOT : Hive : INSERT PARTITION bare column inherits the previous item's value

[fudianchn]

AI disclosure: this issue was prepared with AI coding agents, reviewed and revised line by line by me.

Failing SQL Feature:

In INSERT ... PARTITION, a bare (dynamic) partition column that follows an item with a value inherits that item's value instead of keeping null. The deparsed statement silently turns the dynamic column into a constant assignment, changing the statement's semantics.

Verified combinations (master 1e4e92b and the 5.3 release jar):

PARTITION spec	parsed items
(dtime = '20220403', region)	dtime = '20220403', region = '20220403' (wrong)
(a, b = '1', c)	a = null, b = '1', c = '1' (wrong)
(dtime, hour = '10')	dtime = null, hour = '10' (correct)
(dtime = '20220403', hour = '10')	both correct
(dtime)	dtime = null (correct)

The trigger is: an item without a value following an item with one. Affects both INSERT OVERWRITE TABLE ... PARTITION and INSERT INTO TABLE ... PARTITION (shared Insert() grammar).

Root cause in Partitions() (src/main/jjtree/net/sf/jsqlparser/parser/JSqlParserCC.jjt, around line 3924, introduced with #2323):

List<Partition> Partitions():
{
    List<Partition> partitions = new ArrayList<Partition>();
    Column tableColumn;
    Expression valueExpression = null;
}
{
    (
        tableColumn=Column() [ "=" valueExpression=Expression() ]
        { partitions.add( new Partition (tableColumn, valueExpression)); }
    )
    ( LOOKAHEAD(2) (
            ","
            tableColumn=Column() [ "=" valueExpression=Expression() ]
            { partitions.add( new Partition (tableColumn, valueExpression)); }
    ) )*
    ...

valueExpression is declared once for the whole production and reused for every item. When the optional "=" Expression() is skipped for a bare column, valueExpression still holds the value of an earlier item, which then lands in new Partition(tableColumn, valueExpression).

Suggested fix: reset valueExpression = null at the start of each item. Partition already renders a bare column when the value is null, so no deparser change should be needed.

SQL Example:

Insert insert = (Insert) CCJSqlParserUtil.parse(
        "INSERT OVERWRITE TABLE t PARTITION (dtime = '20220403', region) SELECT code, region FROM src");

for (Partition p : insert.getPartitions()) {
    System.out.println(p.getColumn() + " -> " + p.getValue());
}

Output on master:

dtime -> '20220403'
region -> '20220403'

Expected:

dtime -> '20220403'
region -> null

The deparsed statement is INSERT OVERWRITE TABLE t PARTITION (dtime = '20220403', region = '20220403') SELECT code, region FROM src, so region no longer takes its value dynamically from the SELECT list.

Software Information:

• JSqlParser version: master 1e4e92b, also verified on the 5.3 release jar (Partitions() exists since [feat/refactor] AllColumns and AllTableColumns-Support for JSON_OBJECT #2323; 5.0 does not parse these forms at all)
• Database: Hive (mixed static and dynamic partition columns, see "Dynamic Partition Inserts" in the Hive LanguageManual DML)