[Bug]: SET statement fails when a config property name contains a hyphen

[iGarym]

Version

v4.5.0

Which SQL?

Hive

SQL content

set tez.grouping.max-size = 268435456;
set tez.grouping.min-size = 134217728;
set tez.grouping.split-waves = 1;

JavaScript/TypeScript code

import { HiveSQL, SparkSQL } from 'dt-sql-parser';

const sql = `set tez.grouping.max-size=1;
set tez.grouping.min-size=1;
set tez.grouping.split-waves = 1;`;

console.log(new HiveSQL().validate(sql));
// [
//   { startLine: 1, endLine: 1, startColumn: 21, endColumn: 22,
//     message: "'-' is not valid at this position, expecting a keyword" },
//   { startLine: 2, endLine: 2, startColumn: 21, endColumn: 22,
//     message: "'-' is not valid at this position" },
//   { startLine: 3, endLine: 3, startColumn: 23, endColumn: 24,
//     message: "'-' is not valid at this position" }
// ]

console.log(new SparkSQL().validate(sql));
// []

What happened?

Hive config property names are allowed to contain hyphens, and the statements
above run fine in both Hive and Spark. SparkSQL.validate() accepts them as
well, but HiveSQL.validate() reports a syntax error on every one of them:

line	message
1	'-' is not valid at this position, expecting a keyword
2	'-' is not valid at this position
3	'-' is not valid at this position

Relevant log output

[liuxy0551]

Regarding this issue, here are two pieces of information:

1. For the syntax "set tez.grouping.max-size = 1;", the dt-sql-parser Hive parser reports an error. However, if you use backticks to enclose the parameter name like "set tez.grouping.`max-size` = 1;", the parser does not report an error and passes validation successfully.

2. I checked the Hive official configuration documentation at https://cwiki.apache.org/confluence/display/Hive/Configuration+Properties, and among the configuration properties listed on this page, I did not see any that contain a "-" hyphen.

[iGarym]

Regarding this issue, here are two pieces of information:

1. For the syntax "set tez.grouping.max-size = 1;", the dt-sql-parser Hive parser reports an error. However, if you use backticks to enclose the parameter name like "set tez.grouping.max-size = 1;", the parser does not report an error and passes validation successfully.
2. I checked the Hive official configuration documentation at https://cwiki.apache.org/confluence/display/Hive/Configuration+Properties, and among the configuration properties listed on this page, I did not see any that contain a "-" hyphen.

The page you checked documents Hive's own system configuration properties, whereas these three parameters belong to the underlying parameters of the Apache Tez engine. Even though they may not be listed under standard system parameters, they are still completely valid.

While the official Apache Tez documentation does not provide a comprehensive list of all configuration parameters, various cloud providers offer detailed explanations, for example:

• AWS EMR Serverless User Guide
• Cloudera Technical Blog on Optimizing Hive on Tez Performance
• Azure HDInsight Documentation

Would it be possible to add a more general rule to support configuration property names in the xxx-xxx format? For example, something like this:

// src/grammar/hive/HiveSqlParser.g4#L89
- | KW_SET configPropertiesItem ((DOT | COLON) configPropertiesItem)* EQUAL .*?
+ | KW_SET configProperty EQUAL .*?

+ configProperty
+     : configPropertyPart ((DOT | COLON) configPropertyPart)*
+     ;

+ configPropertyPart
+     : configPropertiesItem (
+         {this.isNextTokenAdjacent()}? MINUS {this.isNextTokenAdjacent()}? + configPropertiesItem
+     )*
+     ;

// src/lib/SQLParserBase.ts

+ public isNextTokenAdjacent() {
+     const previousToken = this.tokenStream.LT(-1);
+     const nextToken = this.tokenStream.LT(1);
+     return previousToken !== null &&
+         nextToken !== null &&
+         previousToken.stop + 1 === nextToken.start;
+ }

The isNextTokenAdjacent method is introduced here to prevent interference from spaces and ensure that statements with invalid spacing like set xxx.a - b are correctly caught as syntax errors.

This depends on whether dt-sql-parser intends to support a broader range of configuration keys rather than being limited to Hive's own system configurations. If you're open to this approach, I'd be happy to submit a PR.

As a side note, I feel that the wildcard implementation for the SET statement in the current Spark parser is a bit too broad.