New SQL parser using ANTLR (replacing JavaCC)

[lvca]

This is a huge task. We inherited the SQL engine from OrientDB. It was originally developed with JavaCC. Unfortunately JavaCC is poorly supported nowadays and has some issues with complex grammar (huge usage of LOOKAHEAD).

Since the new Cypher parser and lexer work pretty well, why not converting our old SQL parser with ANTLR latest version?

[lvca]

Created a global and database setting to switch between parser. By default from v26.1.1 will be the new ANTLR, but in case you can always switch back to the old JavaCC. We're going to remove JavaCC in the next releases, so please report any issues you have with the new parser before switching back JavaCC.

SQL_PARSER_IMPLEMENTATION("arcadedb.sql.parserImplementation", SCOPE.DATABASE,
 "SQL parser implementation to use. 'antlr' (default) uses the new ANTLR4-based parser with improved error messages. " +
 "'javacc' uses the legacy JavaCC-based parser for backward compatibility.",
 String.class, "antlr", Set.of("antlr", "javacc")),

[lvca]

First results are very good already:

==========================================================================================
RESULTS SUMMARY
==========================================================================================
Query Type                     │  JavaCC (µs) │   ANTLR (µs) │    Diff (µs) │     Winner
------------------------------─┼─------------─┼─------------─┼─------------─┼─----------
Simple SELECT                  │          150 │          144 │           -6 │      ANTLR
Complex SELECT (AND/OR)        │          504 │          212 │         -292 │      ANTLR
MATCH Query 1                  │          240 │          314 │          +74 │     JavaCC
MATCH Query 2                  │          223 │          438 │         +215 │     JavaCC
Mixed SQL (10 cmds)            │          576 │          197 │         -379 │      ANTLR
------------------------------─┼─------------─┼─------------─┼─------------─┼─----------
TOTAL                          │         1693 │         1305 │         -388 │      ANTLR
==========================================================================================
Overall: ANTLR is 22.9% faster than JavaCC
==========================================================================================

[lvca]

Latest improvements:

==========================================================================================
RESULTS SUMMARY
==========================================================================================
Query Type                     │  JavaCC (µs) │   ANTLR (µs) │    Diff (µs) │     Winner
------------------------------─┼─------------─┼─------------─┼─------------─┼─----------
Simple SELECT                  │          187 │           98 │          -89 │      ANTLR
Complex SELECT (AND/OR)        │          664 │          225 │         -439 │      ANTLR
SQL Many Parenthesis           │          222 │           36 │         -186 │      ANTLR
MATCH Query 1                  │          285 │          103 │         -182 │      ANTLR
MATCH Query 2                  │          335 │          119 │         -216 │      ANTLR
Mixed SQL (10 cmds)            │          898 │          158 │         -740 │      ANTLR
------------------------------─┼─------------─┼─------------─┼─------------─┼─----------
TOTAL                          │         2591 │          739 │        -1852 │      ANTLR
==========================================================================================
Overall: ANTLR is 71.5% faster than JavaCC
==========================================================================================