Resolve backslash escapes in StringValue.getNotExcapedValue()

[minleejae]

Purpose

Make StringValue.getNotExcapedValue() return the characters a string literal denotes. Today it only collapses '', so:

• A MySQL column default DEFAULT 'it\'s' yields it\'s.
• 'a\\b' yields a\\b.
• PostgreSQL E'it\'s' keeps its backslash.
• The double-quoted "it""s" keeps both quotes.

Current design (As-is)

Whether a backslash is an escape depends on the parser configuration: backslash escapes, or a PostgreSQL E'' string. The StringValue node does not record it, so getNotExcapedValue() cannot resolve backslash escapes. It also always collapses single quotes, whatever the literal's quote character.

Target design (To-be)

• Add StringValue.Escaping (STANDARD, MYSQL_BACKSLASH, POSTGRESQL_ESCAPE_STRING) with a getter, setter and with method.

• The parser records the escaping of every literal it creates from a token:

  • MYSQL_BACKSLASH: no dialect, or MySQL/MariaDB, with backslash escapes enabled.
  • POSTGRESQL_ESCAPE_STRING: PostgreSQL E'' strings, or any literal with backslash escapes enabled.
  • STANDARD: everything else.

• getNotExcapedValue() resolves the literal's own doubled quote and, depending on the escaping:

  • MySQL escapes: \0 \b \n \r \t \Z; \% and \_ keep their backslash; any other escaped character stands for itself.
  • PostgreSQL escapes: \b \f \n \r \t, octal, \x, \u and \U. Octal and hexadecimal escapes are bytes, decoded as UTF-8.

  It still runs in linear time.

• These keep the standard reading: constructed values, dollar-quoted strings, and the other backslash dialects (BigQuery, Databricks, Snowflake). getValue(), toString() and equals() are unchanged.

• Document the API in the usage guide.

Validation

• 9 new tests cover:
  • MySQL and MariaDB escapes, including \0, \Z, \%, \_ and N'';
  • doubled quotes per quote character;
  • the default configuration with backslash escapes;
  • PostgreSQL standard, E'' and dollar-quoted strings, including UTF-8 byte sequences;
  • the other backslash dialects;
  • constructed values;
  • an ALTER TABLE ... DEFAULT column option;
  • linear time on a 1,000,000-escape literal.
• Java 17 offline Gradle spotlessApply check: 10,103 tests, 10,078 passed, 25 skipped, zero failures/errors. The grammar choice conflict check passes.
• Compared getNotExcapedValue() with the bytes each server produces (HEX(...) / convert_to(...)) for 35 literals on MySQL 8.4, MySQL 8.0, MariaDB 11.4 and PostgreSQL 18.3. All are identical.

This PR is independent. A literal that mixes \' and '' additionally needs the lexer fix in #2790 to be tokenized as one string.

Integration validation

Applied #2788, #2789, #2790, #2791 and #2792 together on master (166f0a27):

• Every pair merges without conflicts, and the combined grammar is free of choice conflicts.
• The integrated build passes Java 17 offline Gradle spotlessApply check: 10,150 tests, 10,125 passed, 25 skipped, zero failures/errors.
• With the integrated build, the same server results hold on MySQL 8.4, MySQL 8.0, MariaDB 11.4 and PostgreSQL 18.3:
  • the 95 statement cases of these PRs;
  • the 11 default-mode drop cases;
  • 51 string literal comparisons, including 16 mixed-escape literals that need Lex MySQL backslash escapes consistently in string literals #2790 and Resolve backslash escapes in StringValue.getNotExcapedValue() #2791 together.

[manticore-projects]

Thank you!