[Surajsuthar]

Add dialect-specific precedence for unary + and - so PostgreSQL
binds them more tightly than ^, while preserving generic-dialect behavior.

#2819

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.20%. Comparing base (bac1468) to head (e1774c6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2829   +/-   ##
=======================================
  Coverage   83.20%   83.20%           
=======================================
  Files          42       42           
  Lines       36471    36472    +1     
  Branches    36471    36472    +1     
=======================================
+ Hits        30345    30346    +1     
  Misses       2632     2632           
  Partials     3494     3494           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LucaCappelletti94]

SQLiteDialect::prec_value matches every variant, so it needs the new one too.

-            Precedence::MulDivModOp => 40,
+            Precedence::MulDivModOp | Precedence::UnaryPlusMinus => 40,

I have looked into whether this precedence bug applies also to the other dialects, and it does in some, so it will need to be addressed. DuckDB, MySQL and MariaDB all bind ^ tighter than * (2 * 3 ^ 2 is 18 in DuckDB 1.4.4, and 2 * 3 ^ 1 is 4 in MySQL 8.4 and MariaDB 11.8), while the DuckDB and MySQL dialects parse (2 * 3) ^ n. The DuckDB dialect also parses ^ as BitwiseXor, although it is exponentiation there. Those fixes are outside this PR and can follow separately once the new level lands, since their tables need UnaryPlusMinus above Caret. If you happen to have time, such PRs would be welcomed.

[LucaCappelletti94]

In a little while I will open a PR adding a fuzzer exactly aimed at finding these issues, do use it before the next review round as I think it could be useful. It unfortunately finds still several expression priority issues.

[LucaCappelletti94]

You could rename this seed to postgres_2829_unary_minus_caret.sql, since every other fuzz_seeds/<dialect>_<n>_*.sql name carries the number of the sqlparser PR that added it, and 27094 is the DataFusion issue.