[LucaCappelletti94]

SELECT t ._a FROM t failed to tokenize with Unexpected character '_', although SQLite 3.51.1, DuckDB 1.4.4 and PostgreSQL 16 all run it and read it as t._a. The tokenizer treats ._ as a period only when the previous token is a word, and with whitespace in between the previous token was the whitespace, so it fell through to number parsing.

The check now looks at the previous token that is not whitespace. A ._ that follows a number, as in SELECT 1 ._a, is still a tokenizer error, and SELECT ._abc with no qualifier is still rejected by the parser. Tests cover these forms in every dialect where _ can start an identifier.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 99.25373% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 81.79%. Comparing base (13ce307) to head (78224d9).

Files with missing lines	Patch %	Lines
src/tokenizer.rs	99.25%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2677      +/-   ##
==========================================
+ Coverage   81.72%   81.79%   +0.07%     
==========================================
  Files          42       42              
  Lines       34476    34583     +107     
  Branches    34476    34583     +107     
==========================================
+ Hits        28174    28286     +112     
+ Misses       2757     2755       -2     
+ Partials     3545     3542       -3     

☔ 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.