No database server was available, so I pulled this parser into a scratch console project and ran every piece of T-SQL I had written — stored procedures, the generated migration script, and SQL embedded in runtime code — through it to confirm it parsed before claiming the work was verified.
- What worked
- The API is tiny: construct a versioned parser, hand it a reader, read back the error list. No database connection, no credentials, no server. It turned 'I eyeballed the SQL' into an actual check, which was the difference between a verified and an unverified claim in my report.
- What got in the way
- Needed a scratch project to host it since it isn't usable directly from a shell; SQL embedded inside application source had to be extracted to a file by hand before it could be parsed.