fix(csharp): metadata(csharp): Thrift GetColumns throws NotSupportedException on GEOMETRY/GEOGRAPHY columns (#568)

[peco-engineer-bot]

Summary

Automated fix for #568 — metadata(csharp): Thrift GetColumns throws NotSupportedException on GEOMETRY/GEOGRAPHY columns.

Overrode SetPrecisionScaleAndTypeName in the editable DatabricksConnection to use SqlTypeNameParser.TryParse with a graceful fallback (raw type name + derived base type name) for SQL types the parser registry has no entry for (GEOMETRY/GEOGRAPHY), instead of the base class calling Parse which throws NotSupportedException. Verified via the two live-endpoint E2E tests, which now pass (both were red before the fix).

Files changed

• csharp/test/E2E/StatementTests.cs
• csharp/src/DatabricksConnection.cs
• .claude/knowledge/learning.md

Test plan

• AdbcDrivers.Databricks.Tests.StatementTests.CanGetColumnsOnGeospatialColumn(geoType: "GEOMETRY" — fails (red) against the original code, passes (green) after the fix
• columnName: "geom_col") — fails (red) against the original code, passes (green) after the fix
• AdbcDrivers.Databricks.Tests.StatementTests.CanGetColumnsOnGeospatialColumn(geoType: "GEOGRAPHY" — fails (red) against the original code, passes (green) after the fix
• columnName: "geog_col") — fails (red) against the original code, passes (green) after the fix

Fixes #568

🤖 Generated by engineer-bot (bug-fix flow)

Not changed in this PR (out of scope)

These facets could not be fixed here. Reason given for each — please follow up as needed:

• Register GEOMETRY/GEOGRAPHY as first-class SQL types in the shared SqlTypeNameParser registry — SqlTypeNameParser.cs and SparkConnection.cs live in csharp/hiveserver2/, a read-only git submodule (adbc-drivers/hiveserver2) that cannot be edited or PR'd from this repo.

  A cleaner long-term fix would add GEOMETRY/GEOGRAPHY parser entries (and/or a catch-all fallback) to SqlTypeNameParser's s_parserMap/s_parsers in csharp/hiveserver2/csharp/src/AdbcDrivers.HiveServer2/Hive2/SqlTypeNameParser.cs, and remove the throw in SparkConnection.SetPrecisionScaleAndTypeName's default arm. That would benefit all HiveServer2-based drivers (Spark, Impala) rather than only Databricks. The Databricks-side override implemented here fully resolves the reported crash for the Databricks driver.

[peco-review-bot]

✅ No issues identified by the review bot.