[Tokyohenjin]

Summary

The MSSQL driver currently documents decimal and numeric columns without their precision and scale, and fixed-length char, nchar and binary columns without their length. For example, a decimal(9,2) column appears as just decimal, and char(5) as char. The variable-length types (varchar, nvarchar, varbinary) already include their length, so this makes the output consistent.

Changes

• The column query in analyzeTable now also reads sys.columns.precision and sys.columns.scale.
• convertColumnType formats:
  • decimal / numeric as decimal(p,s) / numeric(p,s)
  • char / binary as char(n) / binary(n)
  • nchar as nchar(n), halving max_length the same way nvarchar does
• Added TestColumnTypePrecisionAndScale, covering explicit precision/scale, the implicit decimal(18,0) default, and the fixed-length types.

Behavior change

Generated docs for existing MSSQL schemas will change wherever these types appear. In particular, a column declared as bare decimal will now show as decimal(18,0), because the values come from sys.columns rather than the DDL text.

Samples

sample/mssql is unaffected, since testdata/ddl/mssql.sql doesn't use any of these types.

sample/azuresql needs regenerating: three decimal columns in testdata/ddl/azuresql.sql will now render as decimal(10,2) / decimal(12,2). I couldn't build the CLI locally due to local IT restrictions. Sorry for the extra step!

sample/azuresql is regenerated in commit 1abe749 and shows the updated values in tables and SVGs.

Testing

go test -tags mssql ./drivers/mssql/ passes against the local SQL Server container.