macinnir / dvc Goto Github PK

Generate code based on datatypes and meta defined in schema?

Currently, when generating models, the field order is randomized (as is normal go functionality with randomized map ordering). Field order should be consistent either with the order in the database or alphabetically.

Either way, the randomization of field order causes noise in git diff when looking at schema changes as large portions of the schema change simply because the ordering of fields is different.

• A table compare has produced two alter statements: "DROP COLUMN..." and "ADD COLUMN..."
• Those two columns have the same data type.

Desired behavior
DVC identifies a column as being renamed when:

• DROP column and CREATE Column are of the same type

Identify if a table has been renamed.

Current Behavior: If a table name does not match, it will be dropped. And the newly named table will be created.

Desired behavior: Rename a table if the old table and the new table have --

• The same # of columns
• The same columns types and names

This will also mean that if a table has been renamed and a column has been changed, the table will be dropped.

Desired Behavior

• Specify enumeration tables in dvc.toml via enums = [ "someTable", ... ]
• Import values into [tableName].schema.json via the dvc import command
• Include enumeration table in comparison when running dvc compare
• Apply enumeration table data (via DELETE all and INSERT) via dvc compare apply