In early weeks of the migration, I tried to implement all the logic that might be required to read the index. But I found it rather inefficient, so later I tried to create various stubs with specific panic messages. So whenever I hit a panic complaining about missing implementation, then I do the migration. It would save me a lot of time to avoid migrate unnecessary stuff. It does help and I have been hitting several panic messages now, including
- Supported slice in SimpleFSIndexInput.
- Added Numeric entry parsing in DocValuesReader.
- Added CompressingStoredFieldsReader and CompressingStoredFieldsIndexReader.
- Error handling of currupted index.
- Fixed FieldInfo read logic.