Advanced Topics#
Alternatives to tdprepview#
There are several viable ways to prepare data on Teradata Vantage. This page compares common alternatives to tdprepview at a glance using inline annotations for quick context.
This whole paragraph should show a tooltip on hover (try moving your mouse anywhere over the line).
Comparison matrix#
| Criterion | tdprepview | Manual SQL | Python ETL | sklearn + ONNX/PMML | In-DB functions |
|---|---|---|---|---|---|
| Runtime performance | ★★★★ |
★★☆ |
★☆ |
★★★☆ |
★★★★ |
| Dev speed / iteration | ★★★★ |
☆ |
★★★ |
★★ |
★★☆ |
| Data movement | None |
None |
High |
None |
None |
| Auditability / SQL visibility | High |
High |
Low–Medium |
Medium |
Medium |
| sklearn pipeline semantics | Yes |
No |
Yes |
Partial |
No |
| Flexibility / custom logic | Medium |
Very High |
Very High |
Medium |
Low |
| Ops simplicity (serve) | High |
Medium |
Low |
High |
Medium–High |
★★★★ excellent · ★★★ good · ★★ fair · ★ poor
☆ half star
TL;DR & how to choose:
- Pick tdprepview for declarative pipelines that emit reviewable SQL and run fully in-DB.
- Choose Manual SQL when you must hand-optimize hot paths.
- Use Python ETL for small datasets or exploratory work.
- Prefer sklearn + ONNX/PMML when your preprocessing and models convert cleanly for in-DB scoring.
- Rely on In-DB functions when you are SQL-savy, the function catalog already covers your needs and you want the simplest, fastest built-ins.