create-a-derived-table update incremental models guidance

[tamsinforbes]

To ensure tables are properly cleaned up when either the insert_overwrite or append incremental strategies are defined we need to better understand what the expected behaviour of these strategies is.

I.e., should a partition always be be defined, should there be logic to only clean up tables when the --full-refresh option is used, etc.?

Investigate other related behaviour such as

• + prefix
  • If models m1, m2, m3, have the same ancestors does dbt run --select +m1 +m2 +m2 rerun those ancestor models 3 times?
    • No it doesn’t - if the depends list overlaps it only runs a unique set; try dbt run --select +m1 +m1 it still only runs a unique set
  • does dbt run --select +m1 include running the seeds that m1 depends on as well as model ancestors.
    • No - seeds must be deployed with the dbt seed command before the models that depend on them are run with dbt run
• seeds how to have versioned seeds, eg lookup_offence
  • just have a column for version?
    • yes - seed will grow linearly but preserves ease of access for users to previous versions
  • or a partition?
    • yes - also partition by release to make performant when users filter by release

Done when

further incrementals training/guidance is added to user guidance