DBT Core Concepts & Architecture Flashcards

What does the `--full-refresh` flag do when running incremental models?

Answer: Forces dbt to drop and recreate the incremental model as a full table

The `--full-refresh` flag causes dbt to drop the existing incremental table and rebuild it from scratch, identical to running it as a table materialization.

What is a dbt adapter?

Answer: A plugin that allows dbt to connect to and compile SQL for a specific data platform

A dbt adapter (e.g., dbt-bigquery, dbt-snowflake) is a plugin that enables dbt to translate its generic operations into platform-specific SQL and API calls.

What does the `vars` block in `dbt_project.yml` enable?

Answer: Defining project-wide variables accessible via the `var()` function

The `vars` block defines project-level variables accessible via `var('variable_name')` in models and macros, and can be overridden at runtime with `--vars`.

Which materialization type creates a permanent, physically stored table in the data warehouse on every run?

Answer: table

The `table` materialization drops and recreates a full physical table in the warehouse on every run, as opposed to a view which is only a stored query.

What is the purpose of the `manifest.json` file generated by dbt?

Answer: Contains a complete representation of the dbt project including all nodes and their relationships

`manifest.json` is a comprehensive artifact capturing all project resources (models, tests, sources, macros) and their metadata, used by dbt docs and CI tooling.

When using dbt Core, which directory stores compiled SQL and run artifacts by default?

Answer: target/

dbt writes compiled SQL, run results, and other artifacts to the `target/` directory by default, which is configurable in `dbt_project.yml`.

What information does `run_results.json` contain after a dbt invocation?

Answer: Execution metadata such as model status, timing, and row counts from the last dbt invocation

`run_results.json` is generated after every dbt invocation and records execution status, timing, and results for each node that was processed.