Lineage

For each node, hodor needs the list of tables whose data the node reads.

dbt parents

Parents come from depends_on.nodes. dbt also adds the to model or source of a foreign key constraint there, so a model with a foreign key builds when the referenced table gets new data, even if its SQL does not read that table.

• Tables and incremental models give their own data version.
• Views and ephemeral models are expanded through the manifest to their own upstream tables, recursively. hodor memoizes the result for the run.
• For any other type (for example metric views), the version is unknown.

Sources

A source gives its Delta data version. A source that is not a Delta table (a foreign table, a view) is unknown.

Deferred parents

With --defer, dbt may compile a parent as the relation of the --state manifest, for example the base table instead of the CI copy. hodor checks which relation the compiled SQL names and takes that node's data version. A deferred view is expanded with the --state manifest, to the tables it reads in that environment. See Slim CI.

External inputs

Table references in the compiled SQL that match no parent or source:

• A Delta path reference such as delta.`<path>` gives the Delta data version of that path.
• read_files(...), other table functions and any other reference are unknown.

Note

A node that reads read_files(...) always builds (unknown_upstream). The DuckDB example select * from range(3) is also an untraceable input.

Tests

The upstreams of a test are the nodes it tests. A relationships test has two: the model it is defined on and the to model, so new data on either side runs it again. A test key combines the test's unique id with a hash of the tested relation names.