Landing families

Five families write the landing layer. Three of them are the per-table shapes a landing table routes to, and this page documents those. The other two are ingest-workflow (databricks/landing/ingest_workflow.jinja, roots workflow and dv), which renders at workflow grain rather than per table, and ingest-piece (databricks/landing/pieces/cloud_files_table.jinja and jdbc_table.jinja), the include-only per-table piece those workflow-grain bodies compose. Both are indexed on the overview page and their field-by-field sections are still to be written.

A landing table renders through one of three shapes, decided in a fixed order. First the ingest runtime is routed from the table’s ingestion settings, the source connector type, and whether the source connection carries Lakeflow Connect settings. Then the file change-data-capture mode is rejected outright (the routed runtime would be change-blind for it). What remains routes: the Lakeflow Connect runtime renders through the run-scoped pipeline resource (never a per-table file), the declarative pipeline runtime takes the ingest-pipeline shape, and every other runtime takes the ingest-notebook shape with the runtime naming which notebook.

ingest-pipeline

Templates: databricks/landing/ingest_pipeline.jinja

Roots: workflow, this, connection (conditional), source (conditional), sources, ingest, dv

The declarative landing pipeline: one pipeline per source connection, replicating source tables verbatim into the landing (bronze) catalog. Because a landing table has no audit columns of its own, there is no methodology root here.

Root: workflow

Field	Type	Meaning
workflow.name	string	landing_ plus the source connection’s raw name; the literal landing_unlinked in a preview with no resolved source connection (a Build throws first).

Root: this

The landing table being written: a table reference with the resolved ingestion block and a Python-safe function name on top.

Field	Type	Meaning
this.name	string	Landing table name, raw.
this.fqn	string	The stored fully qualified name, else composed from the landing catalog, schema, and name.
this.columns[]	array	The landing table’s own columns, ordinal-sorted, system columns included. Elements carry ulid (when present), name, dataType, isPrimaryKey, isNullable.
this.parts.catalog	string	Destination catalog assigned to the landing layer (sample_catalog in an unpinned preview).
this.parts.schema	string	Destination schema (sample_schema in an unpinned preview).
this.parts.name	string	Destination table name.
this.ingestion.ingestMode	string	Resolved ingest mode; defaults to snapshot.
this.ingestion.watermarkColumn	string	The landing column the high-water bound is read from; empty when none. A resolvable high-water-mark filter’s target column outranks the stored setting.
this.ingestion.sourceClause	string	The restriction appended unconditionally to the source read, complete with its leading WHERE; empty when nothing restricts. Python-string escaped. Exactly one of sourceClause and watermarkClause is ever populated.
this.ingestion.watermarkClause	string	The restriction that applies only once a previous run has landed rows, complete with its leading WHERE, carrying the {lower_bound} placeholder, brace-escaped for the formatted string it lands in.
this.ingestion.cdcKeyColumns	string[]	Change-data-capture key columns; empty by default.
this.ingestion.cdcSequenceColumn	string	Change-data-capture sequence column; defaults to __$start_lsn.
this.ingestion.cdcTombstoneHandling	string	Tombstone policy; defaults to delete_row.
this.pythonFunctionName	string	The table name as a Python identifier: sanitized, with every hyphen replaced by an underscore.

Root: connection (conditional)

Present only when the source connection resolved. Always the source connector, never the destination.

Field	Type	Meaning
connection.name	string	Source connection’s raw name.
connection.connectionType	string	Source connector type (for example SQL Server).

Roots: source and sources

source is present only when the source connection, source table, and source schema all resolved; it is omitted rather than emitted empty, so a template’s {% if source %} guard takes the branch a real Build would. sources is always present: a one-element list holding the identical object, or empty. Beyond the table reference fields, source.schema carries the source schema name; source.parts is deliberately absent (a source-system table has no honest catalog).

ingest-notebook

Templates: databricks/landing/auto_loader.jinja, databricks/landing/pyspark_jdbc.jinja, databricks/landing/pyspark_rest.jinja

Roots: this, connection, source (conditional), sources, environment, extract, dv

The per-table landing notebooks. Unlike the pipeline shape there is no workflow root, the connection root is always present (empty strings when unresolved), and the ingestion block is passed raw rather than resolved: the notebooks read change-data-capture settings with no defaults, and resolving them would change rendered bytes.

Root: this

Field	Type	Meaning
this.ulid	string	Landing table identifier (supplied on this shape, unlike the pipeline shape).
this.name	string	Landing table name.
this.fqn	string	The stored fully qualified name, else composed from catalog, schema, and name.
this.columns[]	array	The landing table’s own columns, ordinal-sorted, system columns included.
this.parts.catalog / .schema / .name	string	Destination (bronze) catalog, schema, and table name.
this.ingestion	object	The raw stored ingestion block, empty object when absent. Deliberately unresolved: fields may be missing and templates read them with no defaults.
this.ingestion.ingestMode	string?	Ingest mode, raw.
this.ingestion.watermarkColumn	string?	Watermark column, raw.
this.ingestion.runtimeOverride	string?	The stored runtime override that routed this table to a notebook.
this.ingestion.cdcKeyColumns	string[]?	Raw change-data-capture key columns.
this.ingestion.cdcSequenceColumn	string?	Raw sequence column.
this.ingestion.cdcTombstoneHandling	string?	Raw tombstone policy.

Root: extract

Always present: what the notebook actually extracts from, resolved ahead of render. When a source table is linked, its schema and name; when none is, the landing table’s own (and the builder reports an unlinked-source-self-reference diagnostic so the fallback stays visible). A linked source wins even when its fields are empty strings. Both the raw and the snake-cased forms travel: raw feeds the notebook’s headers and the JDBC table option, snake-cased feeds the physical bronze table name and the quarantine path.

Field	Type	Meaning
extract.schema	string	Source schema when linked; the landing table’s own schema when not.
extract.name	string	Source table name when linked; the landing table’s own name when not.
extract.schemaSnakeCase	string	The schema through the snake-case naming model.
extract.nameSnakeCase	string	The name through the snake-case naming model.

Roots: source, sources, connection, environment

Field	Type	Meaning
source	table ref?	Conditional: present only when a source table resolved (no connection or schema check on this shape). Identical object to sources[0]. Carries the extra source.schema; parts is absent by design.
sources	array	Always present: one element or empty.
connection.name	string	Source connection name; empty string when unresolved (always present on this shape).
connection.connectionType	string	Source connector type; empty string when unresolved.
environment.catalogs.bronze	string	The landing (bronze) catalog the notebook writes into.
environment.schemaPolicy	string	Target connection’s schema policy; defaults to permissive.
environment.landingRoot	string	Target connection’s landing storage root; empty when unset.

lakeflow-entry

Templates: databricks/landing/lakeflow_ingest_object.jinja

Roots: this, source, sources, connection

One Lakeflow Connect ingestion entry. All four roots are unconditional: a table with no source link refuses to render rather than emitting a partial context. Entries carry no column data by design, so this.columns and source.columns are always empty; the entry names tables, it does not enumerate them.

Field	Type	Meaning
this.name	string	Destination table name.
this.fqn	string	Destination catalog.schema.table.
this.columns	array	Always empty by design.
this.parts.catalog / .schema / .name	string	The destination split into parts.
this.key	string	The pipeline resource key: lfc_ plus the snake-cased connection, source schema, and source table.
this.ingestion.ingestMode	string	Resolved ingest mode; snapshot default.
this.ingestion.watermarkColumn	string	The source column a high-water-mark filter restricts (name only, never composed SQL), else the stored setting, else empty. The filter outranks the stored setting.
source.name	string	Source table name.
source.fqn	string	Source schema.table.
source.schema	string	The source schema name (family extra; parts is absent by design).
sources	array	One element, identical object to source.
connection.name	string	Source connection name.
connection.connectionType	string	Source connector type as DeltaVault stores it (for example Azure SQL Database).
connection.connector	string	Precomputed Lakeflow Connect connector key (sqlserver, oracle, postgres, mysql, salesforce, workday, servicenow, netsuite, googleanalytics); an unmapped type lowercases through.

The bundle’s bundle-lakeflow family loops enriched entries carrying these same four roots.