Dependencies#

How Interlace resolves relationships between models.

Automatic Dependency Detection#

Interlace automatically detects dependencies based on function parameters:

@model(name="orders", materialise="table")
def orders() -> ibis.Table:
    return load_orders()

@model(name="order_totals", materialise="table")
def order_totals(orders: ibis.Table) -> ibis.Table:
    # 'orders' parameter creates a dependency on the orders model
    return orders.group_by(orders.customer_id).agg(
        total=orders.amount.sum()
    )

The orders parameter name matches the model name, so Interlace automatically passes the orders table as input.

SQL Dependencies#

In SQL models, Interlace parses your queries and detects dependencies from table references:

SELECT *
FROM orders
WHERE amount > 100

Interlace analyzes FROM and JOIN clauses to build the dependency graph and ensure proper execution order.

Dependency Graph#

Interlace builds a Directed Acyclic Graph (DAG) of all models:

raw_users ──┬──► active_users ──► user_summary
            │
raw_orders ─┴──► enriched_orders

Execution Order#

Models are executed in topological order:

1. Models with no dependencies run first
2. Downstream models wait for their dependencies
3. Independent models run in parallel

Selective Execution#

Run specific models and their upstream dependencies:

# Run only user_summary (and its upstream dependencies)
interlace run user_summary

# Run multiple specific models
interlace run users orders

# Force re-execution (bypass change detection)
interlace run user_summary --force