Function explain

dask/dataframe/dask_expr/diagnostics/_explain.py:20–52

(
    expr: BaseExpr, stage: OptimizerStage = "fused", format: str | None = None
)

def explain(
    expr: BaseExpr, stage: OptimizerStage = "fused", format: str | None = None
) -> None:
    graphviz = import_required(
        "graphviz", "graphviz is a required dependency for using the explain method."
    )

    if format is None:
        format = "png"

    g = graphviz.Digraph(
        STAGE_LABELS[stage], filename=f"explain-{stage}-{expr._name}", format=format
    )
    g.node_attr.update(shape="record")

    expr = optimize_until(expr, stage)

    seen = set(expr._name)
    stack: list[BaseExpr] = [expr]

    while stack:
        node = stack.pop()
        explain_info = _explain_info(node)
        _add_graphviz_node(explain_info, g)
        _add_graphviz_edges(explain_info, g)

        for dep in node.operands:
            if not isinstance(dep, Expr) or dep._name in seen:
                continue
            seen.add(dep._name)
            stack.append(dep)

    g.view()


def _add_graphviz_node(explain_info, graph):