Enum datafusion_optimizer::optimizer::ApplyOrder

pub enum ApplyOrder {
    TopDown,
    BottomUp,
}

If a rule is with ApplyOrder, it means the optimizer will derive to handle children instead of recursively handling in rule. We just need handle a subtree pattern itself.

Notice: sometime result after optimize still can be optimized, we need apply again.

Usage Example: Merge Limit (subtree pattern is: Limit-Limit)

use datafusion_expr::{Limit, LogicalPlan, LogicalPlanBuilder};
use datafusion_common::Result;
fn merge_limit(parent: &Limit, child: &Limit) -> LogicalPlan {
    // just for run
    return parent.input.as_ref().clone();
}
fn try_optimize(plan: &LogicalPlan) -> Result<Option<LogicalPlan>> {
    match plan {
        LogicalPlan::Limit(limit) => match limit.input.as_ref() {
            LogicalPlan::Limit(child_limit) => {
                // merge limit ...
                let optimized_plan = merge_limit(limit, child_limit);
                // due to optimized_plan may be optimized again,
                // for example: plan is Limit-Limit-Limit
                Ok(Some(
                    try_optimize(&optimized_plan)?
                        .unwrap_or_else(|| optimized_plan.clone()),
                ))
            }
            _ => Ok(None),
        },
        _ => Ok(None),
    }
}

Variants

TopDown

BottomUp