Struct KeepEdgeOrder 

pub struct KeepEdgeOrder;

Keeps a graph node’s edge rows in the order the plan produces them, and gives them a fixed order where the plan has none.

DataFusion’s EnforceDistribution puts a round-robin repartition under an operator that benefits from parallel input, such as the FilterExec of a view with a WHERE (a subgraph) or the ProjectionExec that computes a weight expression. Draining its partitions in order would then read the batches out of file order, changing dense ids (and so labels and float summation order) from what ursa-io gives the Python library. This rule runs after the default rules and removes every round-robin repartition below a graph node: a round-robin repartition only adds parallelism, so removing it never changes a result.

What remains may still not preserve order: a join, an aggregate, a hash repartition, a merge of partitions or a UNION produces its rows in an order that can change between runs and thread counts (#263). Over such a plan, the rule sorts the edge rows by (source, destination), and a weighted node’s rows by their weights after that, so that rows that tie are identical in every column the node reads. A traversal’s node-id inputs over such a plan are sorted by value, and so are a graph’s vertex keys, which set the order of its nodes. Order-preserving plans (scans, views over files, filters, projections) are left as they are, in file order. The sort is a SortExec under the node, so EXPLAIN shows it.

Trait Implementations

impl Debug for KeepEdgeOrder

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PhysicalOptimizerRule for KeepEdgeOrder

fn optimize( &self, plan: Arc<dyn ExecutionPlan>, _config: &ConfigOptions, ) -> Result<Arc<dyn ExecutionPlan>>

Rewrite plan to an optimized form.

fn name(&self) -> &str

A human readable name for this optimizer rule

fn schema_check(&self) -> bool

A flag to indicate whether the physical planner should validate that the rule will not change the schema of the plan after the rewriting. Some of the optimization rules might change the nullable properties of the schema and should disable the schema check.

fn optimize_with_context( &self, plan: Arc<dyn ExecutionPlan>, context: &dyn PhysicalOptimizerContext, ) -> Result<Arc<dyn ExecutionPlan>, DataFusionError>

Rewrite plan with access to extended context (statistics registry, etc.).