Module optimizer

Re-exports§

delta_join_solver 🔒: The solver for delta join, which determines lookup order of a join plan. All collection types in this module should be BTree to ensure determinism between runs.
heuristic_optimizer 🔒
logical_optimization 🔒
optimizer_context 🔒
plan_expr_rewriter
plan_expr_visitor 🔒
plan_node: Defines all kinds of node in the plan tree, each node represent a relational expression.
plan_rewriter 🔒
plan_visitor 🔒
property: Define all property of plan tree node, which actually represent property of the node’s result.
rule 🔒: Define all Rule

ExecutionModeDecider
LogicalOptimizer
OptimizationStage
OptimizerContext
PlanRoot: PlanRoot is used to describe a plan. planner will construct a PlanRoot with LogicalNode. and required distribution and order. And PlanRoot can generate corresponding streaming or batch plan with optimization. the required Order and Distribution columns might be more than the output columns. for example:
ReadStorageTableVisitor
RelationCollectorVisitor: TODO(rc): maybe we should rename this to DependencyCollectorVisitor.
SysTableVisitor

PlanPhase: PlanPhase is used to track the phase of the PlanRoot. Usually, it begins from Logical and ends with Batch or Stream, unless we want to construct a PlanRoot from an intermediate phase. Typical phase transformation are:

PlanRewriter: it’s kind of like a PlanVisitor<PlanRef>, but with default behaviour of each rewrite method
PlanVisitor: The visitor for plan nodes. visit all inputs and return the ret value of the left most input, and leaf node returns R::default()

const_eval_exprs 🔒
exist_and_no_exchange_before 🔒
find_version_column_index 🔒
inline_session_timezone_in_exprs 🔒
require_additional_exchange_on_root_in_distributed_mode 🔒: As we always run the root stage locally, for some plan in root stage which need to execute in compute node we insert an additional exhchange before it to avoid to include it in the root stage.
require_additional_exchange_on_root_in_local_mode 🔒: The purpose is same as require_additional_exchange_on_root_in_distributed_mode. We separate them for the different requirement of plan node in different execute mode.