[
https://issues.apache.org/jira/browse/SPARK-7754?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Owen updated SPARK-7754:
-----------------------------
Component/s: SQL
> Use PartialFunction literals instead of objects in Catalyst
> -----------------------------------------------------------
>
> Key: SPARK-7754
> URL: https://issues.apache.org/jira/browse/SPARK-7754
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Reporter: Edoardo Vacchi
>
> Catalyst rules extend two distinct "rule" types: {{Rule[LogicalPlan]}} and
> {{Strategy}} (which is an alias for {{GenericStrategy[SparkPlan]}}).
> The distinction is fairly subtle: in the end, both rule types are supposed to
> define a method {{apply(plan: LogicalPlan)}}
> (where LogicalPlan is either Logical- or Spark-) which returns a transformed
> plan (or a sequence thereof, in the case
> of Strategy).
> Ceremonies asides, the body of such method is always of the kind:
> {code:java}
> def apply(plan: PlanType) = plan match pf
> {code}
> where `pf` would be some `PartialFunction` of the PlanType:
> {code:java}
> val pf = {
> case ... => ...
> }
> {code}
> This is JIRA is a proposal to introduce utility methods to
> a) reduce the boilerplate to define rewrite rules
> b) turning them back into what they essentially represent: function types.
> These changes would be backwards compatible, and would greatly help in
> understanding what the code does. Current use of objects is redundant and
> possibly confusing.
> *{{Rule[LogicalPlan]}}*
> a) Introduce the utility object
> {code:java}
> object rule
> def rule(pf: PartialFunction[LogicalPlan, LogicalPlan]):
> Rule[LogicalPlan] =
> new Rule[LogicalPlan] {
> def apply (plan: LogicalPlan): LogicalPlan = plan transform pf
> }
> def named(name: String)(pf: PartialFunction[LogicalPlan, LogicalPlan]):
> Rule[LogicalPlan] =
> new Rule[LogicalPlan] {
> override val ruleName = name
> def apply (plan: LogicalPlan): LogicalPlan = plan transform pf
> }
> {code}
> b) progressively replace the boilerplate-y object definitions; e.g.
> {code:java}
> object MyRewriteRule extends Rule[LogicalPlan] {
> def apply(plan: LogicalPlan): LogicalPlan = plan transform {
> case ... => ...
> }
> {code}
> with
> {code:java}
> // define a Rule[LogicalPlan]
> val MyRewriteRule = rule {
> case ... => ...
> }
> {code}
> and/or :
> {code:java}
> // define a named Rule[LogicalPlan]
> val MyRewriteRule = rule.named("My rewrite rule") {
> case ... => ...
> }
> {code}
> *Strategies*
> A similar solution could be applied to shorten the code for
> Strategies, which are total functions
> only because they are all supposed to manage the default case,
> possibly returning `Nil`. In this case
> we might introduce the following utility:
> {code:java}
> object strategy {
> /**
> * Generate a Strategy from a PartialFunction[LogicalPlan, SparkPlan].
> * The partial function must therefore return *one single* SparkPlan for
> each case.
> * The method will automatically wrap them in a [[Seq]].
> * Unhandled cases will automatically return Seq.empty
> */
> def apply(pf: PartialFunction[LogicalPlan, SparkPlan]): Strategy =
> new Strategy {
> def apply(plan: LogicalPlan): Seq[SparkPlan] =
> if (pf.isDefinedAt(plan)) Seq(pf.apply(plan)) else Seq.empty
> }
> /**
> * Generate a Strategy from a PartialFunction[ LogicalPlan, Seq[SparkPlan]
> ].
> * The partial function must therefore return a Seq[SparkPlan] for each
> case.
> * Unhandled cases will automatically return Seq.empty
> */
> def seq(pf: PartialFunction[LogicalPlan, Seq[SparkPlan]]): Strategy =
> new Strategy {
> def apply(plan: LogicalPlan): Seq[SparkPlan] =
> if (pf.isDefinedAt(plan)) pf.apply(plan) else Seq.empty[SparkPlan]
> }
> }
> {code}
> Usage:
> {code:java}
> val mystrategy = strategy { case ... => ... }
> val seqstrategy = strategy.seq { case ... => ... }
> {code}
> *Further possible improvements:*
> Making the utility methods `implicit`, thereby
> further reducing the rewrite rules to:
> {code:java}
> // define a PartialFunction[LogicalPlan, LogicalPlan]
> // the implicit would convert it into a Rule[LogicalPlan] at the use sites
> val MyRewriteRule = {
> case ... => ...
> }
> {code}
> *Caveats*
> Because of the way objects are initialized vs. vals, it might be necessary
> reorder instructions so that vals are actually initialized before they are
> used.
> E.g.:
> {code:java}
> class MyOptimizer extends Optimizer {
> override val batches: Seq[Batch] =
> ...
> Batch("Other rules", FixedPoint(100),
> MyRewriteRule // <--- might throw NPE
> val MyRewriteRule = ...
> }
> {code}
> this is easily fixed by factoring rules out as follows:
> {code:java}
> class MyOptimizer extends Optimizer with CustomRules {
> val MyRewriteRule = ...
> override val batches: Seq[Batch] =
> ...
> Batch("Other rules", FixedPoint(100),
> MyRewriteRule
> }
> {code}
> or, even better, further modularizing, e.g using traits or container objects:
> {code:java}
> class MyOptimizer extends Optimizer with CustomRules {
> override val batches: Seq[Batch] =
> ...
> Batch("Other rules", FixedPoint(100),
> MyRewriteRule
> }
> trait CustomRules {
> val MyRewriteRule = ...
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
