[
https://issues.apache.org/jira/browse/DRILL-5879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16217807#comment-16217807
]
ASF GitHub Bot commented on DRILL-5879:
---------------------------------------
Github user sachouche commented on a diff in the pull request:
https://github.com/apache/drill/pull/1001#discussion_r146708658
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/expr/fn/impl/SqlPatternContainsMatcher.java
---
@@ -17,37 +17,166 @@
*/
package org.apache.drill.exec.expr.fn.impl;
-public class SqlPatternContainsMatcher implements SqlPatternMatcher {
+public final class SqlPatternContainsMatcher implements SqlPatternMatcher {
final String patternString;
CharSequence charSequenceWrapper;
final int patternLength;
+ final MatcherFcn matcherFcn;
public SqlPatternContainsMatcher(String patternString, CharSequence
charSequenceWrapper) {
- this.patternString = patternString;
+ this.patternString = patternString;
this.charSequenceWrapper = charSequenceWrapper;
- patternLength = patternString.length();
+ patternLength = patternString.length();
+
+ // The idea is to write loops with simple condition checks to allow
the Java Hotspot achieve
+ // better optimizations (especially vectorization)
+ if (patternLength == 1) {
+ matcherFcn = new Matcher1();
--- End diff --
Padma, I have two reasons to follow the added complexity
1) The new code is encapsulated within the Contains matching logic; doesn't
increase code complexity
2)
o I created a test with the original match logic, pattern and input were
Strings though passed as CharSequence
o Ran the test with the new and old method (1 billion iterations) on MacOS
o pattern length
o The old match method performed in 43sec where as the new one performed in
15sec
o The reason for the speedup is the custom matcher functions have less
instructions (load and comparison)
> Optimize "Like" operator
> ------------------------
>
> Key: DRILL-5879
> URL: https://issues.apache.org/jira/browse/DRILL-5879
> Project: Apache Drill
> Issue Type: Improvement
> Components: Execution - Relational Operators
> Environment: *
> Reporter: salim achouche
> Assignee: salim achouche
> Priority: Minor
> Fix For: 1.12.0
>
>
> Query: select <column-list> from <table> where colA like '%a%' or colA like
> '%xyz%';
> Improvement Opportunities
> # Avoid isAscii computation (full access of the input string) since we're
> dealing with the same column twice
> # Optimize the "contains" for-loop
> Implementation Details
> 1)
> * Added a new integer variable "asciiMode" to the VarCharHolder class
> * The default value is -1 which indicates this info is not known
> * Otherwise this value will be set to either 1 or 0 based on the string being
> in ASCII mode or Unicode
> * The execution plan already shares the same VarCharHolder instance for all
> evaluations of the same column value
> * The asciiMode will be correctly set during the first LIKE evaluation and
> will be reused across other LIKE evaluations
> 2)
> * The "Contains" LIKE operation is quite expensive as the code needs to
> access the input string to perform character based comparisons
> * Created 4 versions of the same for-loop to a) make the loop simpler to
> optimize (Vectorization) and b) minimize comparisons
> Benchmarks
> * Lineitem table 100GB
> * Query: select l_returnflag, count(*) from dfs.`<source>` where l_comment
> not like '%a%' or l_comment like '%the%' group by l_returnflag
> * Before changes: 33sec
> * After changes : 27sec
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
