Benchmark Modification For Cubrid by Qianmin-Jiang · Pull Request #188 · oltpbenchmark/oltpbench

Qianmin-Jiang · 2017-12-15T05:29:22Z

This PR adds modifications on benchmark files to run Cubrid with TPCC, Wikipedia and TPCH.
For TPCH, it does not include any dialect and right now we directly modify the SQLs in /tpch/queries/Q{1-22}.java.

…id-ddl; add jar file to oltpbenchmark/lib

apavlo

Please make the requested changes.

apavlo · 2017-12-16T19:21:42Z

src/com/oltpbenchmark/api/Procedure.java

            // Everyone else can use the regular getGeneratedKeys() method
            else if (is != null) {
-                pStmt = conn.prepareStatement(stmt.getSQL(), is);
+                if (getDatabaseType() == DatabaseType.CUBRID) {


Can you add a comment to explain why you need this special clause?

apavlo · 2017-12-16T19:22:10Z

src/com/oltpbenchmark/benchmarks/tpch/queries/Q1.java

            +     "lineitem "
            + "where "
-            +     "l_shipdate <= date '1998-12-01' - interval '95' day "
+            +     "l_shipdate <= ADDDATE('1998-12-01', interval -'95' day) "


Can you remove all of the changes that you made to the TPC-H queries for Cubrid? We will have to fix the benchmark later to add real dialect files.

Sure, we will remote these.

apavlo · 2017-12-16T19:22:31Z

src/com/oltpbenchmark/benchmarks/wikipedia/WikipediaLoader.java

        if(this.getDatabaseType() == DatabaseType.ORACLE) {
            // Oracle handles quoted object identifiers differently, do not escape names
            sql = SQLUtil.getInsertSQL(catalog_tbl, false);
+        } else if(this.getDatabaseType() == DatabaseType.CUBRID) {


Please add comments to explain what this does.

It means for the generated the SQL queries, include the column names and skip the first auto_increment column.
The field is set to be auto_increment since in the insert queries in the UpdatePage, these fields are not set and they have not null constraints.
Both the loading phase and the procedures are inserting records. In the previous code, in the loading phase, it sets values for all columns. And in the procedures, it does not set value for the auto_increment fields. For Cubrid, it does not update the initial value of the auto_increment fields, while MySQL does this. So if we manually set values for these fields in the loading phase, in the execution phase Cubrid will set these values starting from some initial value which may already exist because of the loading phase. This causes the unique constraint violation. To avoid this, we skip setting these fields in the loading phase.

apavlo · 2017-12-16T19:22:41Z

src/com/oltpbenchmark/benchmarks/wikipedia/WikipediaLoader.java

            // Oracle handles quoted object identifiers differently, do not escape names
            sql = SQLUtil.getInsertSQL(catalog_tbl, false);
+        } else if(this.getDatabaseType() == DatabaseType.CUBRID) {
+            sql = SQLUtil.getInsertSQL(catalog_tbl, true, false, 1, 0);


Please add comments that explain these parameters.

apavlo · 2017-12-16T19:23:11Z

src/com/oltpbenchmark/benchmarks/wikipedia/WikipediaLoader.java

                // Insert the text
                int col = 1;
-                textInsert.setInt(col++, rev_id); // old_id
+                if (this.getDatabaseType() != DatabaseType.CUBRID) {


Why do you need this?

Old_id is an auto_increment field and we skip setting this field here.

apavlo · 2017-12-16T19:24:04Z

src/com/oltpbenchmark/benchmarks/wikipedia/procedures/UpdatePage.java

 			for (Integer otherUserId : wlUser) {
-				ps.setInt(param, otherUserId.intValue());
-				ps.addBatch();
+			    param = 1;


What is this change?

In the previous code, each time it creates a prepared statement, it rebinds the userId field and does not change other fields.
For Cubrid, it complains that in the generated statements, not all parameters are binded. From the log record in Cubrid, in the query that server received, only userId is binded and other params are not. So we rebind all the params here.

apavlo · 2017-12-16T19:24:29Z

src/com/oltpbenchmark/benchmarks/wikipedia/procedures/UpdatePage.java

 				} catch (SQLException esql) {
 					int errorCode = esql.getErrorCode();
-					if (errorCode == 8177)
+					if (errorCode == 8177) 


Is this Cubrid specific?

No, this code exists in the previous code. Sorry we accidentally add a white space. We will remove that.

apavlo · 2017-12-16T19:24:50Z

src/com/oltpbenchmark/util/SQLUtil.java

-//    	sb.append(";");
-    	
-    	return (sb.toString());
+        StringBuilder sb = new StringBuilder();


What did you change in this function?

We used 4 spaces as indentation here and the master branch also has 4 spaces indentation. But the version compared to has 3 indentation. So we keep the 4 spaces indentation here.

… into cubrid_dev

Ubuntu and others added 9 commits December 9, 2017 22:58

tpcc

cdccc9d

ddls

da91781

add configuration file for tpch/wikipedia; add AUTO_INCREMENT to cubr…

632e5f4

…id-ddl; add jar file to oltpbenchmark/lib

modify auto_increment constraint

9d05c27

fixed auto increment bug

35ad298

fixed execute batch bug in wikipedia; now it runs

6635d8a

Recover accidently deleted file

4765c13

Unify fix for Wikipedia loading + cleanup

f1656c6

Fix for executeBatch in Cubrid

42c9e96

apavlo suggested changes Dec 16, 2017

View reviewed changes

Cubrid TPCC config setting

0b50469

dengyuchenkit force-pushed the cubrid_dev branch from a33d37a to 0b50469 Compare December 16, 2017 21:42

Yuchen Deng and others added 5 commits December 16, 2017 17:13

Comments

c4be148

the space and tab

b28a7ca

Merge branch 'cubrid_dev' of https://github.com/Qianmin-Jiang/oltpbench…

4b801f2

… into cubrid_dev

the space and tab

cdb5bd6

add comment to special cubrid prepareStatement method

ab1c013

Conversation

Qianmin-Jiang commented Dec 15, 2017

Uh oh!

apavlo left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

dengyuchenkit Dec 16, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

dengyuchenkit Dec 16, 2017 •

edited

Loading