Problem Obtaining multiple results from MySql Stored Proc via JDBC

Question

I've spent alot of time on this and I'd be really grateful for anyones help please!
I have written a java class to execute a MySQL stored procedure that does an UPDATE and then a SELECT. I want to handle the resultset from the SELECT AND get a count of the number of rows updated by the UPDATE. Even though several rows get updated by the stored proc, getUpdateCount() returns zero.
It's like following the UPDATE with a SELECT causes the updatecount info to be lost. I tried it in reverse: SELECT first and UPDATE last and it works properly. I can get the resultset and the updatecount.
My Stored Procedure:
delimiter $$ CREATE PROCEDURE multiRS( IN drugId int, IN drugPrice decimal(8,2) ) BEGIN UPDATE drugs SET DRUG_PRICE = drugPrice WHERE DRUG_ID > drugId; SELECT DRUG_ID, DRUG_NAME, DRUG_PRICE FROM Drugs where DRUG_ID > 7 ORDER BY DRUG_ID ASC; END $$
In my program (below) callablestatement.execute() returns TRUE even though the first thing I do in my stored proc is an UPDATE not a SELECT. Is this at odds with the JDBC 2 API? Shouldn't it return false since the first "result" returned is NOT a resultset but an updatecount? Does JDBC return any resultsets first by default, even if INSERTS, UPDATES happened in the stored proc before the SELECTs??
Excerpt of my Java Class:
// Create CallableStatement CallableStatement cs = con.prepareCall("CALL multiRS(?,?)"); // Register input parameters ........ // Execute the Stored Proc boolean getResultSetNow = cs.execute(); int updateCount = -1; System.out.println("getResultSetNow: " +getResultSetNow); while (true) { if (getResultSetNow) { ResultSet rs = cs.getResultSet(); while (rs.next()) { // fully process result set before calling getMoreResults() again! System.out.println(rs.getInt("DRUG_ID") +", "+rs.getString("DRUG_NAME") +", "+rs.getBigDecimal("DRUG_PRICE")); } rs.close(); } else { updateCount = cs.getUpdateCount(); if (updateCount != -1) { // it's a valid update count System.out.println("Reporting an update count of " +updateCount); } } if ((!getResultSetNow) && (updateCount == -1)) break; // done with loop, finished all the returns getResultSetNow = cs.getMoreResults(); }
The output of running the program at command line:
getResultSetNow: true 28, Apple, 127.00 35, Orange, 127.00 36, Bananna, 127.00 37, Berry, 127.00 Reporting an update count of 0
During my testing I have noticed:
1. According to the Java documentation execute() returns true if the first result is a ResultSet object; false if the first result is an update count or there is no result. In my java class callablestatement.execute() will return TRUE if in the stored proc the UPDATE is done first and then the SELECT last or vica versa.
2. My java class (above) is coded to loop through all results returned from the stored proc in succession. Running this class shows that any resultsets are returned first and then update counts last regardless of the order in which they appear in the stored proc. Maybe there is nothing unusual here, it may be that Java is designed to return any Resultsets first by default even if they didn't happen first in the stored procedure?
3. In my stored procedure, if the UPDATE happens last then callablestatement.getUpdateCount() will return the correct number of updated rows.
4. If the UPDATE is followed by a SELECT (see above) then callablestatement.getUpdateCount() will return ZERO even though rows were updated.
5. I tested it with the stored proc doing SELECT - UPDATE - SELECT and again getUpdateCount() returns ZERO.
6. I tested it with the stored proc doing SELECT - UPDATE - SELECT - UPDATE and this time getUpdateCount() returns the number rows updated by the last UPDATE and not the first.
My Setup:
Mac OS X 10.3.9
Java 1.4.2
mysql database 5.0.19
mysql-connector 5.1.10 (connector/J)
Maybe I have exposed a bug in JDBC?
Thanks for your help.

Answer 1

plica10 wrote:
Jschell thank you for your response.
I certainly don't mean to be rude but I often get taken that way. I like to state facts as I see them. I'd love to be proved wrong because then I would understand why my code doesn't work!
Doesn't matter to me if you are rude or not. Rudeness actually makes it more entertaining for me so that is a plus. Nothing I have seen suggests rudeness.
In response to your post:
When a MySql stored procedure has multiple sql statements such as SELECT and UPDATE these statements each produce what the Java API documentation refers to as 'results'. A Java class can cycle through these 'results' using callableStatement dot getMoreResults(), getResultSet() and getUpdateCount() to retrieve the resultset object produced by Select queries and updateCount produced by Inserts, Deletes and Updates.
As I read your question it seems to me that you have already proven that it does not in fact do that?
You don't have to read this but in case you think I'm mistaken, there is more detail in the following website under the heading 'Using the Method execute':
http://docsrv.sco.com/JDK_guide/jdbc/getstart/statement.doc.html#1000107
Sounds reasonable. But does not your example code prove that this is not what happens for the database and driver that you are using?
Myself I dont trust update counts at all since, in my experience some databases do not return them. And per other reports sometimes they don't return the correct value either.
So there are two possibilities - your code is wrong or the driver/database does not do it. For me I would also question whether in the future the driver/database would continue to behave the same if you did find some special way to write your SQL so it does do it. And of course you would also need to insure that every proc that needed this would be written in this special way. Hopefully not too many of those.
So this functionality is built into java but is not in common use amongst programmers. My java class did successfully execute a stored proc which Selected from and then finally Updated a table. My code displayed the contents of the Select query and told me how many rows were affected by the update.
It isn't "built into java". It isn't built into jdbc either. If it works at all then the driver and by proxy the database are responsible for it. I suspect that you would be hard pressed to find anything in the JDBC spec that requires what that particular link claims. I believe it is difficult to find anything that suggests that update counts in any form are required.
So you are left with hoping that a particular driver does do it.
I suppose it is rare that you would want to do things this way. Returning rowcounts in OUT parameters would be easier but I want my code to be modular enough to cover the situation where a statement may return more than one ResultSet object, more than one update count, or a combination of ResultSet objects and update counts. The sql may need to be generated dynamically, its statements may be unknown at compile time. For instance a user might have a form that allows them to build their own queries...
Any time I see statements like that it usually makes me a bit uncomfortable. If I am creating data layers I either use an existing framework or I generate the code. For the latter there is no generalization of the usage. Every single operation is laid out in its own code.
And I have in fact created generalized frameworks in the past before. I won't do it again. Benefits of the other idioms during maintenance are too obvious.