[jbosstools-dev] RE: JBIDE-2173 & JBIDE-2162 - performance improvement (for MySQL?)

[Vitali Yemialyanchyk]

>> Any news ?
I have attached screenshots with unit test for JBIDE-2173:
1.  HSQLDB_JUnit_Screens.zip
2.  MySQL_JUnit_Screens.zip
3.  Oracle_JUnit_Screens.zip
4.  Postgre_JUnit_Screens.zip
You can check it all easily. Just unzip and look at pictures.
You can see what in both cases result the same - so patch
JBIDE-2173_20080630.zip should be correct.
For testing both, open JDBCReader -> readDatabaseSchema -> comment/uncomment
// method 1: - old method
// method 2: - new one
Remember what I update some unit tests for better work with dbs Oracle &
MySQL and
update some ambiguous places - like IncrementalSchemaReadingTest and others.
So if you apply the patch without my junit updates and try to run tests for
both cases you can get some other results - but I expect this as unit tests
problems.
JBIDE-2229 - JBIDE-2229_20080529.zip - my junits update patch.
IncrementalSchemaReadingTest is still have no patch to download and apply.
IncrementalSchemaReadingTest has the iterator dependents - it expected first
element "CHILD",
In general case thin is not a true.


>> I'm sorry, but I don't understand your point. "is just disputable" ?
At that moment I did not see all benefits and drawbacks of one and other
solution - so
I just wanted to discuss both. Now I see what current solution is better -
so
we can terminate this topic (the topic about "incorporate quotes process
into DatabaseCollector").

Vitali

-----Original Message-----
From: Max Rydahl Andersen [mailto:max.andersen at redhat.com] 
Sent: Thursday, August 07, 2008 2:05 PM
To: Vitali Yemialyanchyk
Cc: jbosstools-dev at lists.jboss.org
Subject: Re: JBIDE-2173 & JBIDE-2162 - performance improvement (for MySQL?)

>>> Did you create a patch for this ? Does all unittests passes ? I assume
> alot of logic
>>> has to change now that foreignkeys are in "reverse" order ?
>
> Yes I have created patch for this ->
> https://jira.jboss.org/jira/browse/JBIDE-2173 -> JBIDE-2173_20080630.zip.
> I added some info about process time before and after the patch.
> I compare unittests before patch and after - it had the same number of
> Runs/Errors/Failures in both cases.
> You could check the patch to find possible problems - as I expect it  
> should
> work -
> may be it's logic not so smooth - we could discuss it to find better
> solution.
> I'll compare all unittests once more for better confidence and report you
> about all problems.

Any news ?

>>> sorry, but I don't see what is obvious in that ?
>
> You could see this changes in JBIDE-2173_20080630.zip ->
> org.hibernate.cfg.reveng_patch.txt ->
>>>
> -				Table fkTable = dbs.getTable(fkSchema,
> fkCatalog, fkTableName);
> +				Table fkTable =
> dbs.getTable(quote(fkSchema), quote(fkCatalog), quote(fkTableName));
>>>
> I find this error in some unittest for quotes (when I tested
> JBIDE-2173_20080630 patch with unittestes) ->
> This is a way to add table into dbs:
>
> Table table = dbs.addTable(quote(getSchemaForModel(schemaName)),
> getCatalogForModel(catalogName), quote(tableName));
>
> You see - we add quotes for schema & table (if it's necessary to have  
> quotes
> added),
> but then we try to get table -> dbs.getTable we have fkSchema without  
> quotes
> - and get an error.
> In the case without my patch this situation was no reproducible.
>
>>> > I think the better way incorporate quotes process into
> DatabaseCollector.
>>> Why ?
>>> Please explain.
>
> My statement is disputable - " incorporate quotes process into
> DatabaseCollector " means
> We can use
>
>  Table fkTable = dbs.getTable(fkSchema, fkCatalog, fkTableName);
> &
>  Table table = dbs.addTable(getSchemaForModel(schemaName),
> getCatalogForModel(catalogName), tableName);
>
> And forget about quotes outside DatabaseCollector,

I can't see how we can forget about quotes outside db collector ?

> But seems we can't easy and beautiful transfer the function quote() from
> JDBCReader to DatabaseCollector offsprings.
> So statement is just disputable.

I'm sorry, but I don't understand your point. "is just disputable" ?

/max