RE: Zeus compatibility with xerces 2.4.x

Hi Florent,
Why don't you use your own generator instead of SimpleGenerator,
this would soove your problem right away. 
Now, it is probably a good suggestion not to use trim considering 
the evolution of xerces, so I have  applied your patch.
thanks,
christophe

> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of Florent BENOIT
> Sent: Wednesday, April 09, 2003 2:21 PM
> To: [email protected]
> Cc: [email protected]; [email protected];
> [email protected]
> Subject: Zeus compatibility with xerces 2.4.x
> 
> 
> 	Hello,
> 
> JOnAS needs to upgrade its xerces version (1.3.0) to the latest
> available version (2.4.0) for two reasons :
> 1/
> In order to use Axis with JOnAS, the xerces version must be changed.
> This doesn't work with the xerces version 1.3.0
> 2/ 
> The JDK from IBM (IBMJava2-14) includes the xerces 2.4.0 
> version in the
> jre/lib/xml.jar file.
> So when we compile JOnAS or JOnAS examples with this JDK, this is the
> xerces 2.4.0 version which is used and not the version packaged into
> JOnAS(1.3.0)
> 
> But, when we are using the xerces 2.4.0 we've got some problems :
> 
> Here is a short example :
> 
> When we declared an ejb-ql tag in an ejb-jar.xml, we can put the query
> on different lines like this :
> <ejb-ql><![CDATA[
>                 SELECT OBJECT(c) 
>                 FROM Cmp2_Customer c 
>                 WHERE c.lastName LIKE ?1 AND c.firstName LIKE ?2
>                 ]]>
> </ejb-ql>
> 
> When we use xerces 1.3.0 there is no problems.
> But, when xerces 2.4.0 is used there are some problems :
> The different lines of the query are collapsed :
> SELECT OBJECT(c)FROM Cmp2_Customer cWHERE c.lastName LIKE ?1 AND
> c.firstName LIKE ?2
>                 ^                   ^
>                 |                   |
>              No space           No space
> 
> The difference in the behavior between the two xerces can be explained
> by this :
> xerces 2.4.0 call the characters() SAX method for each line 
> of the query
> xerces 1.3.0  call only once the characters() SAX method for 
> the ejb-ql
> tag .
> 
> But why the spaces are deleted ??? This is due to the trim() method
> which is called in the charcaters() method.
> In the xerces 1.3.0 this is not a problem as the lines are send in one
> call so, after the trim, we've got the result query :
> 
> "SELECT OBJECT(c) 
>                 FROM Cmp2_Customer c 
>                 WHERE c.lastName LIKE ?1 AND c.firstName LIKE ?2"
> 
> 
> In the xerces 2.4.0 case we've got the result, with all the collapsed
> lines :
> "SELECT OBJECT(c)""FROM Cmp2_Customer c""WHERE c.lastName LIKE ?1 AND
> c.firstName LIKE ?2"
> 
> 
> In the xml specification, the CDATA section implies that the 
> characters
> must be kept as is. This is not the case when we are using 
> zeus because
> the spaces are removed with the trim().
> 
> A simple solution could be to delete the call to the trim() method in
> the file org.enhydra.zeus.generator.SimpleGenerator.java at the line
> 2638 :
> 
>         writer.write("        String text = new String(ch, start,
> len);\n");
>         writer.write("        text = text.trim();\n");  <-------------
> TRIM
> 
> A more complex solution :
> When using xerces 2.4.0 the startCDATA method contains 
> parameters which
> are not present in the 1.3.0 version.
> The trim shouldn't have to be done for the CDATA section (use of the
> startCDATA and endCDATA method)
> But this implies a lot of changes to allow the use of the both xerces
> version with zeus.
> 
> It would be great if a new version of zeus could be released very soon
> with the "simple solution" patch.
> Otherwise, we can't use the JDK from IBM or use Axis with JOnAS.
> (This is not a solution to put the SELECT FROM WHERE on only one line)
> 
> Thanks for a quick answer.
> 
> Best regards.
> 
> Guillaume, Xavier, Florent.
>