I suppose, nobody really cares :-) but I entered a Jira anyway (TUSCANY-4075) I looked further into it for potential side effects and commented in Jira - was not sure about sequences and lists but that should be safe too (and at least not worse than today).
I tested it and did not find any problems so far. Still, if there are comments, I'd be glad for opinions/...!! Rgds, tge On Tue, 2013-02-19 at 16:32 +0100, Thomas Gentsch wrote: > Hello all, > > we are using both C++ and Java SDO in a project and discovered some > misbehavior in the C++ components with XML data converted from/to SDO if > the XML contains either escaped chars of CDATA. Java seems to do it > mostly right (see below) > > When looking at the SDO (C++ M3) code and searching on the web (e.g. > [1]) it looks as if this topic seemed a bit, well, incomplete in the C++ > world. > > The problem (C++): > - loading an XML with CDATA inside works nicely, the CDATA remains > intact, therefore saving works nicely too. However, if I do a > DataObjectPtr->getCString(), I get the CDATA in the returned value - > means as a user I have to deal with that :-| > - loading an XML with escaped (e.g. <) works too, libxml2 converts > these chars. getCString() returns the real text (e.g. "<"), but > saving does not re-insert the escaping - i.e. the resulting XML is > not usable anymore (TUSCANY-1553) > In Java this looks much better and quite as I'd expect it to: > - loading XML with either constructs works > - using getCString() just returns the real text with the escaped > sections converted > - saving works too, CDATA are lost but are rather converted back to > escaped XML - this is not the *original* XML anymore but at least it > is valid and logically it is the same as the input > - Example: > Input XML: > <tns1:name>ü<>bla blub <![CDATA[ <<>> ]]></tns1:name> > getCString() in Java: > "ü<>bla blub <<>> " > Saving this as XML: > <tns1:name>ü<>bla blub <<>> </tns1:name> > The only questionable thing is the saved "ü" ... to be > converted back to ü or ü ? > > Anyway, now the question: As it seems there were discussions going on > when SDO C++ has been implemented - has the approach above (as in Java) > ever been considered and, if so, why has it not been followed? > I believe that this would have been also much simpler than it is today: > - while parsing > - the cdata handler function of the SAX2 handler just > appends the text returned by libxml2 > - escaped chars are converted by libxml2 anyway > - the property value now contains the real text > (e.g. "ü<>bla blub <<>> ") and returns it just as-is in getCString() > - setting that property also just sets the passed-in value > - saving the property just calls libxml2 xmlTextWriterWriteString() > which should escape the special chars > > Another advantage is that users don't need to worry about (un)escaping > special chars or CDATA as today. > > Any insight is very welcome! > Regards, > tge
