TODO
-----
Things to be done (and comments).
------------------------------------------
2003-10-01
-OMAttribution works a bit different that OMAttribution in ROML.
In fact we are using an extra class OMAttrPair to hold the
attribution pairs. Re-think that.
-OMApplication and OMError return NULL in methods getElementAt(index)
when index does not correspond to a valid value. ROML throws an
exception (IndexOutOfBound?). This behavior of ROML is new
(it used to returns NULL in older versions) and I personally do
not like it. Returning NULL would allow things like
while (pObject=getElementAt(i) != NULL){
//some code.
i++;
}
That change on ROML made many lines of previous code to be broken
(also NewApp.getElementAt(i)=OldApp.getElementAt(i+1)).
-I added a DOMReader class that uses xerces++ to read the DOM tree.
It is not yet complete (add support for bindings, attributions and
so on). No support for extra and extended attributes yet.
-Add a SAXReader (using xerces): faster?
-Add streaming functionality (to support the binary encoding).
That would be very useful for the transmitting big
OMObjects.
-OMFloat does not check for float values. The same for OMString
(this is also the case for ROML, AAAA is read a valid
integer).
-The method getType() returns an integer instead of a string (as in
ROML). Preprocesor constants, OMA, OMS, OMATTR,...
are defined at omcommon.h. Working with integers is
faster...
-I added attributes and extended attributes. Again that problem
of OMObject being virtual and things like
.
.
.
Here id is an extended attribute (of OMApplication) and ref is a
normal attribute of OMA. See implementation.
-OMFloat. I notice in ROML OMFloat dec or hex are stored as
properties of OMFloat not as attibutes. What happen is someone
uses getAttribute("dec")? So, I'm storing dec or hex as
attributes.
E.R.B
---------------------------------------------