Hello again
I would like to draw your attention to the xmlRemoveID function in the valid.c file of the libxml2 tree. I have noticed a bug in that this function does not actually do what it is intended for and remove the ID from the internal ID hash table. I have attached a patch that rectifiesNo that's intended. The IDness is defined by the presence of the attribute. I think your patch is wrong w.r.t. the definition of IDness in the XML spec.
I appreciate that the 'IDness' is only defined by the presence of the
attribute and the attribute's declaration as an ID type in the DTD.
However, the problem is that when I rename an attribute with xmlSetProp
that is an ID, the value is changed in the xml tree, but the respective
document's ID hash table is not updated to the correct value. As a result
of this, future xmlGetID lookups may give false positives for attributes
that were removed. So, it seems that I am supposed to remove the ID
explicitely from the documents IDs hash table with xmlRemoveID, which is
documented as removing "the given attribute from the ID table maintained
internally." In the function's present form, this appears to always fail
due to a miscast pointer:
xmlAttrPtr cur;
.
.
.
cur = xmlHashLookup(table, ID);
if (cur != attr) {
xmlFree(ID);
return(-1);
}
The 'table' pointer, which refers to a document's 'ids' table has inserted
into it xmlIDPtrs by the xmlAddID function. Here, they are being cast into
a xmlAttrPtr and compared with the xmlAttrPtr parameter 'attr'.
Also, I would like to request either some documentation or functionality changes of the following functions: xmlSetProp (and xmlNewProp) - these functions will NOT call xmlRemoveID or xmlAddID on any ID properties. I think that they either should, if necessary, or it be explicitly documented that they do not.Substanciate why you think it's wrong. I do think the current behaviour is the right one w.r.t. the XML spec. The fact that an attribute is of type ID is dependant on the DTD, this does not change with xmlSetProp/xmlNewProp. So the IDness property is directly related to 1/ the definition in the DTD 2/ the presence of the attribute on the element.
I understand that an attribute being an ID is dependent on the DTD. What I am trying to get at is that xmlUnsetProp and xmlFreeProp will _check_ the property to see if it as an ID with xmlIsID and then (attempt to) remove the internal hash table reference with xmlRemoveID, whereas xmlSetProp will not perform any such check. To me, it would seem logical to make this behavior symmetrical with xmlUnsetProp, or document what each does and doesn't do more clearly. Such an addition would involve checking the document for DTD, checking the target elements 'ATTLIST's for the attribute named and using xmlAddID if it would be required. I can appreciate that perhaps not many people would find much use in this additional convenience. It is not really a problem for me either way. Please find attached a sample program and xml file that should demonstrate the xmlRemoveID IDs hash table problem. I have hopefully included enough commentary in the source code for you to be able to comprehend its behavoiral objectives. If I have misinterpreted the desired use of any of the function calls, then please accept my apologies and my polite request that you point out where I have gone wrong. If you need any further clarifications of the example, please don't hesitate to contact me. If you find the behavior to be incorrect aswell, then please refer to my original posting for a patch which should fix (or at the very least help you locate) the problem. I can well understand how this error has come to be present, as I have found little evidence for the xmlRemoveID function's usage in any programs. So, I hope this is of help. Thank you -bill ---- William Robinson baggins elitemail org http://www.hobbiton.cjb.net/
Attachment:
libxml-ids-error.c
Description: Text document
Attachment:
libxml-ids-error.xml
Description: Text document