[Expat-checkins] expat/lib xmlparse.c,1.44,1.45

[Karl Waclawek]

> On Thu, Jun 13, 2002 at 02:53:40PM -0400, Karl Waclawek wrote:
> > Btw, to be honest, I don't like this "reverse indentation" style
> > just to make it fit into 80 character lines.
> 
> Me neither; I didn't know it was present in the Expat sources. Do you have a
> pointer?

I am not sure I have expressed myself understandably.
What I mean is: when you read xmlparse.c (especially), indentations
seems to go to the left, as the "block nesting level" increases - 
apparently to stay within a line length of 80.

Just open up xmlparse.c and check any of the switch cases in doContent,
for example. Follow the indentations - you will see.

On the other hand - maybe this is standard C style.
As you may have guessed, I am not really a hardcore C/C++ programmer.

Here is a snippet:

<snippet>
    case XML_TOK_EMPTY_ELEMENT_NO_ATTS:
      if (startElementHandler || endElementHandler) {
 const char *rawName = s + enc->minBytesPerChar;
 enum XML_Error result;
 BINDING *bindings = NULL;
 TAG_NAME name;
 name.str = poolStoreString(&tempPool, enc, rawName,
       rawName + XmlNameLength(enc, rawName));
 if (!name.str)
   return XML_ERROR_NO_MEMORY;
 poolFinish(&tempPool);
 result = storeAtts(parser, enc, s, &name, &bindings);
 if (result)
   return result;
 poolFinish(&tempPool);
 if (startElementHandler)
   startElementHandler(handlerArg, name.str, (const XML_Char **)atts);
 if (endElementHandler) {
   if (startElementHandler)
     *eventPP = *eventEndPP;
   endElementHandler(handlerArg, name.str);
 }
 poolClear(&tempPool);
 while (bindings) {
   BINDING *b = bindings;
   if (endNamespaceDeclHandler)
     endNamespaceDeclHandler(handlerArg, b->prefix->name);
   bindings = bindings->nextTagBinding;
   b->nextTagBinding = freeBindingList;
   freeBindingList = b;
   b->prefix->binding = b->prevPrefixBinding;
 }
      }
      else if (defaultHandler)
 reportDefault(parser, enc, s, next);
      if (tagLevel == 0)
 return epilogProcessor(parser, next, end, nextPtr);
      break;
    case XML_TOK_END_TAG:
      if (tagLevel == startTagLevel)
        return XML_ERROR_ASYNC_ENTITY;
      else {
 int len;
 const char *rawName;
 TAG *tag = tagStack;
 tagStack = tag->parent;
 tag->parent = freeTagList;
 freeTagList = tag;
 rawName = s + enc->minBytesPerChar*2;
 len = XmlNameLength(enc, rawName);
 if (len != tag->rawNameLength
     || memcmp(tag->rawName, rawName, len) != 0) {
   *eventPP = rawName;
   return XML_ERROR_TAG_MISMATCH;
 }
 --tagLevel;
</snippet>
 
> IMO, I'd be quite happy to toss anything like that. In fact, a big revamp to
> make everything consistent would be quite nice.

Let's check with Fred. Maybe he likes it the way it is.
 
> One caution (and I guess this applies to untabifying): if people generate
> patches against 1.95.3, then we won't be able to apply them to the source
> code. Thus, it is usually safest to do big revamps right before a release
> (when the patch queue has been emptied).

Absolutely. 1.95.4 is coming up soon, it seems.
I only have the InternalEntityRefhandler still on my plate,
and it is currently being discussed on the mailing list.

Karl