pattern.c

xml开源解析代码.版本为libxml2-2.6.29,可支持GB3212.网络消息发送XML时很有用.

 *
 * Returns -1 in case of failure and 0 in case of success.
 */
static int
xmlStreamCompile(xmlPatternPtr comp) {
    xmlStreamCompPtr stream;
    int i, s = 0, root = 0, flags = 0, prevs = -1;
    xmlStepOp step;

    if ((comp == NULL) || (comp->steps == NULL))
        return(-1);
    /*
     * special case for .
     */
    if ((comp->nbStep == 1) &&
        (comp->steps[0].op == XML_OP_ELEM) &&
	(comp->steps[0].value == NULL) &&
	(comp->steps[0].value2 == NULL)) {
	stream = xmlNewStreamComp(0);
	if (stream == NULL)
	    return(-1);
	/* Note that the stream will have no steps in this case. */
	stream->flags |= XML_STREAM_FINAL_IS_ANY_NODE;
	comp->stream = stream;
	return(0);
    }

    stream = xmlNewStreamComp((comp->nbStep / 2) + 1);
    if (stream == NULL)
        return(-1);
    if (comp->dict != NULL) {
        stream->dict = comp->dict;
	xmlDictReference(stream->dict);
    }

    i = 0;        
    if (comp->flags & PAT_FROM_ROOT)
	stream->flags |= XML_STREAM_FROM_ROOT;

    for (;i < comp->nbStep;i++) {
	step = comp->steps[i];
        switch (step.op) {
	    case XML_OP_END:
	        break;
	    case XML_OP_ROOT:
	        if (i != 0)
		    goto error;
		root = 1;
		break;
	    case XML_OP_NS:
		s = xmlStreamCompAddStep(stream, NULL, step.value,
		    XML_ELEMENT_NODE, flags);		
		if (s < 0)
		    goto error;
		prevs = s;
		flags = 0;		
		break;	    
	    case XML_OP_ATTR:
		flags |= XML_STREAM_STEP_ATTR;
		prevs = -1;
		s = xmlStreamCompAddStep(stream,
		    step.value, step.value2, XML_ATTRIBUTE_NODE, flags);
		flags = 0;
		if (s < 0)
		    goto error;
		break;
	    case XML_OP_ELEM:		
	        if ((step.value == NULL) && (step.value2 == NULL)) {
		    /*
		    * We have a "." or "self::node()" here.
		    * Eliminate redundant self::node() tests like in "/./."
		    * or "//./"
		    * The only case we won't eliminate is "//.", i.e. if
		    * self::node() is the last node test and we had
		    * continuation somewhere beforehand.
		    */
		    if ((comp->nbStep == i + 1) &&
			(flags & XML_STREAM_STEP_DESC)) {
			/*
			* Mark the special case where the expression resolves
			* to any type of node.
			*/
			if (comp->nbStep == i + 1) {
			    stream->flags |= XML_STREAM_FINAL_IS_ANY_NODE;
			}
			flags |= XML_STREAM_STEP_NODE;			
			s = xmlStreamCompAddStep(stream, NULL, NULL,
			    XML_STREAM_ANY_NODE, flags);
			if (s < 0)
			    goto error;
			flags = 0;
			/*
			* If there was a previous step, mark it to be added to
			* the result node-set; this is needed since only
			* the last step will be marked as "final" and only
			* "final" nodes are added to the resulting set.
			*/
			if (prevs != -1) {
			    stream->steps[prevs].flags |= XML_STREAM_STEP_IN_SET;
			    prevs = -1;
			}
			break;	

		    } else {
			/* Just skip this one. */
			continue;
		    }
		}
		/* An element node. */		
	        s = xmlStreamCompAddStep(stream, step.value, step.value2,
		    XML_ELEMENT_NODE, flags);		
		if (s < 0)
		    goto error;
		prevs = s;
		flags = 0;		
		break;		
	    case XML_OP_CHILD:
		/* An element node child. */
	        s = xmlStreamCompAddStep(stream, step.value, step.value2,
		    XML_ELEMENT_NODE, flags);		
		if (s < 0)
		    goto error;
		prevs = s;
		flags = 0;
		break;	    
	    case XML_OP_ALL:
	        s = xmlStreamCompAddStep(stream, NULL, NULL,
		    XML_ELEMENT_NODE, flags);		
		if (s < 0)
		    goto error;
		prevs = s;
		flags = 0;
		break;
	    case XML_OP_PARENT:	
	        break;
	    case XML_OP_ANCESTOR:
		/* Skip redundant continuations. */
		if (flags & XML_STREAM_STEP_DESC)
		    break;
	        flags |= XML_STREAM_STEP_DESC;
		/*
		* Mark the expression as having "//".
		*/
		if ((stream->flags & XML_STREAM_DESC) == 0)
		    stream->flags |= XML_STREAM_DESC;
		break;
	}
    }    
    if ((! root) && (comp->flags & XML_PATTERN_NOTPATTERN) == 0) {
	/*
	* If this should behave like a real pattern, we will mark
	* the first step as having "//", to be reentrant on every
	* tree level.
	*/
	if ((stream->flags & XML_STREAM_DESC) == 0)
	    stream->flags |= XML_STREAM_DESC;

	if (stream->nbStep > 0) {
	    if ((stream->steps[0].flags & XML_STREAM_STEP_DESC) == 0)
		stream->steps[0].flags |= XML_STREAM_STEP_DESC;	    
	}
    }
    if (stream->nbStep <= s)
	goto error;
    stream->steps[s].flags |= XML_STREAM_STEP_FINAL;
    if (root)
	stream->steps[0].flags |= XML_STREAM_STEP_ROOT;
#ifdef DEBUG_STREAMING
    xmlDebugStreamComp(stream);
#endif
    comp->stream = stream;
    return(0);
error:
    xmlFreeStreamComp(stream);
    return(0);
}

/**
 * xmlNewStreamCtxt:
 * @size: the number of expected states
 *
 * build a new stream context
 *
 * Returns the new structure or NULL in case of error.
 */
static xmlStreamCtxtPtr
xmlNewStreamCtxt(xmlStreamCompPtr stream) {
    xmlStreamCtxtPtr cur;

    cur = (xmlStreamCtxtPtr) xmlMalloc(sizeof(xmlStreamCtxt));
    if (cur == NULL) {
	ERROR(NULL, NULL, NULL,
		"xmlNewStreamCtxt: malloc failed\n");
	return(NULL);
    }
    memset(cur, 0, sizeof(xmlStreamCtxt));
    cur->states = (int *) xmlMalloc(4 * 2 * sizeof(int));
    if (cur->states == NULL) {
	xmlFree(cur);
	ERROR(NULL, NULL, NULL,
	      "xmlNewStreamCtxt: malloc failed\n");
	return(NULL);
    }
    cur->nbState = 0;
    cur->maxState = 4;
    cur->level = 0;
    cur->comp = stream;
    cur->blockLevel = -1;
    return(cur);
}

/**
 * xmlFreeStreamCtxt:
 * @stream: the stream context
 *
 * Free the stream context
 */
void
xmlFreeStreamCtxt(xmlStreamCtxtPtr stream) {
    xmlStreamCtxtPtr next;

    while (stream != NULL) {
        next = stream->next;
        if (stream->states != NULL)
	    xmlFree(stream->states);
        xmlFree(stream);
	stream = next;
    }
}

/**
 * xmlStreamCtxtAddState:
 * @comp: the stream context
 * @idx: the step index for that streaming state
 *
 * Add a new state to the stream context
 *
 * Returns -1 in case of error or the state index if successful
 */
static int
xmlStreamCtxtAddState(xmlStreamCtxtPtr comp, int idx, int level) {
    int i;
    for (i = 0;i < comp->nbState;i++) {
        if (comp->states[2 * i] < 0) {
	    comp->states[2 * i] = idx;
	    comp->states[2 * i + 1] = level;
	    return(i);
	}
    }
    if (comp->nbState >= comp->maxState) {
        int *cur;

	cur = (int *) xmlRealloc(comp->states,
				 comp->maxState * 4 * sizeof(int));
	if (cur == NULL) {
	    ERROR(NULL, NULL, NULL,
		  "xmlNewStreamCtxt: malloc failed\n");
	    return(-1);
	}
	comp->states = cur;
        comp->maxState *= 2;
    }
    comp->states[2 * comp->nbState] = idx;
    comp->states[2 * comp->nbState++ + 1] = level;
    return(comp->nbState - 1);
}

/**
 * xmlStreamPushInternal:
 * @stream: the stream context
 * @name: the current name
 * @ns: the namespace name
 * @nodeType: the type of the node
 *
 * Push new data onto the stream. NOTE: if the call xmlPatterncompile()
 * indicated a dictionary, then strings for name and ns will be expected
 * to come from the dictionary.
 * Both @name and @ns being NULL means the / i.e. the root of the document.
 * This can also act as a reset.
 *
 * Returns: -1 in case of error, 1 if the current state in the stream is a
 *    match and 0 otherwise.
 */
static int
xmlStreamPushInternal(xmlStreamCtxtPtr stream,
		      const xmlChar *name, const xmlChar *ns,
		      int nodeType) {
    int ret = 0, err = 0, final = 0, tmp, i, m, match, stepNr, desc;
    xmlStreamCompPtr comp;
    xmlStreamStep step;
#ifdef DEBUG_STREAMING
    xmlStreamCtxtPtr orig = stream;
#endif

    if ((stream == NULL) || (stream->nbState < 0))
        return(-1);

    while (stream != NULL) {
	comp = stream->comp;

	if ((nodeType == XML_ELEMENT_NODE) &&
	    (name == NULL) && (ns == NULL)) {
	    /* We have a document node here (or a reset). */
	    stream->nbState = 0;
	    stream->level = 0;
	    stream->blockLevel = -1;
	    if (comp->flags & XML_STREAM_FROM_ROOT) {
		if (comp->nbStep == 0) {
		    /* TODO: We have a "/." here? */
		    ret = 1;
		} else {
		    if ((comp->nbStep == 1) &&
			(comp->steps[0].nodeType == XML_STREAM_ANY_NODE) &&
			(comp->steps[0].flags & XML_STREAM_STEP_DESC))
		    {
			/*
			* In the case of "//." the document node will match
			* as well.
			*/
			ret = 1;
		    } else if (comp->steps[0].flags & XML_STREAM_STEP_ROOT) {
			/* TODO: Do we need this ? */
			tmp = xmlStreamCtxtAddState(stream, 0, 0);
			if (tmp < 0)
			    err++;
		    }
		}
	    }
	    stream = stream->next;
	    continue; /* while */
	}

	/*
	* Fast check for ".".
	*/
	if (comp->nbStep == 0) {
	    /*
	     * / and . are handled at the XPath node set creation
	     * level by checking min depth
	     */
	    if (stream->flags & XML_PATTERN_XPATH) {
		stream = stream->next;
		continue; /* while */
	    }
	    /*
	    * For non-pattern like evaluation like XML Schema IDCs
	    * or traditional XPath expressions, this will match if
	    * we are at the first level only, otherwise on every level.
	    */
	    if ((nodeType != XML_ATTRIBUTE_NODE) &&
		(((stream->flags & XML_PATTERN_NOTPATTERN) == 0) ||
		(stream->level == 0))) {
		    ret = 1;		
	    }
	    stream->level++;
	    goto stream_next;
	}
	if (stream->blockLevel != -1) {
	    /*
	    * Skip blocked expressions.
	    */
    	    stream->level++;
	    goto stream_next;
	}

	if ((nodeType != XML_ELEMENT_NODE) &&
	    (nodeType != XML_ATTRIBUTE_NODE) &&
	    ((comp->flags & XML_STREAM_FINAL_IS_ANY_NODE) == 0)) {
	    /*
	    * No need to process nodes of other types if we don't
	    * resolve to those types.
	    * TODO: Do we need to block the context here?
	    */
	    stream->level++;
	    goto stream_next;
	}

	/*
	 * Check evolution of existing states
	 */
	i = 0;
	m = stream->nbState;
	while (i < m) {
	    if ((comp->flags & XML_STREAM_DESC) == 0) {
		/*
		* If there is no "//", then only the last
		* added state is of interest.
		*/
		stepNr = stream->states[2 * (stream->nbState -1)];
		/*
		* TODO: Security check, should not happen, remove it.
		*/
		if (stream->states[(2 * (stream->nbState -1)) + 1] <
		    stream->level) {
		    return (-1);
		}
		desc = 0;
		/* loop-stopper */
		i = m;
	    } else {
		/*
		* If there are "//", then we need to process every "//"
		* occuring in the states, plus any other state for this
		* level.
		*/		
		stepNr = stream->states[2 * i];

		/* TODO: should not happen anymore: dead states */
		if (stepNr < 0)
		    goto next_state;

		tmp = stream->states[(2 * i) + 1];

		/* skip new states just added */
		if (tmp > stream->level)
		    goto next_state;

		/* skip states at ancestor levels, except if "//" */
		desc = comp->steps[stepNr].flags & XML_STREAM_STEP_DESC;
		if ((tmp < stream->level) && (!desc))
		    goto next_state;
	    }
	    /* 
	    * Check for correct node-type.
	    */
	    step = comp->steps[stepNr];
	    if (step.nodeType != nodeType) {
		if (step.nodeType == XML_ATTRIBUTE_NODE) {
		    /*
		    * Block this expression for deeper evaluation.
		    */
		    if ((comp->flags & XML_STREAM_DESC) == 0)
			stream->blockLevel = stream->level +1;
		    goto next_state;
		} else if (step.nodeType != XML_STREAM_ANY_NODE)
		    goto next_state;
	    }	    
	    /*
	    * Compare local/namespace-name.
	    */
	    match = 0;
	    if (step.nodeType == XML_STREAM_ANY_NODE) {
		match = 1;
	    } else if (step.name == NULL) {
		if (step.ns == NULL) {
		    /*
		    * This lets through all elements/attributes.
		    */
		    match = 1;
		} else if (ns != NULL)
		    match = xmlStrEqual(step.ns, ns);
	    } else if (((step.ns != NULL) == (ns != NULL)) &&
		(name != NULL) &&
		(step.name[0] == name[0]) &&
		xmlStrEqual(step.name, name) &&
		((step.ns == ns) || xmlStrEqual(step.ns, ns)))
	    {
		match = 1;	    
	    }	 
#if 0 
/*
* TODO: Pointer comparison won't work, since not guaranteed that the given
*  values are in the same dict; especially if it's the namespace name,
*  normally coming from ns->href. We need a namespace dict mechanism !
*/
	    } else if (comp->dict) {
		if (step.name == NULL) {
		    if (step.ns == NULL)
			match = 1;
		    else
			match = (step.ns == ns);
		} else {
		    match = ((step.name == name) && (step.ns == ns));
		}
#endif /* if 0 ------------------------------------------------------- */	    
	    if (match) {		
		final = step.flags & XML_STREAM_STEP_FINAL;
		if (desc) {
		    if (final) {
			ret = 1;
		    } else {
			/* descending match create a new state */
			xmlStreamCtxtAddState(stream, stepNr + 1,
			                      stream->level + 1);
		    }
		} else {
		    if (final) {
			ret = 1;
		    } else {
			xmlStreamCtxtAddState(stream, stepNr + 1,
			                      stream->level + 1);
		    }
		}
		if ((ret != 1) && (step.flags & XML_STREAM_STEP_IN_SET)) {
		    /*
		    * Check if we have a special case like "foo/bar//.", where
		    * "foo" is selected as well.
		    */
		    ret = 1;
		}
	    }	    
	    if (((comp->flags & XML_STREAM_DESC) == 0) &&
		((! match) || final))  {
		/*
		* Mark this expression as blocked for any evaluation at
		* deeper levels. Note that this includes "/foo"
		* expressions if the *pattern* behaviour is used.
		*/
		stream->blockLevel = stream->level +1;
	    }
next_state:
	    i++;
	}

	stream->level++;

	/*
	* Re/enter the expression.
	* Don't reenter if it's an absolute expression like "/foo",
	*   except "//foo".
	*/
	step = comp->steps[0];