started messing seriously with per-document dict and element and attribute
* SAX2.c parser.c parserInternals.c xmlreader.c: started messing
seriously with per-document dict and element and attribute nodes
reuse in the xmlReader. This seems to lead to an interesting
speedup of the xmlReader already.
Daniel
diff --git a/ChangeLog b/ChangeLog
index 1bb2fb9..fc91395 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,10 @@
+Wed Sep 17 12:23:41 CEST 2003 Daniel Veillard <daniel@veillard.com>
+
+ * SAX2.c parser.c parserInternals.c xmlreader.c: started messing
+ seriously with per-document dict and element and attribute nodes
+ reuse in the xmlReader. This seems to lead to an interesting
+ speedup of the xmlReader already.
+
Wed Sep 17 01:07:56 CEST 2003 Daniel Veillard <daniel@veillard.com>
* dict.c include/libxml/dict.h: do string allocations in large
diff --git a/SAX2.c b/SAX2.c
index d37bbbd..ff26fcf 100644
--- a/SAX2.c
+++ b/SAX2.c
@@ -1642,6 +1642,17 @@
else
ret->name = xmlStrdup(localname);
+ /* link at the end to preserv order, TODO speed up with a last */
+ if (ctxt->node->properties == NULL) {
+ ctxt->node->properties = ret;
+ } else {
+ xmlAttrPtr prev = ctxt->node->properties;
+
+ while (prev->next != NULL) prev = prev->next;
+ prev->next = ret;
+ ret->prev = prev;
+ }
+
if ((__xmlRegisterCallbacks) && (xmlRegisterNodeDefaultValue))
xmlRegisterNodeDefaultValue((xmlNodePtr)ret);
} else {
diff --git a/parser.c b/parser.c
index 944a656..0840de3 100644
--- a/parser.c
+++ b/parser.c
@@ -648,7 +648,8 @@
goto mem_error;
}
- xmlHashAddEntry2(ctxt->attsSpecial, fullname, fullattr, (void *) type);
+ xmlHashAddEntry2(ctxt->attsSpecial, fullname, fullattr,
+ (void *) (long) type);
return;
mem_error:
@@ -919,7 +920,7 @@
ctxt->name = value;
ctxt->pushTab[ctxt->nameNr * 3] = (void *) prefix;
ctxt->pushTab[ctxt->nameNr * 3 + 1] = (void *) URI;
- ctxt->pushTab[ctxt->nameNr * 3 + 2] = (void *) nsNr;
+ ctxt->pushTab[ctxt->nameNr * 3 + 2] = (void *) (long) nsNr;
return (ctxt->nameNr++);
mem_error:
xmlErrMemory(ctxt, NULL);
@@ -10902,6 +10903,8 @@
ctxt->userData = user_data;
else
ctxt->userData = ctxt;
+ if (ctxt->dict != NULL) xmlDictFree(ctxt->dict);
+ ctxt->dict = oldctxt->dict;
oldsax = ctxt->sax;
ctxt->sax = oldctxt->sax;
@@ -10912,6 +10915,7 @@
newDoc = xmlNewDoc(BAD_CAST "1.0");
if (newDoc == NULL) {
ctxt->sax = oldsax;
+ ctxt->dict = NULL;
xmlFreeParserCtxt(ctxt);
return(-1);
}
@@ -10924,6 +10928,7 @@
BAD_CAST "pseudoroot", NULL);
if (ctxt->myDoc->children == NULL) {
ctxt->sax = oldsax;
+ ctxt->dict = NULL;
xmlFreeParserCtxt(ctxt);
if (newDoc != NULL)
xmlFreeDoc(newDoc);
@@ -10941,6 +10946,7 @@
*/
ctxt->loadsubset |= XML_SKIP_IDS;
}
+ ctxt->dictNames = oldctxt->dictNames;
xmlParseContent(ctxt);
if ((RAW == '<') && (NXT(1) == '/')) {
@@ -10987,6 +10993,7 @@
}
ctxt->sax = oldsax;
+ ctxt->dict = NULL;
xmlFreeParserCtxt(ctxt);
if (newDoc != NULL)
xmlFreeDoc(newDoc);
diff --git a/parserInternals.c b/parserInternals.c
index 7c9e471..2dfeca9 100644
--- a/parserInternals.c
+++ b/parserInternals.c
@@ -2383,6 +2383,26 @@
xmlHashFree(ctxt->attsDefault, (xmlHashDeallocator) xmlFree);
if (ctxt->attsSpecial != NULL)
xmlHashFree(ctxt->attsSpecial, NULL);
+ if (ctxt->freeElems != NULL) {
+ xmlNodePtr cur, next;
+
+ cur = ctxt->freeElems;
+ while (cur != NULL) {
+ next = cur->next;
+ xmlFree(cur);
+ cur = next;
+ }
+ }
+ if (ctxt->freeAttrs != NULL) {
+ xmlAttrPtr cur, next;
+
+ cur = ctxt->freeAttrs;
+ while (cur != NULL) {
+ next = cur->next;
+ xmlFree(cur);
+ cur = next;
+ }
+ }
#ifdef LIBXML_CATALOG_ENABLED
if (ctxt->catalogs != NULL)
diff --git a/xmlreader.c b/xmlreader.c
index 54778bd..eea9a28 100644
--- a/xmlreader.c
+++ b/xmlreader.c
@@ -135,6 +135,345 @@
static const char *xmlTextReaderIsEmpty = "This element is empty";
+/************************************************************************
+ * *
+ * Our own version of the freeing routines as we recycle nodes *
+ * *
+ ************************************************************************/
+
+static void xmlTextReaderFreeNode(xmlTextReaderPtr reader, xmlNodePtr cur);
+static void xmlTextReaderFreeNodeList(xmlTextReaderPtr reader, xmlNodePtr cur);
+
+/**
+ * xmlTextReaderFreeEntityWrapper:
+ * @entity: An entity
+ * @name: its name
+ *
+ * Deallocate the memory used by an entities in the hash table.
+ */
+static void
+xmlTextReaderFreeEntityWrapper(xmlEntityPtr entity,
+ const xmlChar *name ATTRIBUTE_UNUSED) {
+ if (entity == NULL) return;
+
+ if ((entity->children) && (entity->owner == 1) &&
+ (entity == (xmlEntityPtr) entity->children->parent)) {
+ xmlDocPtr doc;
+ xmlTextReaderPtr reader = NULL;
+ doc = entity->doc;
+ if (doc != NULL)
+ reader = doc->_private;
+ xmlTextReaderFreeNodeList(reader, entity->children);
+ }
+ if (entity->name != NULL)
+ xmlFree((char *) entity->name);
+ if (entity->ExternalID != NULL)
+ xmlFree((char *) entity->ExternalID);
+ if (entity->SystemID != NULL)
+ xmlFree((char *) entity->SystemID);
+ if (entity->URI != NULL)
+ xmlFree((char *) entity->URI);
+ if (entity->content != NULL)
+ xmlFree((char *) entity->content);
+ if (entity->orig != NULL)
+ xmlFree((char *) entity->orig);
+ xmlFree(entity);
+}
+
+/**
+ * xmlTextReaderFreeEntitiesTable:
+ * @table: An entity table
+ *
+ * Deallocate the memory used by an entities hash table.
+ */
+static void
+xmlTextReaderFreeEntitiesTable(xmlEntitiesTablePtr table) {
+ xmlHashFree(table, (xmlHashDeallocator) xmlTextReaderFreeEntityWrapper);
+}
+
+/**
+ * xmlTextReaderFreeDtd:
+ * @cur: the DTD structure to free up
+ *
+ * Free a DTD structure.
+ */
+static void
+xmlTextReaderFreeDtd(xmlTextReaderPtr reader, xmlDtdPtr cur) {
+ if (cur == NULL) return;
+
+ if (cur->children != NULL) {
+ xmlNodePtr next, c = cur->children;
+
+ /*
+ * Cleanup all the DTD comments they are not in the DTD
+ * indexes.
+ */
+ while (c != NULL) {
+ next = c->next;
+ if ((c->type == XML_COMMENT_NODE) || (c->type == XML_PI_NODE)) {
+ xmlUnlinkNode(c);
+ xmlTextReaderFreeNode(reader, c);
+ }
+ c = next;
+ }
+ }
+
+ if (cur->name != NULL) xmlFree((char *) cur->name);
+ if (cur->SystemID != NULL) xmlFree((char *) cur->SystemID);
+ if (cur->ExternalID != NULL) xmlFree((char *) cur->ExternalID);
+ /* TODO !!! */
+ if (cur->notations != NULL)
+ xmlFreeNotationTable((xmlNotationTablePtr) cur->notations);
+
+ if (cur->elements != NULL)
+ xmlFreeElementTable((xmlElementTablePtr) cur->elements);
+ if (cur->attributes != NULL)
+ xmlFreeAttributeTable((xmlAttributeTablePtr) cur->attributes);
+ if (cur->pentities != NULL)
+ xmlFreeEntitiesTable((xmlEntitiesTablePtr) cur->pentities);
+
+ if (cur->entities != NULL)
+ xmlTextReaderFreeEntitiesTable((xmlEntitiesTablePtr) cur->entities);
+
+ xmlFree(cur);
+}
+
+/**
+ * xmlTextReaderFreeProp:
+ * @reader: the xmlTextReaderPtr used
+ * @cur: the node
+ *
+ * Free a node.
+ */
+static void
+xmlTextReaderFreeProp(xmlTextReaderPtr reader, xmlAttrPtr cur) {
+ if (cur == NULL) return;
+
+ /* Check for ID removal -> leading to invalid references ! */
+ if ((cur->parent != NULL) && (cur->parent->doc != NULL) &&
+ ((cur->parent->doc->intSubset != NULL) ||
+ (cur->parent->doc->extSubset != NULL))) {
+ if (xmlIsID(cur->parent->doc, cur->parent, cur))
+ xmlRemoveID(cur->parent->doc, cur);
+ }
+ if (cur->children != NULL)
+ xmlTextReaderFreeNodeList(reader, cur->children);
+
+ if ((reader != NULL) && (reader->ctxt != NULL) &&
+ (xmlDictOwns(reader->ctxt->dict, cur->name) != 1) &&
+ (cur->name != NULL))
+ xmlFree((xmlChar *)cur->name);
+ if ((reader != NULL) && (reader->ctxt != NULL)) {
+ cur->next = reader->ctxt->freeAttrs;
+ reader->ctxt->freeAttrs = cur;
+ } else {
+ xmlFree(cur);
+ }
+}
+
+/**
+ * xmlTextReaderFreePropList:
+ * @reader: the xmlTextReaderPtr used
+ * @cur: the first property in the list
+ *
+ * Free a property and all its siblings, all the children are freed too.
+ */
+static void
+xmlTextReaderFreePropList(xmlTextReaderPtr reader, xmlAttrPtr cur) {
+ xmlAttrPtr next;
+ if (cur == NULL) return;
+ while (cur != NULL) {
+ next = cur->next;
+ xmlTextReaderFreeProp(reader, cur);
+ cur = next;
+ }
+}
+
+/**
+ * xmlTextReaderFreeNodeList:
+ * @reader: the xmlTextReaderPtr used
+ * @cur: the first node in the list
+ *
+ * Free a node and all its siblings, this is a recursive behaviour, all
+ * the children are freed too.
+ */
+static void
+xmlTextReaderFreeNodeList(xmlTextReaderPtr reader, xmlNodePtr cur) {
+ xmlNodePtr next;
+ if (cur == NULL) return;
+ if (cur->type == XML_NAMESPACE_DECL) {
+ xmlFreeNsList((xmlNsPtr) cur);
+ return;
+ }
+ if ((cur->type == XML_DOCUMENT_NODE) ||
+ (cur->type == XML_HTML_DOCUMENT_NODE)) {
+ xmlFreeDoc((xmlDocPtr) cur);
+ return;
+ }
+ while (cur != NULL) {
+ next = cur->next;
+ /* unroll to speed up freeing the document */
+ if (cur->type != XML_DTD_NODE) {
+
+ if ((cur->children != NULL) &&
+ (cur->type != XML_ENTITY_REF_NODE))
+ xmlTextReaderFreeNodeList(reader, cur->children);
+ if (((cur->type == XML_ELEMENT_NODE) ||
+ (cur->type == XML_XINCLUDE_START) ||
+ (cur->type == XML_XINCLUDE_END)) &&
+ (cur->properties != NULL))
+ xmlTextReaderFreePropList(reader, cur->properties);
+ if ((cur->type != XML_ELEMENT_NODE) &&
+ (cur->type != XML_XINCLUDE_START) &&
+ (cur->type != XML_XINCLUDE_END) &&
+ (cur->type != XML_ENTITY_REF_NODE)) {
+ if (cur->content != NULL) xmlFree(cur->content);
+ }
+ if (((cur->type == XML_ELEMENT_NODE) ||
+ (cur->type == XML_XINCLUDE_START) ||
+ (cur->type == XML_XINCLUDE_END)) &&
+ (cur->nsDef != NULL))
+ xmlFreeNsList(cur->nsDef);
+
+ /*
+ * we don't free element names here they are interned now
+ */
+ if (cur->type == XML_ELEMENT_NODE) {
+ if ((reader != NULL) && (reader->ctxt != NULL) &&
+ (xmlDictOwns(reader->ctxt->dict, cur->name) != 1) &&
+ (cur->name != NULL))
+ xmlFree((xmlChar *)cur->name);
+ } else if ((cur->type != XML_TEXT_NODE) &&
+ (cur->type != XML_COMMENT_NODE) &&
+ (cur->name != NULL))
+ xmlFree((xmlChar *)cur->name);
+ if ((cur->type == XML_ELEMENT_NODE) &&
+ (reader != NULL) && (reader->ctxt != NULL)) {
+ cur->next = reader->ctxt->freeElems;
+ reader->ctxt->freeElems = cur;
+ } else {
+ xmlFree(cur);
+ }
+ }
+ cur = next;
+ }
+}
+
+/**
+ * xmlTextReaderFreeNode:
+ * @reader: the xmlTextReaderPtr used
+ * @cur: the node
+ *
+ * Free a node, this is a recursive behaviour, all the children are freed too.
+ * This doesn't unlink the child from the list, use xmlUnlinkNode() first.
+ */
+static void
+xmlTextReaderFreeNode(xmlTextReaderPtr reader, xmlNodePtr cur) {
+ if (cur->type == XML_DTD_NODE) {
+ xmlTextReaderFreeDtd(reader, (xmlDtdPtr) cur);
+ return;
+ }
+ if (cur->type == XML_NAMESPACE_DECL) {
+ xmlFreeNs((xmlNsPtr) cur);
+ return;
+ }
+ if (cur->type == XML_ATTRIBUTE_NODE) {
+ xmlTextReaderFreeProp(reader, (xmlAttrPtr) cur);
+ return;
+ }
+
+ if ((cur->children != NULL) &&
+ (cur->type != XML_ENTITY_REF_NODE))
+ xmlTextReaderFreeNodeList(reader, cur->children);
+ if (((cur->type == XML_ELEMENT_NODE) ||
+ (cur->type == XML_XINCLUDE_START) ||
+ (cur->type == XML_XINCLUDE_END)) &&
+ (cur->properties != NULL))
+ xmlTextReaderFreePropList(reader, cur->properties);
+ if ((cur->type != XML_ELEMENT_NODE) &&
+ (cur->type != XML_XINCLUDE_START) &&
+ (cur->type != XML_XINCLUDE_END) &&
+ (cur->type != XML_ENTITY_REF_NODE)) {
+ if (cur->content != NULL) xmlFree(cur->content);
+ }
+ if (((cur->type == XML_ELEMENT_NODE) ||
+ (cur->type == XML_XINCLUDE_START) ||
+ (cur->type == XML_XINCLUDE_END)) &&
+ (cur->nsDef != NULL))
+ xmlFreeNsList(cur->nsDef);
+
+ /*
+ * we don't free names here they are interned now
+ */
+ if (cur->type == XML_ELEMENT_NODE) {
+ if ((reader != NULL) && (reader->ctxt != NULL) &&
+ (xmlDictOwns(reader->ctxt->dict, cur->name) != 1) &&
+ (cur->name != NULL))
+ xmlFree((xmlChar *)cur->name);
+ } else if ((cur->type != XML_TEXT_NODE) &&
+ (cur->type != XML_COMMENT_NODE) &&
+ (cur->name != NULL))
+ xmlFree((xmlChar *)cur->name);
+ if ((cur->type == XML_ELEMENT_NODE) &&
+ (reader != NULL) && (reader->ctxt != NULL)) {
+ cur->next = reader->ctxt->freeElems;
+ reader->ctxt->freeElems = cur;
+ } else {
+ xmlFree(cur);
+ }
+
+}
+
+/**
+ * xmlTextReaderFreeDoc:
+ * @reader: the xmlTextReaderPtr used
+ * @cur: pointer to the document
+ *
+ * Free up all the structures used by a document, tree included.
+ */
+static void
+xmlTextReaderFreeDoc(xmlTextReaderPtr reader, xmlDocPtr cur) {
+ xmlDtdPtr extSubset, intSubset;
+
+ if (cur == NULL) return;
+
+ /*
+ * Do this before freeing the children list to avoid ID lookups
+ */
+ if (cur->ids != NULL) xmlFreeIDTable((xmlIDTablePtr) cur->ids);
+ cur->ids = NULL;
+ if (cur->refs != NULL) xmlFreeRefTable((xmlRefTablePtr) cur->refs);
+ cur->refs = NULL;
+ extSubset = cur->extSubset;
+ intSubset = cur->intSubset;
+ if (intSubset == extSubset)
+ extSubset = NULL;
+ if (extSubset != NULL) {
+ xmlUnlinkNode((xmlNodePtr) cur->extSubset);
+ cur->extSubset = NULL;
+ xmlTextReaderFreeDtd(reader, extSubset);
+ }
+ if (intSubset != NULL) {
+ xmlUnlinkNode((xmlNodePtr) cur->intSubset);
+ cur->intSubset = NULL;
+ xmlTextReaderFreeDtd(reader, intSubset);
+ }
+
+ if (cur->children != NULL) xmlTextReaderFreeNodeList(reader, cur->children);
+
+ if (cur->version != NULL) xmlFree((char *) cur->version);
+ if (cur->name != NULL) xmlFree((char *) cur->name);
+ if (cur->encoding != NULL) xmlFree((char *) cur->encoding);
+ if (cur->oldNs != NULL) xmlFreeNsList(cur->oldNs);
+ if (cur->URL != NULL) xmlFree((char *) cur->URL);
+ xmlFree(cur);
+}
+
+/************************************************************************
+ * *
+ * The reader core parser *
+ * *
+ ************************************************************************/
#ifdef DEBUG_READER
static void
xmlTextReaderDebug(xmlTextReaderPtr reader) {
@@ -795,6 +1134,8 @@
} while ((reader->ctxt->node == NULL) &&
((reader->mode != XML_TEXTREADER_MODE_EOF) &&
(reader->mode != XML_TEXTREADER_DONE)));
+ if (reader->ctxt->myDoc != NULL)
+ reader->ctxt->myDoc->_private = reader;
if (reader->ctxt->node == NULL) {
if (reader->ctxt->myDoc != NULL) {
reader->node = reader->ctxt->myDoc->children;
@@ -875,7 +1216,7 @@
(reader->node->prev->type != XML_DTD_NODE)) {
xmlNodePtr tmp = reader->node->prev;
xmlUnlinkNode(tmp);
- xmlFreeNode(tmp);
+ xmlTextReaderFreeNode(reader, tmp);
}
goto node_found;
@@ -908,7 +1249,7 @@
*/
if (oldnode->type != XML_DTD_NODE) {
xmlUnlinkNode(oldnode);
- xmlFreeNode(oldnode);
+ xmlTextReaderFreeNode(reader, oldnode);
}
goto node_end;
@@ -1097,6 +1438,7 @@
return(NULL);
}
+#if 0
/**
* xmlTextReaderReadBase64:
* @reader: the xmlTextReaderPtr used
@@ -1112,8 +1454,10 @@
* instance is not positioned on an element or -1 in case of error.
*/
int
-xmlTextReaderReadBase64(xmlTextReaderPtr reader, unsigned char *array,
- int offset, int len) {
+xmlTextReaderReadBase64(xmlTextReaderPtr reader,
+ unsigned char *array ATTRIBUTE_UNUSED,
+ int offset ATTRIBUTE_UNUSED,
+ int len ATTRIBUTE_UNUSED) {
if ((reader == NULL) || (reader->ctxt == NULL))
return(-1);
if (reader->ctxt->wellFormed != 1)
@@ -1140,8 +1484,10 @@
* instance is not positioned on an element or -1 in case of error.
*/
int
-xmlTextReaderReadBinHex(xmlTextReaderPtr reader, unsigned char *array,
- int offset, int len) {
+xmlTextReaderReadBinHex(xmlTextReaderPtr reader,
+ unsigned char *array ATTRIBUTE_UNUSED,
+ int offset ATTRIBUTE_UNUSED,
+ int len ATTRIBUTE_UNUSED) {
if ((reader == NULL) || (reader->ctxt == NULL))
return(-1);
if (reader->ctxt->wellFormed != 1)
@@ -1152,6 +1498,7 @@
TODO
return(0);
}
+#endif
/************************************************************************
* *
@@ -1235,6 +1582,7 @@
}
ret->ctxt->_private = ret;
ret->ctxt->linenumbers = 1;
+ ret->ctxt->dictNames = 1;
ret->allocs = XML_TEXTREADER_CTXT;
/*
* use the parser dictionnary to allocate all elements and attributes names
@@ -1297,7 +1645,7 @@
#endif
if (reader->ctxt != NULL) {
if (reader->ctxt->myDoc != NULL) {
- xmlFreeDoc(reader->ctxt->myDoc);
+ xmlTextReaderFreeDoc(reader, reader->ctxt->myDoc);
reader->ctxt->myDoc = NULL;
}
if ((reader->ctxt->vctxt.vstateTab != NULL) &&
@@ -2937,6 +3285,7 @@
* Utilities *
* *
************************************************************************/
+#ifdef NOT_USED_YET
/**
* xmlBase64Decode:
* @in: the input buffer
@@ -3111,3 +3460,4 @@
}
#endif
+#endif /* NOT_USED_YET */