Re: [Libvirt-cim] [PATCH 2 of 3] [RFC][CU] Add EI CIMXML parser

10 Feb 2008


      On Fri, Feb 08, 2008 at 03:18:00PM -0800, Dan Smith wrote:
...
# HG changeset patch
# User Dan Smith <danms@us.ibm.com>
# Date 1202512655 28800
# Node ID f6559a2f55abab85baeabd56faaf7804c3ea30f3
# Parent  03126c9af41607dcaa8b197736dd5059c108261a
[RFC][CU] Add EI CIMXML parser
This handles strings, bools, integers, and arrays of those.  It doesn't yet
handle nested objects, as we will need for RASD.HostResource[].  I need
to do some further investigation of what the XML for that will look like.
I have tested this by taking an existing XML for DefineSystem() and
changing the MemResourceAllocationSettingData field to this:
<INSTANCE CLASSNAME="Xen_MemResourceAllocationSettingData">
    <PROPERTY NAME="CreationClassName" TYPE="string">
      <VALUE>Xen_MemResourceAllocationSettingData</VALUE>
    </PROPERTY>
    <PROPERTY NAME="InstanceID" TYPE="string">
      <VALUE>test/mem</VALUE>
    </PROPERTY>
    <PROPERTY NAME="ResourceType" TYPE="uint16">
       <VALUE>4</VALUE>
    </PROPERTY>
    <PROPERTY NAME="VirtualQuantity" TYPE="uint64">
       <VALUE>256 </VALUE>
    </PROPERTY>
    <PROPERTY.ARRAY NAME="Foo" TYPE="string">
      <VALUE.ARRAY>
        <VALUE>Bar</VALUE>
        <VALUE>Baz</VALUE>
      </VALUE.ARRAY>
    </PROPERTY.ARRAY>
  </INSTANCE>
Oh the escaping fun :-)

[...]
...
+static CMPIType get_node_value(const CMPIBroker *broker,
+                               xmlNode *node,
+                               const char *tstr,
+                               CMPIValue *val)
+{
+        CMPIType type = CMPI_null;
+        char *content = NULL;
+
+        if (node->type != XML_ELEMENT_NODE) {
+                CU_DEBUG("Non-element node: %s", node->name);
+                goto out;
+        }
+
+        if (!STREQC((char *)node->name, "value")) {
+                CU_DEBUG("Expected <VALUE> but got <%s>", node->name);
+                goto out;
+        }
Assuming I understand the example, there is also no namespace in those
element, checking node->ns == NULL might be a good idea (or not if namespaced)
...
+        content = get_content(node);
+        CU_DEBUG("Node content: %s", content);
+        type = get_property_value(broker, tstr, content, val);
+        free(content);
+ out:
+        return type;
+}
+
+static CMPIType parse_array(const CMPIBroker *broker,
+                            const char *tstr,
+                            xmlNode *root,
+                            CMPIArray **array)
+{
+        xmlNode *value;
+        CMPIValue *list = NULL;
+        int size = 0;
+        int cur = 0;
+        CMPIStatus s;
+        CMPIType type = CMPI_null;
+        int i;
+
+        for (value = root->children; value; value = value->next) {
+
+                if (value->type != XML_ELEMENT_NODE)
+                        continue;
+
+                if (cur == size) {
+                        CMPIValue *tmp;
+
+                        size += 4;
+                        tmp = realloc(list, sizeof(CMPIValue) * size);
when i have realloc in loop I usually do things like size *= 2 to escape
a linear growing model.

[..]
...
+        for (child = root->children; child; child = child->next) {
+                if (child->type == XML_ELEMENT_NODE){
check namespace here ?
...
+                        if (STREQC((char *)child->name, "property"))
+                                parse_property(broker, child, *inst);
+        CU_DEBUG("Escape is: %s", escape);
+
+        if (STREQC(escape, "<"))
+                *result = '<';
+        else if (STREQC(escape, ">"))
+                *result = '>';
+        else if (STREQC(escape, """))
+                *result = '\"';
Add & -> & and ' -> ''' would be good for completeness

[...]
...
+int cu_parse_ei_xml(const CMPIBroker *broker,
+                    const char *ns,
+                    const char *xml,
+                    CMPIInstance **instance)
+{
+        xmlDoc *doc = NULL;
+        xmlNode *root = NULL;
+        int ret = 0;
+
+        doc = xmlReadMemory(xml, strlen(xml), NULL, NULL, 0);
I would add (XML_PARSE_NOENT | XML_PARSE_NOCDATA | XML_PARSE_NONET)
to ask the parser to replace entities references, change CDATA section into
normal text and forbid network access (should not occur, but better safe than
sorry :-)
...
+        if (doc == NULL) {
+                CU_DEBUG("Error reading decoded XML from memory");
+                goto out;
+        }
+
+        root = xmlDocGetRootElement(doc);
+        if (root == NULL) {
+                CU_DEBUG("Error getting root XML node");
+                goto out;
+        }
+
Looks fine, maybe a small check that no memory allocation is lost is in
order though, it's easy to forget something.

Daniel

-- 
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard      | virtualization library  http://libvirt.org/
veillard@redhat.com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine  http://rpmfind.net/