how to write a CDATA node using libxml2?

问题:

I'm using libxml2 to read/write xml files. Now I'm trying to write a CDATA node.

Here is what I tried:

nodePtr = xmlNewChild( parentPtr, NULL, "foo", NULL );
xmlNodeSetContentLen( nodePtr, "<![CDATA[\nTesting 1 < 2\n]]>", len );

However, this results in the following encoded text:

<foo>&lt;![CDATA[
Testing 1 &lt; 2
]]&gt;</foo>

I'm thinking that perhaps there might be a CDATA-specific libxml2 API. Or maybe I have to call something else to tell libxml2 not to automatically encode the node content?

回答1:

Figured it out. The trick is in knowing that CDATA text content is actually a child and not a part of the current node, and the critical API to call is xmlNewCDataBlock(). Using the same example as above:

nodePtr = xmlNewChild( parentPtr, NULL, "foo", NULL );
cdataPtr = xmlNewCDataBlock( doc, "Testing 1 < 2", 13 );
xmlAddChild( nodePtr, cdataPtr );

This will produce the following xml:

<foo><![CDATA[Testing 1 < 2]]></foo>

回答2:

I cannot say for all versions of libxml2, but according to libxml2-2.9.4 the doc part of returning node of xmlNewChild comes from its parent. Also the parent of child node returned from xmlNewCDataBlock is set by doc parameter. So the following would be a good practice:

const char str[] = "said the kitty";
xmlNodePtr node = xmlNewNode(NULL, BAD_CAST "meow");
xmlNodePtr cdata_node = xmlNewCDataBlock(node->doc, BAD_CAST str, strlen(str));
xmlAddChild(node, cdata_node);

The resulting xml is

<meow><![CDATA[said the kitty]]></meow>

And it would not matter if node is part of an xmlDoc or not