John Fleck jfleck@inkstain.net 2002, 2003 John Fleck 1 June 4, 2002 Initial draft 2 June 12, 2002 retrieving attribute value added 3 Aug. 31, 2002 freeing memory fix 4 Nov. 10, 2002 encoding discussion added 5 Dec. 15, 2002 more memory freeing changes 6 Jan. 26. 2003 add index 7 April 25, 2003 add compilation appendix 8 July 24, 2003 add XPath example 9 Feb. 14, 2004 Fix bug in XPath example 7 Aug. 24, 2004 Fix another bug in XPath example Libxml is a freely licensed C language library for handling XML, portable across a large number of platforms. This tutorial provides examples of its basic functions. Libxml is a C language library implementing functions for reading, creating and manipulating XML data. This tutorial provides example code and explanations of its basic functionality. Libxml and more details about its use are available on the project home page. Included there is complete API documentation. This tutorial is not meant to substitute for that complete documentation, but to illustrate the functions needed to use the library to perform basic operations. The tutorial is based on a simple XML application I use for articles I write. The format includes metadata and the body of the article. The example code in this tutorial demonstrates how to: Parse the document. Extract the text within a specified element. Add an element and its content. Add an attribute. Extract the value of an attribute. Full code for the examples is included in the appendices. Libxml declares a number of data types we will encounter repeatedly, hiding the messy stuff so you do not have to deal with it unless you have some specific need. xmlChar xmlChar A basic replacement for char, a byte in a UTF-8 encoded string. If your data uses another encoding, it must be converted to UTF-8 for use with libxml's functions. More information on encoding is available on the libxml encoding support web page. xmlDoc xmlDoc A structure containing the tree created by a parsed doc. xmlDocPtr is a pointer to the structure. xmlNodePtr xmlNodePtr and xmlNode A structure containing a single node. xmlNodePtr is a pointer to the structure, and is used in traversing the document tree. file parsing Parsing the file requires only the name of the file and a single function call, plus error checking. Full code: xmlDocPtr doc; xmlNodePtr cur; doc = xmlParseFile(docname); if (doc == NULL ) { fprintf(stderr,"Document not parsed successfully. \n"); return; } cur = xmlDocGetRootElement(doc); if (cur == NULL) { fprintf(stderr,"empty document\n"); xmlFreeDoc(doc); return; } if (xmlStrcmp(cur->name, (const xmlChar *) "story")) { fprintf(stderr,"document of the wrong type, root node != story"); xmlFreeDoc(doc); return; } Declare the pointer that will point to your parsed document. Declare a node pointer (you'll need this in order to interact with individual nodes). Check to see that the document was successfully parsed. If it was not, libxml will at this point register an error and stop. encoding One common example of an error at this point is improper handling of encoding. The XML standard requires documents stored with an encoding other than UTF-8 or UTF-16 to contain an explicit declaration of their encoding. If the declaration is there, libxml will automatically perform the necessary conversion to UTF-8 for you. More information on XML's encoding requirements is contained in the standard. Retrieve the document's root element. Check to make sure the document actually contains something. In our case, we need to make sure the document is the right type. "story" is the root type of the documents used in this tutorial. element retrieving content Retrieving the content of an element involves traversing the document tree until you find what you are looking for. In this case, we are looking for an element called "keyword" contained within element called "story". The process to find the node we are interested in involves tediously walking the tree. We assume you already have an xmlDocPtr called doc and an xmlNodPtr called cur. cur = cur->xmlChildrenNode; while (cur != NULL) { if ((!xmlStrcmp(cur->name, (const xmlChar *)"storyinfo"))){ parseStory (doc, cur); } cur = cur->next; } Get the first child node of cur. At this point, cur points at the document root, which is the element "story". This loop iterates through the elements that are children of "story", looking for one called "storyinfo". That is the element that will contain the "keywords" we are looking for. It uses the libxml string comparison function, xmlStrcmp. If there is a match, it calls the function parseStory. void parseStory (xmlDocPtr doc, xmlNodePtr cur) { xmlChar *key; cur = cur->xmlChildrenNode; while (cur != NULL) { if ((!xmlStrcmp(cur->name, (const xmlChar *)"keyword"))) { key = xmlNodeListGetString(doc, cur->xmlChildrenNode, 1); printf("keyword: %s\n", key); xmlFree(key); } cur = cur->next; } return; } Again we get the first child node. Like the loop above, we then iterate through the nodes, looking for one that matches the element we're interested in, in this case "keyword". When we find the "keyword" element, we need to print its contents. Remember that in XML, the text contained within an element is a child node of that element, so we turn to cur->xmlChildrenNode. To retrieve it, we use the function xmlNodeListGetString, which also takes the doc pointer as an argument. In this case, we just print it out. Because xmlNodeListGetString allocates memory for the string it returns, you must use xmlFree to free it. In addition to walking the document tree to find an element, Libxml2 includes support for use of XPath expressions to retrieve sets of nodes that match a specified criteria. Full documentation of the XPath API is here. XPath allows searching through a document for nodes that match specified criteria. In the example below we search through a document for the contents of all keyword elements. A full discussion of XPath is beyond the scope of this document. For details on its use, see the XPath specification. Full code for this example is at . Using XPath requires setting up an xmlXPathContext and then supplying the XPath expression and the context to the xmlXPathEvalExpression function. The function returns an xmlXPathObjectPtr, which includes the set of nodes satisfying the XPath expression. xmlXPathObjectPtr getnodeset (xmlDocPtr doc, xmlChar *xpath){ xmlXPathContextPtr context; xmlXPathObjectPtr result; context = xmlXPathNewContext(doc); result = xmlXPathEvalExpression(xpath, context); if(xmlXPathNodeSetIsEmpty(result->nodesetval)){ xmlXPathFreeObject(result); printf("No result\n"); return NULL; First we declare our variables. Initialize the context variable. Apply the XPath expression. Check the result and free the memory allocated to result if no result is found. The xmlPathObjectPtr returned by the function contains a set of nodes and other information needed to iterate through the set and act on the results. For this example, our functions returns the xmlXPathObjectPtr. We use it to print the contents of keyword nodes in our document. The node set object includes the number of elements in the set (nodeNr) and an array of nodes (nodeTab): for (i=0; i < nodeset->nodeNr; i++) { keyword = xmlNodeListGetString(doc, nodeset->nodeTab[i]->xmlChildrenNode, 1); printf("keyword: %s\n", keyword); xmlFree(keyword); } The value of nodeset->Nr holds the number of elements in the node set. Here we use it to iterate through the array. Here we print the contents of each of the nodes returned. Note that we are printing the child node of the node that is returned, because the contents of the keyword element are a child text node. element writing content Writing element content uses many of the same steps we used above — parsing the document and walking the tree. We parse the document, then traverse the tree to find the place we want to insert our element. For this example, we want to again find the "storyinfo" element and this time insert a keyword. Then we'll write the file to disk. Full code: The main difference in this example is in parseStory: void parseStory (xmlDocPtr doc, xmlNodePtr cur, char *keyword) { xmlNewTextChild (cur, NULL, "keyword", keyword); return; } The xmlNewTextChild function adds a new child element at the current node pointer's location in the tree, specified by cur. file saving Once the node has been added, we would like to write the document to file. Is you want the element to have a namespace, you can add it here as well. In our case, the namespace is NULL. xmlSaveFormatFile (docname, doc, 1); The first parameter is the name of the file to be written. You'll notice it is the same as the file we just read. In this case, we just write over the old file. The second parameter is a pointer to the xmlDoc structure. Setting the third parameter equal to one ensures indenting on output. attribute writing Writing an attribute is similar to writing text to a new element. In this case, we'll add a reference URI to our document. Full code:. A reference is a child of the story element, so finding the place to put our new element and attribute is simple. As soon as we do the error-checking test in our parseDoc, we are in the right spot to add our element. But before we do that, we need to make a declaration using a data type we have not seen yet: xmlAttrPtr newattr; We also need an extra xmlNodePtr: xmlNodePtr newnode; The rest of parseDoc is the same as before until we check to see if our root element is story. If it is, then we know we are at the right spot to add our element: newnode = xmlNewTextChild (cur, NULL, "reference", NULL); newattr = xmlNewProp (newnode, "uri", uri); First we add a new node at the location of the current node pointer, cur. using the xmlNewTextChild function. Once the node is added, the file is written to disk just as in the previous example in which we added an element with text content. attribute retrieving value Retrieving the value of an attribute is similar to the previous example in which we retrieved a node's text contents. In this case we'll extract the value of the URI we added in the previous section. Full code: . The initial steps for this example are similar to the previous ones: parse the doc, find the element you are interested in, then enter a function to carry out the specific task required. In this case, we call getReference: void getReference (xmlDocPtr doc, xmlNodePtr cur) { xmlChar *uri; cur = cur->xmlChildrenNode; while (cur != NULL) { if ((!xmlStrcmp(cur->name, (const xmlChar *)"reference"))) { uri = xmlGetProp(cur, "uri"); printf("uri: %s\n", uri); xmlFree(uri); } cur = cur->next; } return; } The key function is xmlGetProp, which returns an xmlChar containing the attribute's value. In this case, we just print it out. If you are using a DTD that declares a fixed or default value for the attribute, this function will retrieve it. encoding Data encoding compatibility problems are one of the most common difficulties encountered by programmers new to XML in general and libxml in particular. Thinking through the design of your application in light of this issue will help avoid difficulties later. Internally, libxml stores and manipulates data in the UTF-8 format. Data used by your program in other formats, such as the commonly used ISO-8859-1 encoding, must be converted to UTF-8 before passing it to libxml functions. If you want your program's output in an encoding other than UTF-8, you also must convert it. Libxml uses iconv if it is available to convert data. Without iconv, only UTF-8, UTF-16 and ISO-8859-1 can be used as external formats. With iconv, any format can be used provided iconv is able to convert it to and from UTF-8. Currently iconv supports about 150 different character formats with ability to convert from any to any. While the actual number of supported formats varies between implementations, every iconv implementation is almost guaranteed to support every format anyone has ever heard of. A common mistake is to use different formats for the internal data in different parts of one's code. The most common case is an application that assumes ISO-8859-1 to be the internal data format, combined with libxml, which assumes UTF-8 to be the internal data format. The result is an application that treats internal data differently, depending on which code section is executing. The one or the other part of code will then, naturally, misinterpret the data. This example constructs a simple document, then adds content provided at the command line to the document's root element and outputs the results to stdout in the proper encoding. For this example, we use ISO-8859-1 encoding. The encoding of the string input at the command line is converted from ISO-8859-1 to UTF-8. Full code: The conversion, encapsulated in the example code in the convert function, uses libxml's xmlFindCharEncodingHandler function: xmlCharEncodingHandlerPtr handler; size = (int)strlen(in)+1; out_size = size*2-1; out = malloc((size_t)out_size); … handler = xmlFindCharEncodingHandler(encoding); … handler->input(out, &out_size, in, &temp); … xmlSaveFormatFileEnc("-", doc, encoding, 1); handler is declared as a pointer to an xmlCharEncodingHandler function. The xmlCharEncodingHandler function needs to be given the size of the input and output strings, which are calculated here for strings in and out. xmlFindCharEncodingHandler takes as its argument the data's initial encoding and searches libxml's built-in set of conversion handlers, returning a pointer to the function or NULL if none is found. The conversion function identified by handler requires as its arguments pointers to the input and output strings, along with the length of each. The lengths must be determined separately by the application. To output in a specified encoding rather than UTF-8, we use xmlSaveFormatFileEnc, specifying the encoding. compiler flags Libxml includes a script, xml2-config, that can be used to generate flags for compilation and linking of programs written with the library. For pre-processor and compiler flags, use xml2-config --cflags. For library linking flags, use xml2-config --libs. Other options are available using xml2-config --help. &STORY; &KEYWORD; &XPATH; &ADDKEYWORD; &ADDATTRIBUTE; &GETATTRIBUTE; &CONVERT; A number of people have generously offered feedback, code and suggested improvements to this tutorial. In no particular order: Daniel Veillard Marcus Labib Iskander Christopher R. Harris Igor Zlatkovic Niraj Tolia David Turover