Edit report at http://bugs.php.net/bug.php?id=53655&edit=1
ID: 53655 Updated by: rricha...@php.net Reported by: olav dot morken at uninett dot no Summary: Improve speed of DOMNode::C14N() on large XML documents -Status: Open +Status: Assigned Type: Feature/Change Request Package: DOM XML related PHP Version: 5.3.4 -Assigned To: +Assigned To: rrichards Block user comment: N Private report: N Previous Comments: ------------------------------------------------------------------------ [2011-01-05 08:33:02] olav dot morken at uninett dot no Description: ------------ The C14N() function appears to have a runtime that is O(N^2) (or possibly worse?) depending on input size, which means that it becomes very slow as the input grows. For example, an input with around 196000 nodes takes about 290 seconds, while an input with 486000 nodes takes 2200 seconds. Note that this problem only occurs when canonicalizing a subtree of the docuemnt. If we canonicalize the whole document, it completes almost immediately. The problem is that canonicalization uses an XPath expression to find the nodeset that should be canonicalized. Evaluation of the XPath expression takes a lot of time as the input size grows, but the libxml2 xmlC14NDocSaveTo() function also has to do a lookup in the nodeset returned by the XPath expression for every node it encounters. I believe a better solution would be to do this like it is done in the xmlsec library. This library use the xmlC14NExecute()-function instead, which accepts a callback that determines whether a node should be included in the result. This should make the speed of canonicalization linear with the input size. Test script: --------------- <?php $doc = new DOMDocument(); $doc->load('some-large-xml-file.xml'); $start = microtime(TRUE); $doc->documentElement->C14N(FALSE, FALSE); echo "Done in " . (microtime(TRUE) - $start) . " seconds.\n"; ------------------------------------------------------------------------ -- Edit this bug report at http://bugs.php.net/bug.php?id=53655&edit=1