Virtuoso implements XSLT 1.0 transformations as SQL callable functions. In general, Virtuoso's XSLT processor follows the XSLT 1.0 standard as far as possible without loss of database-related functionality. The most important difference is in error handling. While XSLT rules assume that almost any error must be recovered, Virtuoso will stop the transformation in case of data access error or in case of serious arithmetic error, such as divide by zero. This is due to the greater complexity of the Virtuoso XSLT processor's environment. Standalone processors generally accept a set of files as input parameters, the only output is the resulting transformation as a file, and all configuration parameters are listed on the command line. Virtuoso's XSLT processor may retrieve both data and stylesheets from local filesystems, local tables, remote datasources, or the Web. Parameters of the stylesheet transformation may be prepared by complicated Virtuoso PL procedures and so on. Plain XSLT processors may continue producing "dirty" output; Virtuoso's processor will stop upon encountering serious in order to produce helpful error diagnostics.
Another important difference is that Virtuoso has one processor for XSLT, XPath and XQuery. The user will find this convenient because an application may use XSLT-specific functions in XPath expressions, XQuery functions in XSLT, and so on. If extension functions are defined for the XSLT processor, they may also be used in XPath and XQuery expressions without any additional operations. One caveat is that this may invalidate some rules related to the document order of nodes in node-sets. Both the XPath 1.0 and XSLT 1.0 standards say that some functions must process the node that is first in document order from a given node-set. In such cases Virtuoso's processor will process the first node found in the node-set, regardless of whether it is actually the first in document order or not. This is done because the old XPath rule is nonsensical if nodes of the node-set are retrieved from more than one document or are created in the query. The processor follows the old rule for 99% of real expressions. Specifically, it fully matches the standards' rules for all XPath expressions that are free of reverse-order axes and non-standard functions.