How Does doc() Differ from document() in XSLT 2.0?

In XSLT development, accessing external XML data transitioned significantly between XSLT 1.0 and XSLT 2.0. While XSLT 1.0 relied exclusively on the versatile but complex document() function, XSLT 2.0 introduced doc() alongside document() to provide a simpler, standardized mechanism based on XPath 2.0 and the XML Query data model. This article breaks down the technical differences between doc() and document(), examining their URI resolution rules, handling of multiple arguments, node sequencing, and error behavior.

Core Purpose and Origins

The document() function was designed specifically for XSLT 1.0 to load external resources, process secondary XML files, and read lookup tables. Because XPath 1.0 lacked a native mechanism for opening external documents, document() was an XSLT-specific instruction.

With XSLT 2.0 and XPath 2.0, the core specification integrated the doc() function into the XPath 2.0 Functions and Operators standard. The goal was to provide a streamlined, predictable function for retrieving a single document, aligning XPath with XQuery.

Key Differences

1. Argument Types and Document Counts

doc($uri): Accepts a single string argument (or an empty sequence) representing a single URI. It returns exactly one document node (document-node()), or an empty sequence if the argument is an empty sequence.
document($uri-sequence, [$base-node]): Accepts a node-set, sequence, or string as its first argument. If passed a node-set or sequence of URIs, it fetches all referenced XML documents and returns their document nodes merged into a single node collection.

2. Base URI Resolution

doc($uri): Resolves relative URIs against the static base URI of the stylesheet expression where the call appears. It does not accept a second argument for dynamic base URI overriding.
document($uri, $base-node): Allows an optional second argument (a node). If supplied, relative URIs in the first argument resolve against the base URI of that specific node rather than the stylesheet, allowing dynamic URI resolution based on source document structure.

3. Whitespace and Node Stripping

doc(): Does not apply the stylesheet’s xsl:strip-space or xsl:preserve-space declarations to the loaded document in standard XPath implementations unless explicitly handled by the XSLT processor's document builder settings.
document(): Integrates directly into the XSLT processing model, routinely applying whitespace stripping rules defined by xsl:strip-space in the importing stylesheet.

4. Error Handling and Missing Resources

doc(): Triggers a dynamic error (FODC0002) if the resource cannot be retrieved or contains malformed XML. To handle missing files safely, XSLT 2.0 pairs it with doc-available($uri).
document(): Historically permitted processors to return an empty node-set rather than raising a fatal error if a document was not found, depending on the XSLT 1.0 implementation.

Comparison Summary

Feature	`doc()` (XPath 2.0 / XSLT 2.0)	`document()` (XSLT 1.0 / 2.0)
Standard Origin	XPath 2.0 / XQuery 1.0	XSLT 1.0 / XSLT 2.0
Input Arguments	Single URI string (or empty)	Sequence/Node-set of URIs, optional base node
Return Value	Single document node or empty	Node-set of multiple document nodes
Base URI Default	Static base URI of stylesheet	Static base URI or dynamic node base
Safety Check	Use with `doc-available()`	Implementation-dependent empty fallback

When to Use Which

In modern XSLT 2.0 and 3.0 stylesheets, doc() is preferred for loading individual external XML resources due to its strict typing and cross-compatibility with XQuery. The document() function remains useful when you need to load a batch of documents simultaneously from a list of URI nodes or require dynamic base URI resolution using a secondary context node.