XProc next CG

XProc 3.0 — An overview

As of 2018-02-10

Short reminder on XProc

XProc is an XML pipeline language
Create XML centric workflows like
- Expand document with XInclude
- Validate document with RelaxNG, Schematron or ..
- Transform to HTML or PDF with XSLT or XSL-FO
W3C recommendation in 2010 - Working group closed in 2016
W3C community group took over in September 2016: Meetings in Amsterdam, Prague, London and Aachen (2017) and Prague (2018)
About 15 active members
Four editors of the specs

Overall Goal: Improve usability of XProc

The only type of document processable in XProc (1.0) were well-formed XML documents.

The new model: A document is a tuple of a representation and its properties

Properties are metadata of the document, e.g. content-type, base-uri, user defined properties
The representation is the data structure the processor uses to refer to the content.

XProc 3.0 currently knows four document types

Text value templates (as known from XSLT):
```
<p:inline><doc>{$a}</doc><p:inline>
```
will put the content of $a as children into <doc/>

Attribute value templates: Instead of

<p:add-attribute attribute-name="foo">
  <p:with-option name="attribute-value" 
   select="concat($a,$b)" />
</p:with-option

You can now simply write:

<p:add-attribute attribute-name="foo"
  attribute-value="{concat($a,$b)}" />

Variables and options in XProc 1.0 were untyped atomics or strings
XProc 3.0 uses sequences as defined in XDM 3.1 as basis, including maps and arrays
Sequence type syntax can be used to declare and check expected types

Variables now can hold nodes or even documents

<p:variable name="theDoc" select="doc('my-doc')"/>
<p:identity>
   <p:with-input>
     <wrapper>{$theDoc}</wrapper>
   </p:with-input>
<p:identity>

In XProc 1.0 you could only use built-in functions from XPath and XProc
Often not handy when you had to write the same XPath expression over and over again

New in XProc 3.0:

<p:import-functions href="my-lib" 
  content-type="application/xslt+xml" 
  namespace="http://example.com/foo"/>

will import all functions (and variables?) within the given namespace from the URI

Variable declarations everywhere. Variables shadowed by name
No more parameters ports. Maps are used instead to provide values to XSLT, XQuery etc.
Maps are also used as bags for serialization parameters
New compound step: <p:if /> No need to declare the same output ports on every branch of try/catch or <p:choose/> any more
Enhanced try/catch: Multiple catch clauses instead of just one; p:finally
Renamed port binding to p:with-input to mark the difference between declaring and binding ports
Ever wondered what <p:namespaces/> is for? Don't worry, it is gone!

XProc 1.0:

<p:input port="source">
 <p:inline>
  <doc />
 </p:inline>
</p:input>

XProc 3.0

<p:with-input><doc /><p:with-input>

XProc 1.0:

<p:input port="source">
  <p:document href="some-thing" />
</p:input>

XProc 3.0

<p:with-input href="some-thing" />

XProc 1.0:

<p:input port="source">
  <p:pipe step="some" port="thing" />
</p:input>

XProc 3.0

<p:with-input pipe="thing@some" />

<p:with-input pipe="@some" />

<p:with-input pipe="thing" />

<p:with-input pipe="port1@step1 port2@step2 port3@step3 />

Specs split into three documents: core, (required) step library and (optional) step library
Last call draft (of the core spec) “this spring”
Core spec and step spec “candidate releases” ready by June
XML Calabash 2 and MorganaXProc 2 implementations commensurate with CR by June
Erik Siegel’s Programmer Reference for 3.0 also planned for June