Sphinx a user-friendly assertion language David North Senior Developer, CoreFiling Eurofiling Workshop, Luxembourg, December 2013.

Sphinx a user-friendly assertion language

David NorthSenior Developer, CoreFilingEurofiling Workshop, Luxembourg, December 2013

What is Sphinx?

• Domain-specific language for making assertions about XBRL facts

• High-level, semantic, based on an n-dimensional world

• Not XML• Designed for humans to write and

edit directly• Proprietary

History

• 2009 – first prototypes• 2011 – stable language reaches 1.0• 2012 – version 2.0 incorporating real-

world feedback requiring breaking changes

• 2013 – (later abandoned) attempt to offer the Sphinx IP to XBRL International

The requirements Sphinx grew from• XBRL US (XUSI) – development of the

XBRL US Consistency Checks for quality/consistency of SEC filings

• CIMA – regulator needing to check quality/consistency of incoming instances

• Replaced a stopgap spreadsheet-based solution

Who uses it?

• XUSI• CIMA• HMRC

Proprietary: Disadvantages

• Not an XII standard• Fewer options for implementations• Difficult for a regulator to mandate

its usage by filers• Can't be part of DTS

Proprietary: Advantages

• High quality tools available from CoreFiling

• Fast evolution of the language in response to customer feedback

What can it be used for?

• Rapid creation of business and accounting rules

• Typical checks might include:– X must be reported– Co-constraints (if X is reported, Y must

be too)– Comparing data across dimensions

The language

• Based on item expressions• The most basic: [ ]• Represents the table of all items in

the instance. Operations on it apply to every cell in turn

• Restrict by axes (“aspects” in XBRL formula)– FooConcept[ ] – restriction to facts with

FooConcept– [Geography=UK, Product=Cheese] –

restricts dimensions

Lining up

How is the following expression evaluated?

Revenue[ ] = SalesRevenue[ ] + RentRevenue[ ]

Examples

Please refer to accompanying file “Sphinx Examples.txt”

Tools

• Rules authored in SpiderMonkey, potentially alongside taxonomy development– Can test against instances here

• Can plug authored “rulebase” into other CoreFiling tools both for GUI-based instance validation and command-line/web based validation (e.g. filter gateway).

Sphinx vs XBRL formula

• Sphinx does lining up (implicit filtering) by default

• Sphinx cannot do fact creation• Sphinx cannot use XPath• User-defined Sphinx functions are in-

language only (no ability to make them implementation-defined)

Questions?