Information Technology Reference
In-Depth Information
All functions from the domain model can now be explicitly defined in terms
of the element values. In this way the axioms (that refer to these functions) from
the domain model now impose well-formedness conditions on the elements of a
language description. Additional axioms that impose well-formedness conditions
on the
rct
,
ppt
and
sst
values are added, so that these axioms provide a
static
semantics
for the DSL.
Concrete Syntax and Data Collector Implementation.
Two alternative
solutions to the implementation of the concrete DSL have been made using the
Extensible Markup Language
XML
[39] and the Unified Modelling Language
UML
[34], respectively. Below we outline the XML solution that is documented
in [14]. In [3] it is described how the DSL is defined by a UML 2.0 profile in the
second solution.
The
concrete syntax
of the language has been defined by an XML document
type definition (DTD). For the elements of the RSL abstract syntax, correspond-
ing XML elements are defined. The
static semantics
has been implemented using
the extensible style sheet language XSL [40]. In a systematic way each RSL axiom
expressing a well-formedness requirement has been transformed into a template
that tests whether the requirement is fulfilled.
AGUIbased
data collector
for creating DSL descriptions in the required XML
syntax has been developed using XForms [38].
For the convenience of users, a
graphical representation
of DSL descriptions
(XML documents) has been developed. This was done using XSLT and HTML.
In Fig. 3 the graphical representation of some sample interlocking tables for the
network given in Fig. 2 are shown.
5
Generating Applications from Domain-Specific
Descriptions
According to our method three generators taking a statically well-formed DSL
description
as argument are required (Fig. 1). The primary one is the gen-
erator producing a control system model
D
M
. The second generator produces
the behavioural model
of the physical environment and the third one gener-
ates the safety properties
Φ
, to be checked to hold for the concurrent composition
of the control system model
P
M
and the physical model
P
. In this section we
outline the basic concepts of the generator for
M
. The other generators are
developed in a similar way.
Components of the Controller Model Generator.
The implemented gen-
erator for controller models consists of two parts:
1. A configurable library of generic code that is re-usable for all control sys-
tems to be generated: The code comprises generic versions of the control
algorithms, the data structures carrying dynamic state information needed
for performing control decisions, and the static configuration data structures.
Search WWH ::
Custom Search