330 likes | 340 Views
Explore one approach for mapping TeleAtlas road data to EuroRoadS using XMI, GML, and model transformation techniques. Identify model subsets, transformation challenges, and tools like Rational Rose and Enterprise Architect. Understand issues with XMI imports and workarounds. Enhance your model for GML support.
E N D
From Model to Data Transformation – One approach Mapping TeleAtlas Data to EuroRoadS Anders Friis-Christensen, JRC
From Model to Data Transformation Outline: • Model • EuroRoadS subset • XMI/GML • Visio -> XMI -> Rational / EA -> • XMI -> GML schema • TeleAtlas Road data (shape) -> GML (+schema) • Model transformation • EuroRoads GML <-> TeleAtlas GML • XSL • Data Transformation • XSLT • Population of data • Verification of data
Creating EuroRoadS subset • A subset of the network exchange model • The network model is a common base for roads • Including: • Attributes (modeled as abstract classes) • Geometry (GML data types) • No topology • No routing info
XMI Generation from Visio Newline in empty tag <Foundation.Core.Classifier.feature></Foundation.Core.Classifier.feature> not allowed. Changed 198 places. Newline in empty tag <Foundation.Core.AssociationEnd.qualifier></Foundation.Core.AssociationEnd.qualifier> not allowed. Changed 192 places. Newline in empty tag <Foundation.Core.Namespace.ownedElement></Foundation.Core.Namespace.ownedElement> not allowed. Changed 4 places. Undefined references: line 5901: <Foundation.Core.Classifier.associationEnd> <Foundation.Core.AssociationEnd xmi.idref="UID92E0A85B-FFCA-4D3E-AFA3-8A94E5121744"/> <- <Foundation.Core.AssociationEnd xmi.idref="UIDFA3023F9-5631-4AF5-85C0-761190195B8C"/> <- </Foundation.Core.Classifier.associationEnd> deleted line 6054: <Foundation.Core.Classifier.associationEnd> <Foundation.Core.AssociationEnd xmi.idref="UID958794C3-5D75-4A43-89E3-F0B6E1546E5B"/> <Foundation.Core.AssociationEnd xmi.idref="UID59BE8B25-EFB3-4BEF-8811-B438964F63A3"/> <- <Foundation.Core.AssociationEnd xmi.idref="UID657C99EC-7244-47BD-B84B-BBEEF3A69330"/> <- </Foundation.Core.Classifier.associationEnd> deleted line 12659: <Foundation.Core.GeneralizableElement.specialization> <Foundation.Core.Generalization xmi.idref="UID496EBF06-CABE-46F1-87C0-F66CDEAD17D8"/> <- <Foundation.Core.Generalization xmi.idref="UID7EA823B3-EB2F-4160-B6C8-F8518860E7F0"/> </Foundation.Core.GeneralizableElement.specialization> deleted • Use of Microsoft XMIExport.dll • Invoked by ESRI add-on or by Lars Vikstroms application • Produces non-valid XMI supposedly based on UML 1.1 • Several errors: • Newline in empty tags, more than 400 errors (only in XML spy 2004, error in parser) • Undefined references (associations ends and generalization) • Must be an error in the model, old references not deleted properly by Visio
XMI Import (Based on Visio export) • Rational Rose • Does not read UML 1.1 XMI • Possible workarounds?? • Enterprise Architect • Based on OMG UML 1.1 XMI 1.0 dtd, success on import • However, several errors. E.g.: • Generalizations not imported • Enumerations not imported • Association multiplicity not imported
XMI Import (Based on Visio export) • ArgoUML 0.18: • Possible to import XMI • Errors: • Multiplicities not imported • Export UML 1.3 XMI 1.0 (without multiplicities) • Rational Rose: • Import new XMI possible • No import of enumerations • Enterprise Architect: • Same as RR As a curiosity: ArgoUML cannot read its own import if it is saved as a project. The XMI parsing fails.
XMI – Issues • Conclusion: • Microsoft Visio is not interoperable (choose another software for UML, if it is foreseen that exchange of models is required) • Various vendors of XML editors/importers have errors in their parsers which complicates the process of validating XMI • Tools do not support same versions of XMI • Tools interpret XMI in different ways • It is very difficult to validate the content of the XMI, i.e. if the model imported is correct even though the XMI is validated. • Report John Herring (communicated by Morten) • EA problem with RR XMI: • It does not import data types <<DataType>> (see also the example EA diagram that was send to you • It does not see realization ("inheritance" from interfaces and types) in the same way as Rose. It transforms those into a type of dependency
Prepare Sub-set Model in RR to support GML • Creation of GML schemas is done by ShapeChange tool (Interactive Instruments): • Converts XMI 1.0 to GML 2.1.2 and 3.1 (3.1.1?) • Other XMI than export from RR?? • Experiments from EA (Clemens) • ShapeChange put different requirements on the UML model (and hence the resulting XMI)
Multiple inheritance: ER_RoadnetElement inherited directly from both ER_RoadFeature and ER_IdentifiableObject Prepare Sub-set Model in RR to support GML
Two associations between ER_RoadAttribute and ER_RoadnetElement Prepare Sub-set Model in RR to support GML
Added multiplicities to associations and made them bidirectional Prepare Sub-set Model in RR to support GML
Added a concrete class with a value of string datatype in order to store data Prepare Sub-set Model in RR to support GML
Prepare Sub-set Model in RR to support GML Added a validityFrom and validityTo Instead of TM_Period in order of simplicity and to support GML 2.1.2
Other UGAS requirements • Additionally following rules needed to be applied the model • Model must be represented in a package stereotyped with <<Application Schema>> • The XML namespace (a URI) and an abbreviation must be present as tagged values
Resulting GML <?xml version="1.0" encoding="UTF-8"?> <!--Output of ShapeChange has been modified JRC-AFC--> <schema xmlns="http://www.w3.org/2001/XMLSchema" xmlns:cslWorkshop="http://www.jrc.it/inspire/cslWorkshop" xmlns:gml="http://www.opengis.net/gml" xmlns:sc="http://www.interactive-instruments.de/ShapeChange" xmlns:scXML="http://www.isotc211.org/scXML" xmlns:smXML="http://www.isotc211.org/smXML" xmlns:xlink="http://www.w3.org/1999/xlink" targetNamespace="http://www.jrc.it/inspire/cslWorkshop" elementFormDefault="qualified" version="0.1"> <import namespace="http://www.opengis.net/gml" schemaLocation="feature.xsd"/> <!--Added FeatureCollection in order to validate GML containing multiple feature members. JRC-AFC--> <element name="FeatureCollection" type="cslWorkshop:FeatureCollectionType" substitutionGroup="gml:_FeatureCollection"/> <complexType name="FeatureCollectionType"> <complexContent> <extension base="gml:AbstractFeatureCollectionType"/> </complexContent> </complexType> <!--Type and element definitions of this package--> <complexType name="ER_RoadnetLinkType" abstract="true"> <annotation> <documentation>Note: The attribute TM_Period in the original EuroRoadS model has been replaced by validityFrom and validityTo for simplicity reasons</documentation> </annotation> <complexContent> <extension base="cslWorkshop:ER_RoadnetElementType"> <sequence> <element name="validityFrom" type="date" minOccurs="0"/> <element name="validityTo" type="date" minOccurs="0"/> <element name="curve" type="gml:LineStringPropertyType" minOccurs="0"/> </sequence> </extension> </complexContent> </complexType> <complexType name="ER_RoadnetLinkPropertyType"> <sequence> <element ref="cslWorkshop:ER_RoadnetLink" minOccurs="0"/> </sequence> <attributeGroup ref="gml:AssociationAttributeGroup"/> </complexType> <complexType name="ER_RoadnetLinkPropertyByValueType"> <sequence> <element ref="cslWorkshop:ER_RoadnetLink"/> </sequence> </complexType> <element name="ER_RoadnetLink" type="cslWorkshop:ER_RoadnetLinkType" abstract="true" substitutionGroup="cslWorkshop:ER_RoadnetElement"/> <complexType name="ER_IdentifiableObjectType" abstract="true"> <complexContent> <!--Added FeatureCollection in order to validate GML containing multiple feature members. JRC-AFC--> <element name="FeatureCollection" type="cslWorkshop:FeatureCollectionType" substitutionGroup="gml:_FeatureCollection"/> <complexType name="FeatureCollectionType"> <complexContent> <extension base="gml:AbstractFeatureCollectionType"/> </complexContent> </complexType>
Resulting GML • Because of the many classes in the model, the GML is complex • I did not work further with GML 3.1 schemas • Complex model • 3.1 schemas not valid (according several parsers), experience in producing 3.1.1? • Tools support (3.1 and 3.1.1)? • Any experiences??
Data transformation • Produce GML from TeleAtlas data • FME GML • Create schema option • No support for 3.1 or 3.1.1?? • TeleAtlas based on GDF • Should be no problems when mapping attributes • Except that the attributes are named: ID, FEATTYP, FT, F_JNCTID, F_JNCTTYP, T_JNCTID, T_JNCTTYP, PJ, METERS, FRC, NETCLASS, NETBCLASS, NET2CLASS, NAME, NAMELC, SOL, NAMETYP, CHARGE, ROUTENUM, RTETYP, RTEDIR, RTEDIRVD, PROCSTAT, FOW, SLIPRD, FREEWAY, BACKRD, TOLLRD, RDCOND, STUBBLE, PRIVATERD, CONSTATUS, ONEWAY, F_BP, T_BP, F_ELEV, T_ELEV, KPH, MINUTES, POSACCUR, CARRIAGE, LANES
FID (TA) mapped to ID:permanentID (should actually have been ID (TA) as FID is FME generated) We only map data to ER_RoadLink Name (TA) mapped to StringAttribute KPH (TA) mapped to speedLimit lineStringProperty (TA) mapped to GM_Curve FOW (TA) mapped to FormOfWay using lookup table Always ER_RoadElement FRC (TA) mapped to NationalRoadClass using lookup table XSL FeatureCollection properties (e.g. bbox) are mapped
Resulting GML • Validates to EuroRoadS submodel GML schema • Complex attribute types not possible to import via FME or ArcGIS • Requires flat structure • After short! trial, no success in publishing via GeoServer WFS
Approach(1) • Automatically (preferably no manual intervention) • Easy set-up • - Minimum cost • - Support from vendors/technologies • - Support a SOA • Pragmatic solution • Feasibility of approach • Does it satisfy requirements: • It is automatic. • Creating XSL files can be complicated, especially if the semantics of attributes to be transformed vary too much • Time-consuming (costly) • XSLT is a standardized methodology, GML is supported by the important software producers • We shall later see similar approach in a service architecture (Snowflake, Finnish Geodetic Ins. etc..) • Would best support: • s
Approach • Requires that the two models that need to be mapped do not differentiate too much: • Attribute types, feature types, etc. • No (immediate) generalization mechanisms • i.e., the approach is scale dependent • Aggregation (attributes, features) is complicated • Requires in-depth knowledge of both models • From domain to low level spatio-temporal aspects of models • How to test that a mapping occurs error free • Validation of data? • Constraints? • Other possibilities: • Materialized views and stored procedures for mapping • Create DB schema based on ‘complex’ UML or GML schemas?