PRATA: A System for XML Publishing, Inte- gration and View Maintenance

We present PRATA, a system that supports the following in a uniform framework: (a) XML publishing, i.e., converting data from databases to an XML document, (b) XML integration, i.e., extracting data from multiple, distributed databases, and integrating the data into a single XML document, and (c) incremental maintenance of published or integrated XML data (view), i.e., in response to changes to the source databases, efficiently propagating the source changes to the XML view by computing the corresponding XML changes. A salient feature of the system is that publishing, integration and view maintenance are schema-directed: they are conducted strictly following a user-specified (possibly recursive and complex) XML schema, and guarantee that the generated or modified XML document conforms to the predefined schema. Figure 1. PRATA System Architecture PRATA consists of three main modules: • Schema-Directed XML Publishing • XML Integration • Incremental Maintenance of XML Views To our knowledge, PRATA is the first and the only system that is capable of supporting all of these. Schema-Directed Publishing •This module allows users to specify mappings from a relational database schema R to a predefined XML schema D, via a GUI and in a novel language Attribute Translation Grammar (ATG) that we proposed in [2]. •The ATG approach for publishing relational data in XML is given as follows, by using a simplified example taken from the IUPHAR (International Union of Pharmacology) Receptor Database [4]. ————————————————————— Source relational schema R0: chapters(chapter id, name) receptors(receptor id, chapter id, name, code) refs(ref id, chapter id, year, title) cite(ref id, receptor id) ————————————————————— Target DTD D0: /* #PCDATA is omitted here. */ ————————————————————— ATG σ0: Semantic Attributes: /*omitted*/ Semantic Rules: db→ family* Q1: $family← select chapter id, name from chapters family→ name, receptors, references $fname = ($family.name), $references = ($family.chapter id), $receptors = (0, $family.chapter id, ∅) receptors→ receptor* Q2: $receptor← case $receptors.tag of 0: select receptor id, name, $receptors.ids from receptors where chapter id = $receptors.id 1: select a.receptor id, a.name, $receptors.ids from receptors a, cite b, cite c where b.receptor id = $receptors.id and b.ref id = c.ref id and b.receptor id <> c.receptor id and a.receptor id = c.receptor id and a.receptor id not in $receptors.ids receptor→ name, receptors $rname = ($receptor.name), $receptors = (1, $receptor.receptor id, $receptor.ids ∪ $receptor.receptor id) references→ reference* Q3: $reference← select title, year from refs where chapter id = $references.chapter id reference→ title, year $year = ($reference.year), $title = ($reference.title) A→ S /* A is one of name, title, year */ $S = ($A.val) ————————————————————— •Given an ATG σ0 and an IUPHAR database instance I of R0 as above, the system automatically generates an XML document (view) σ0(I) of I such that σ0(I) is guaranteed to conform to the given DTD D0 as above. •We successfully generate the XML document of the whole IUPHAR database using ATG grammar. •One can write ATG “programs” easily with basic knowledge of SQL and DTD.