Method for converting relational data into XML6785673Abstract A method for converting relational data to XML (eXtensible Markup Language) is provided. The method can use a greedy algorithm to efficiently construct materialized XML views of relational databases. A greedy algorithm designed for XML view definition queries is provided for decomposing a large query into smaller queries and determining which query will run faster without actually running the query. Claims What is claimed: Description FIELD OF THE INVENTION
<?xml encoding = "US-ASCII"?>
<!ELEMENT supplier (company, product*)>
<!ELEMENT product (name, category, description, retail,
sale?, report*)>
<!ATTLIST product ID ID>
<!ELEMENT company (#PCDATA)>
<!ELEMENT name (#PCDATA)>
<!ELEMENT retail (#PCDATA)>
<!ELEMENT sale (#PCDATA)>
<!ELEMENT report (#PCDATA)>
<!ATTLIST report code (size.vertline.defective.vertline.style)
#REQUIRED>
The above code includes the supplier's name and a list of available products. Each product element includes an item name, a category name, a brief description, a retail price, an optional sale price, and zero or more trouble reports. The content of a retail or sale element typically is a currency value. A trouble report includes a code attribute, indicating the class of problem; the report's content may be the customer's comments. Most importantly, this DTD can be used by suppliers and resellers, and it can be a public document. Consider now a particular supplier whose business data is organized according to the relational schema. An illustrative schema of a supplier's relational database (* denotes key) is depicted in the code below.
Clothing(*pid, item, category, description, price, cost)
SalePrice(*pid, price)
Problems(pid, code, comments)
The Clothing table contains tuples with a product id (the table's key), an item name, category name, item description, price, and cost. The SalePrice table contains sale prices and has key field pid and the Problem table contains trouble codes of products and their reports. The above code shows a third-normal form relational schema, designed for the supplier's particular business needs. The schema can be proprietary. For example, the supplier may not want to reveal the attribute cost in Clothing. The supplier's task is to convert its relational data into a valid XML view conforming to the DTD and make the XML view available to resellers. In this example, it is assumed that the supplier exports a subset of its inventory, in particular, its stock of winter outerwear that it wants to sell at a reduced price at the end of the winter season. Once the XML views of a suppliers' data are available, the reseller can access that data by formulating queries over the XML view. Some examples of such queries may include: 1) retrieve products whose sale price is less than 50% of the retail price; 2) count the number of "defective" reports for a product; and 3) compute minimum and maximum cost of outerwear stock. As these queries might suggest, the reseller is typically interested only in a small subset of the information provided by the suppliers. Those skilled in the art will recognize that these queries could be formulated as SQL queries over the supplier's relational database, but relational schemas can differ from supplier to supplier and may not be accessible by the reseller. I. Architecture of SilkRoute FIG. 1 shows an illustrative architecture of SilkRoute according to the present invention. SilkRoute 100 serves as middleware between a relational database server (RDBMS) 110 and an application 120 accessing data over a distributed network, such as the Web/Intranet 130. The distributed network may be a public or private network. According to the invention, the database administrator starts by writing a view query that defines the XML virtual view of the database. In a preferred implementation of the present invention, the view query is an RXL query. The view query is typically complex, because it transforms the relational data into a deeply nested XML view. The resulting view query is virtual, meaning that it is not evaluated, but kept in source code. Typically, applications contact SilkRoute 100 to request data. An application 120 only "sees" the virtual XML view, not the underlying relational database. To access the data, the application 120 can formulate a user query in XML-QL over the virtual view and send it to SilkRoute 100. Together, the view query (e.g., RXL view query) and the user query (e.g., XML-QL user query) can be passed to the query composer module 102 in SilkRoute 100. The query composer module 102 computes the composition and produces a new view query (e.g., RXL query), called the executable query. The answer to the executable query typically includes only a small fragment of the database, e.g., one data item, a small set of data items, or an aggregate value. The result of SilkRoute 100 is an XML document, as specified by the user query (e.g., XML-QL user query). Once computed, the executable query is passed to the translator 104, which partitions the executable query into a data-extraction part, e.g., one or more SQL queries, and an XML-construction part, e.g., an XML template. The translator 104 also may take as an input a description of the relational schema and uses the relational schema to perform syntax checking of the RXL query (e.g., to ensure that the relations named in the RXL query exist in the relational database) and to determine the capabilities of the SQL dialect used by the relational database (e.g., does the SQL dialect support inner joins?). Until now, SilkRoute 100 has manipulated only query source code, but no data. At this point, the data extraction part (e.g., SQL queries) is sent to the RDBMS server 110, which returns one tuple stream per each query (e.g., SQL query) in the data extraction part. The XML generator module 106 merges the tuple streams with the XML-construction part and produces the XML document, which is then returned to the application 120. This scenario is probably the most common use of SilkRoute. However those skilled in the art will recognize that minor changes to the information flow in FIG. 1 can permit other scenarios. For example, the data administrator may export the entire database as one large XML document by materializing the view query. This can be done by passing the view query directly to the translator. In another scenario, the result of query composition could be kept virtual for later composition with other user queries. This is useful, for example, when one wants to define a new XML view from an existing composed view. A. The View Query: RXL Next, RXL (Relational to XML transformation Language) is described. RXL essentially combines the extraction part of SQL, i.e., a from and a where clause (possibly followed by sort by and/or group by clauses) with the construction part of XML-QL, i.e., the construct clause. As a first example, consider this RXL query, which defines a fragment of an XML view:
from Clothing $C
where $c.category = "outerwear"
construct <product>
<name>$c.item</name>
<category>$c.category</category>
<retail>$c.price</retail>
</product>
Given a database like that in the supplier's schema, the query can produce an XML fragment like the following:
<product> <name>... </name> <category>...
</category> <retail>... </retail> </product>
<product> <name>... </name> <category>...
</category> <retail>... </retail> </product>
A root element is missing; later it will be explained how to add one. As in SQL, the from clause declares variables that iterate over tables. Variable names start with a $. In this example, $c is a tuple variable that iterates over the Clothing table. The where clause contains zero or more filters (Boolean predicates) over column expressions. The column expression $c.item refers to the item attribute value of $c and in this case, requires that it equal the string "outerwear". The construct clause specifies the XML value, called an XML template, in terms of the bound column expressions. RXL has three powerful features that make it possible to create arbitrarily complex XML structures: nested queries, Skolem functions, and block structure. An example of a nested query is:
construct <view>{
from Clothing $c
construct <product>
<name>$c.item</name>
{ from Problems $p
where $p.pid = $c.cid
construct <report>$p.comments</report>
}
</product>
} </view>
The outer query has no from or where clauses, only a <construct> clause for the root element <view>. The first sub-query builds one <product> element for each row in Clothing. Its inner sub-query creates zero or more <report> sub-elements, one for each report associated with that product. Those skilled in the art and familiar with SQL will recognize this as a left-outer join of Clothing with Problems followed by a group by on Clothing. Skolem functions allow the way elements are grouped to be controlled. Recall that in XML an attribute with type ID contains a value that uniquely identifies the element in the document, i.e., a key. In RXL, the distinguished attribute ID always has type ID, and its value is a Skolem term, which is used to control grouping and element creation. For example, in the following:
from Clothing $c
construct <category ID=Cat($c.category) name=$c.category>
<product>$c. item</product>
</category>
Cat is a Skolem function and Cat ($c.category) is a Skolem term whose meaning is that only one <category> element exists for every value of $c.category, and it includes all products in that category:
<category> <product>p1</product>
<product>p2<product> </category>
<category> <product>p3</product>
<product>p4</product> </category>
Without the ID attribute and its Skolem term, the query would create one <category> element for each row in Clothing:
<category> <product>p1</product>
</category>
<category> <product>p2</product>
</category>
When Skolem terms are missing, RXL introduces them automatically. Since Skolem terms could be used to define arbitrary graphs, RXL enforces semantic constraints that guarantee a view always defines a tree, and therefore, a well-formed XML document. For example, the Skolem term of a sub-element includes all the variables of its the parent element. Finally, the block structure allows RXL to construct parts of complex elements independently. The query below shows an illustrative multi-block RXL view query containing two blocks.
construct
<view ID=View( )>
{ from Clothing $c
construct <product ID=Prod($c.item)>
<name ID=Name($c.item)>$c.item</name>
<price ID=Price($c.item,
$c.price)>$c.price</price>
</product>}
{ from Clearance $d
where $d.disc > 50
construct <product ID=Prod($d.prodname)>
<name ID=Name(
$d.prodname)>$d.prodname</name>
<discount
ID=Discount($d.prodname,$d.disc)>$d.disc</discount>
</product>
</view>
The first block creates elements of the form: <product><name>n</name><price>p</price></ product> for each product name in Clothing. The second block creates elements of the form: <product><name>n</name><discount>d</discount> </product> for each product name in Clearance. It is to be assumed that Clearance(*prodname, disc) is part of the supplier's schema. When the same product name occurs both in Clothing and Clearance, then the two elements will have the same ID key and can be merged into: <product><name>n</name><price>p</price>< discount>d</discount></product> Those skilled in the art and familiar with SQL will recognize this as a full outer join. The below code contains the complete view query, RXL view query (V), for the supplier relational schema example described above.
1. construct
2. <supplier 1D=Supp( )>
3. <company ID=Comp ( )>"Acme Clothing"</company>
4. {
5. fromClothing$c
6. where $c.category = "outerwear"
7. construct
8. <product ID=Prod($c.pid)>
9. <name ID=Name($c.pid,$c.item)>$c.item</name>
10. <category
ID=Cat($c.pid,$c.category)>$c.category</category>
11.
<descriptionID=Desc($c.pid,$c.description)>$c.description</
description>
12. <retail
ID=Retail($c.pid,$c.price)>$c.price</retail>
13. { from SalePrice $s
14. where $s.pid = $c.pid
15. construct
16. <sale
ID#Sale($c.pid,$s.pid,$s.price)>$s.price</retail>
17. }
18. { from Problems $p
19. where $p.pid = $c.pid
20. construct
21. <report code=$p.code
ID=Prob($c.pid,$p.pid,$p.code,$p.comments)>
22. $p.comments
23. </report>
24. }
25. </product>
26. }
27. </supplier>
Lines 1, 2, and 27 create the root <supplier> element. Notice that the Skolem term Supp( ) has no variables, meaning that one <supplier> element is created. The outer-most clause constructs the top-level element supplier and its company child element. The first nested clause (lines 4-26) contains the query fragment described above, which constructs one product element for each "outerwear" item. Within this clause, the nested clause (lines 13-17) expresses a join between the Clothing and Sale Price tables and constructs a sale element with the product's sale price nested within the outer product element. The last nested clause (lines 18-24) expresses a join between the Clothing and Problem tables and constructs one report element containing the problem code and customer's comments; the report elements are also nested within the outer product element. Notice that the Skolem term of product guarantees that all product elements with the same identifier are grouped together. Usually Skolem terms can be inferred automatically, but they have been included explicitly, because they are relevant to query composition described herein. B. The User Query: XML-QL Applications do not access the relational data directly, but through the XML view. To do so, applications provide user queries in XML-QL, a query language specifically designed for XML. XML-QL queries contain a where clause followed by a construct clause. The where clause contains an arbitrary number of XML patterns and filters. The construct clause is identical to that in RXL. In the example described herein, the reseller can retrieve all products with a sale price less than half of the retail price using the XML-QL user query (U) below:
1. construct
2. <results>{
3. where <supplier>
4. <company>$company</company>
5. <product>
6. <name>$name</name>
7. <retail>$retail</retail>
8. <sale>$sale</sale>
9. </product>
10. </supplier>in "http://acme.com:/products.xml",
11. $sale <0.5 * $retail
12. construct
13. <result ID=Result($company)>
14. <supplier>$company</supplier>
15. <name>$name</name>
16. </result>
17. </results>
The where clause includes a pattern (lines 3-10) and a filter (line 11). A pattern's syntax is similar to that of XML data, but also may contain variables, whose names start with $. Filters are similar to RXL (and SQL). The meaning of a query is as follows. First, all variables in the where clause are bound in all possible ways to the contents of elements in the XML document. For each such binding, the construct clause constructs an XML value. Grouping is expressed by Skolem terms in the construct clause. In this example, the construct clause produces one result element for each value of $company. Each result element contains the supplier's name and a list of name elements containing the product names. In this example, the answer to the user query includes a small fraction of the relational database, i.e., only those products that are heavily discounted. C. The Query Composer The query composer module 102 of SilkRoute 100 takes a user query and the RXL view query and generates a new RXL query, which is equivalent to the user query evaluated on the materialized view. In the example described herein, the view query is the RXL view query (V) above, the user query is the XML-QL user query (U), and the composed query, RXL query (C) is shown below.
construct
<results>
{ from Clothing $c, SalePrice $s
where $c.category = "outerwear",
$c.pid = $s.pid,
$s.price < 0.5 * $c.retail
construct
<result ID=Result("Acme Clothing")>
<supplier>"Acme Clothing"</supplier>
<name ID=Name($c.pid, $c.item)>$c.item</name>
</result>
}
</results>
The composed query combines fragments of the view query and user query. Those fragments from the user query are highlighted. The composed query extracts data from the relational database in the same way as the view query. It also includes the user filter $s.price<0.5 $c.retail and structures the result as in the user query. The details of the composition are subtle, and a complete description of the composition algorithm is described later herein. The composed query is referred to as executable, because it is typically translated into SQL queries and sent to the relational database engine. The answer of the executable query is quite small--the same as that of the user query. In general, it is more efficient to execute the composed query, instead of materializing the view query, because composed queries often contain constraints on scalar values that can be evaluated using indexes in the relational database. Such indices are of little or no use when evaluating a view query. For example, consider a user query that specifies the condition: $s.price between 80 and 100. This condition is propagated into the executable query, and then into the SQL query, and can be evaluated efficiently if an index exists on price. In contrast, an index on price is useless when materializing the view query directly. D. Translator and XML Generator The translator 104 takes an RXL query and decomposes it into one or more SQL queries and an XML template. The RDBMS 110 server or engine executes the SQL queries, and their flat results (streams of tuples) are converted into XML by the XML generator 106. The translator 104 also takes a source description, which is an XML document specifying systems information needed to contact the source: the protocol (e.g. JDBC), the connection string, and a source-specific query driver. The driver translates RXL expressions into the source's query language, which is typically a dialect of SQL. Although one skilled in the art will appreciate that other query languages can be supported. For example, the executable RXL query (C) is translated into the following SQL query:
select c.pid as pid, c.item as item
from Clothing c, SalePrice s
where c.category = "outerwear",
c.pid = s.pid,
s.price < 0.5 * c.retail
sort by c.pid
and into the XML template:
<results>
<result ID=Result("Acme Clothing")>
<supplier>"Acme Clothing"</supplier>
<name ID=Name($pid, $item)>$item</name>
</result>
</results>
where the variables $pid and $item refer to the attributes pid and item in the SQL query's select clause; the template generation is described in more detail in section II, part A below. After translation, the SQL query is sent to the relational engine, RDBMS 110 and the resulting tuple stream is fed into the XML generator 106, which produces the XML output. In this example, the translation requires only one SQL query. In general, there may be several ways to translate a complex RXL query into one or more SQL queries and to merge tuple streams into the XML result. Choosing an efficient evaluation strategy may be important when the RXL query returns a large result, e.g., if the entire XML view if materialized. SilkRoute can have one or more evaluation strategies, which can generate one SQL query for each disjunct of an RXL sub-query, which should be in disjunctive-normal form (DNF). Each SQL query has a sort-by clause, making it possible for the XML generator 106 to merge the queries into an XML document in a single pass. E. Alternative Approaches The above example of the present invention has been described in terms of a general approach for exporting relational data into XML. Other approaches are possible, and in some cases, may be more desirable. Currently, the most widely used Web interfaces to relational databases are HTML forms with CGI scripts. A script can translate user inputs into SQL queries, and the query answers can be rendered in HTML. The answers could be generated just as easily in XML. Forms interfaces may be appropriate for casual users, but may not be appropriate for data exchange between applications, because they limit the application to only those queries that are predetermined by the form interface. Aggregate queries, for example, are rarely offered by form interfaces. In another alternative implementation of the invention, the data provider can either pre-compute the materialized view or compute the view on demand whenever requested by an application. This alternative can be feasible when the XML view is small and the application needs to load the entire XML view in memory, e.g., using the DOM (document object module defined by the World Wide Web Consortium DOM Recommendation) interface. However, pre-computed views are not dynamic (i.e., their data can become stale) and are not acceptable when data freshness is critical. Another alternative implementation of the present invention uses a native XML database engine, which can store XML data and process queries in some XML query language. XML engines will not replace relational databases, but a high-performance XML engine might be appropriate to use in data exchange. For example, one could materialize an XML view using SilkRoute and store the result in an XML engine that supports an XML query language, thus avoiding the query composition cost done in SilkRoute. However, XML engines may not match the performance of commercial SQL engines anytime soon. In addition, this approach can suffer from data staleness, and incur a high space cost (e.g., for disk space) because it duplicates the entire data in XML. II. Query Composition In this section, the query composition algorithm is described. As discussed previously, an RXL query, such as V, takes a relational database as an input and returns an XML document as an output. The XML-QL user query, such as U, which is written against V, takes an XML document as an input and returns an XML document. For any database D, the result of U can be computed by first materializing V(D), denoted as XMLD, and then computing U(XMLD). The query composition problem is to construct an equivalent RXL query C, where C=U ? V. In other words, it would be desirable to construct an RXL query C that is guaranteed to yield the same result as U and V for any database D, that is, C(D)=U(V(D)). C takes as an input a relational database and returns an XML document. With C, the construction of the intermediate result XMLD is skipped. As an example, RXL view query (V), and XML-QL user query (U) can be used with the result of the composition, C, being composed RXL query (C). Before describing the details, a brief intuitive description is given. The key observation is that all XML components (tags, attributes, #PCDATA) present in XMLD are explicitly mentioned in the construct clause(s) of RXL view query (V). When XML-QL user query (U) is evaluated on XMLD, its patterns are matched with these components. The key idea is to evaluate XML-QL user query (U) on the templates of RXL view query (V) directly, without constructing XMLD. During this evaluation only the patterns are considered and not the filters occurring in user query (U). In this example, user query (U) has a unique pattern that mentions <supplier>, <company>, <product>, <name>, <retail>, and <sale> with a particular nesting, and all these tags also occur in the templates of view query (V) under the same nesting. RXL view query (V) is shown again below, this time after the matching, with the matched tags in bold font.
construct
<supplier ID=Supp( )>
<company ID=Comp( )>"Acme Clothing"</company>
{
from Clothing $c
where $c.category = "outerwear"
construct
<product ID=Prod($c.pid)>
<name ID=Name($c.pid,$c.item)>$c.item</name>
<category
ID=Cat($c.pid,$c.category)>$c.category</category>
<retail ID=Retail($c.pid,$c.price)>$c.price</retail>
{ from SalePrice $s
where $s.pid = $c.pid
construct
<sale ID=Sale($c.pid,$s.pid,$s.price)>$s.price</sale>
}
{ from Problems $p
where $p.pid = $c.pid
construct
<report code=$p.code
ID=Prob($c.pid,$p.pid,$p.code,$p.comments)>
$p.comments
</report>
}
</product>
}
</supplier>
That is, the RXL view query (V) is shown with patterns from the XML-QL user query (U) highlighted. Once the matching is done, the composed query (C) can be constructed in a second step, as follows. The construct clause of the composed query (C) is the same as the construct clause of the XML-QL user query (U), modulo variable renaming. The from and where clauses of the composed query (C) include both of the "relevant" from and where clauses in the view query (V) and of all the where filter conditions in the user query (U), modulo variable renaming. This completes the construction of composed query (C). In this example, the "relevant" from and where clauses are:
from Clothing $c, SalePrice $s
where $c.category = "outerwear", $s.pid = $c.pid
and the where filter condition in user query (U) is $sale<0.5 $retail which becomes the following after variable renaming:
where $s.price < 0.5 * $c.retail
Inspection of the composed RXL query (C) shown above indicated that the from and where clauses, together, form the from and where clauses of the composed query (C). FIG. 2 depicts the architecture of query composition according to the present invention. The pattern matcher 140 implements a first step, which involves evaluating user queries (U) on view query (V) templates. During the first step, the user query (U) patterns are matched with view query (V) templates. The result is a solutions relation, R, in which each tuple represents one match. Multiple matches may occur if the patterns contain alternation, e.g., <company.vertline.organization>, or Kleene-star operators, e.g., <*.supplier>, or tag variables <$elm>. A rewriter 150 carriers out a second step by taking the remaining clauses (the from and where of the view query (V) and the construct of the user query (U)) and the relation R, and rewriting each solution tuple into one RXL clause. The result is the composed query C. The illustrative query composition technique can be viewed as an example of partial evaluation, where the patterns are evaluated at composition time (a.k.a. compile time) on view query (V) templates, and the filters and constructors are evaluated at run time when the new RXL view is evaluated. Section II, parts A-D of the description describe the internal representation of view and user queries and provide a detailed description of an illustrative composition algorithm according to the present invention. A pseudo code version of the algorithm appears in section III. A. Step 1: Pattern Matching In Step 1, the solutions relation R that contains all matchings of user query (U) patterns with view query (V) templates can be constructed. 1. Construct the View Tree. For the composition algorithm, the view query V may be represented by a data structure called a view tree, which includes a global template and a set of datalog rules. The global template can be obtained by merging all view query (V) templates from all its construct clauses. Nodes from two different templates may be merged if and only if they have the same Skolem function. Hence, each Skolem function occurs exactly once in the view tree. The datalog rules are non-recursive. Their heads are the Skolem functions names, and their bodies include relation names and filters. The datalog rules can be constructed as follows. For each occurrence of a Skolem function F in a view query (V), one rule is constructed of the form F(x, y, . . . ):-body, where body is the conjunction of all from and where clauses in the scope where F occurs. When a rule is associated with a Skolem function, then that rule guards the Skolem function and its corresponding XML element. In both the template and datalog rules, the tuple variables used in RXL can be replaced by column variables. Below is the template of the view tree for the RXL query on the left and the datalog rules of the view tree for the RXL query on the right according to the illustrative example of the present invention described herein.
<supplier ID=Supp( )> Supp( ) :- true
<company ID=Comp( )>Acme Clothing Comp( ) :- true
</company>
<product ID=Prod($cpid)> Prod($cpid) :-Clothing($cpid,
_,$category, _,_) ,$category= "outerwear"
<name ID=Name($cpid,$citem)> Name($cpid, $citem) :-
Clothing($cpid, $citem, $category, _,_),
$category = "outerwear"
$citem
</name>
<category ID=Cat($cpid,$ccategory)> Cat($cpid, $ccategory) :-
Clothing($cpid, _, $category, _,_),
$category = "outerwear"
$ccategory
</category>
<retail ID=Retail($cpid,$cprice)> Retail($cpid, $cprice) :-
Clothing($cpid, _, $category, _, $cprice),
$category = "outerwear"
$cprice
</retail>
<sale ID=Sale($cpid,$spid,$sprice)> Sale($cpid, $spid, $sprice)
:-
$sprice Clothing($cpid, _, $category,
_,_), $category = "outerwear",
</sale> SalePrice($spid, $sprice), $cpid
= $spid
<report ID=Rep($cpid,$ppid,$pcode,$pcmnts) Rep($cpid, $ppid, $pcode,
$pcmnts) :-Clothing($cpid, _, $category, _,_),
code=$Pcode>
$category = "outerwear",
$pcmnts Problems($ppid, $pcode,
$pcmnts), $cpid = $ppid
</report>
</product>
</supplier>
The unique supplier element is guarded by the rule Supp( ):-true, which is always true, because no predicate expression guards the element's creation. The retail elements are guarded by the rule: Retail($cpid, $cprice):-Clothing ($cpid, _, $category, _, $cprice), $category="outerwear" which means that one retail element is created for each value of cpid and cprice that satisfies the table expression on the right-hand side. There is only one datalog rule for each Skolem function, because each function occurs once in the query view (V). 2. Evaluate User View (U) on the View Tree. Next, the patterns of user view (U) can be matched with the template of view query (V). To simplify presentation, it is assumed that the user view (U) includes a single block as represented by Equation 1: U=construct<elm>{where P, W construct T}</elm> (1) where T denotes the template, P denotes all patterns, and W denotes all filters. New, temporary variables in U's patterns can be introduced, with one variable for the ID attribute of each element in the pattern. In this example, U has a single pattern and six new variables are added, one temporary variable for each element in the pattern, as shown below.
<supplier ID=$t1>
<company ID=$t2>$company</company>
<product ID=$t3>
<name ID=$t4>$name</name>
<retail ID=$t5>$retail</retail>
<sale ID=$t6>$sale</sale>
</product>
</supplier>
The necessity of these variables and how to handle multi-block user queries are described in section II, part A(3) below. Next, U's patterns on V's template are evaluated in the standard way of evaluating patterns on a tree. In general, there may be zero, one, or more results. The results can be represented as a table R, with one column for each variable in U, and one row for each result. The values in the table are #PCDATA, Skolem terms, variables, tag names, attribute values, and attribute names, which occur in V's template. In this example, the step results in the following table R:
$t1 $t2 $company $t3 $t4 $name $t5
$retail $t6 $sale
Supp( ) Comp( ) Acme Prod($cpid) Name($cpid, $citem Retail($cpid,
$cprice Sale($cpid, $sprice
Clothing $cprice) $cprice)
$spid,
$sprice)
The column names correspond to the variables in U's single pattern shown above. The single row in R means that there exists only one matching of U's pattern with V's template. The row specifies that U's variable $name is bound to $citem in V, the variable $t3 is bound to the Skolem term Prod($cpid), and the variable $company is bound to the #PCDATA "Acme Clothing". B. Step 2: Query Rewriting In Step 2, the table R can used to construct the composed query C. Each row in R represents one match, and composed query C is the union of all possible matches. In particular, composed query C includes several parallel blocks, which denote a union in RXL. In each block, the from and where clauses contain the "relevant" datalog rules, that is the rules for the Skolem functions in the corresponding row. The construct clause of the block contains the template of the user view U. Recall that U includes a single block (Eq. 1), and that T denotes its template, P its patterns, and W its filters. Let the rows in R be r.sub.1 . . . r.sub.k. Then C includes several parallel blocks:
C = construct { <elm>{B.sub.1 } ... {B.sub.k } </elm> }
with one or more blocks corresponding to each row r.sub.i. In the next section, how blocks corresponding to one row, r.sub.i, in R are constructed is described. 1. Contruct One Block To construct the from and where clauses of one block, the clauses are represented as one datalog rule. Then, the rule is converted into a from-where clause. Let F.sub.1 . . . F.sub.n be the Skolem functions that occur in the row r.sub.i. Recall that the view tree associates one or more datalog rules to each Skolem function. Assume that there is a unique datalog rule for each Skolem function: F.sub.1 :-body.sub.1 . . . F.sub.n -body.sub.n. The block's construct clause is S.sub.0 (T) where S.sub.0 is a variable substitution defined below. For each datalog rule F.sub.i, one variable substitution S.sub.i is applied. The body of the new datalog rule is the union of all bodies after variable substitution, plus S.sub.0 (W). Thus, the new rule has the form: Q(S.sub.0 (x), S.sub.0 (y), . . . ):-S.sub.0 (W), S.sub.1 (body.sub.1), . . . S.sub.n (body.sub.n) where x, y, . . . are the variables in U's template T. Next, Q is minimized, and rewritten as a from-where clause: all relation names appearing in the from clause, and all filters appearing in the where clause. This completes the construction of one block. 2. Variable Substitutions Next, the substitutions of S.sub.0 and S.sub.1 . . . S.sub.n are defined. For all the datalog rules F.sub.1 . . . F.sub.n, the substitutions S.sub.1 . . . S.sub.n are constructed so that the expressions S.sub.1 (body.sub.1) . . . S.sub.n (body.sub.n) all have distinct variables, with one exception. For every two columns t.sub.j, t.sub.k in R, where the variable t.sub.j corresponds to an element that is the parent of t.sub.k 's element, all variables in S.sub.j (F.sub.j ( . . . )) can be shared with S.sub.k (F.sub.k ( . . . )). To compute S.sub.0, the substitutions S.sub.1 . . . S.sub.n are applied to the entire row r.sub.i and drop all columns in r.sub.i that correspond to the temporary variables $t1 . . . $t2. The new row is S.sub.0, which maps U's variables to variables, constants, and Skolem terms. When there is more than one datalog rule per Skolem function, the resulting datalog program is converted into disjunctive normal form, i.e., a disjunction of multiple conjunctive datalog rules, before generating the RXL blocks. For each conjunctive rule, the construction above can be applied to obtain one block and take the union of all such blocks. In this case, more than one block for one row r.sub.i can be obtained. In this example, table R has one row that contains the Skolem terms Supp( ), Comp( ), Prod($cpid), Name($cpid, $citem), Retail(Scpid, $cprice), and Sale(Scpid, $spid, $sprice). Their corresponding datalog rules are shown in the view tree for the RXL query in section II, part A(1) above. Next, the substitutions S.sub.1, . . . , S.sub.6 are computed such that the rules have disjoint variables with the exception of variables that have parent/child relationships. In this example, the variable $t3 is the parent of variables $t4, $t5, $t6; see the pattern in section II, part A(2) above. Therefore the Skolem term Prod($cpid) shares the variable $cpid with that in Name ($cpid, $citem), Retail ($cpid,$cprice), and Sale ($cpid, $spid, $sprice). Otherwise, all variables must be distinct. The modified rules are:
Supp( ) :- true
Comp( ) :- true
Prod($cpid) :- Clothing($cpid, _, $category1, _,_), $category1
= "outerwear"
Name($cpid, $citem) :- Clothing($cpid, $citem, $category2, _,_),
$category2 = "outerwear"
Retail($cpid, $cprice) :- Clothing($cpid, _, $category3, _, $cprice),
$category3 = "outerwear"
Sale($cpid, $spid, $sprice) :- Clothing($cpid, _, $category4, _,_)
$category4 = "outerwear",
SalePrice($spid, $sprice), $cpid = $spid
The substitution S.sub.0 is obtained directly from the table R, by dropping all columns corresponding to the new variables $t1, . . . , $t6:
S.sub.0 =
$company $name $retail $sale
Acme Clothing $citem $cprice $price
The template T of U is in the user query shown in section I, part B. The filter W of U is $sale<0.5*$retail. Only the variables $company and $name occur in T, so S.sub.0 ($company) and S.sub.0 ($name) need to be included in the rule's head; $company, however, is a constant, therefore the rule becomes:
Q($citem) :- Clothing($cpid, -, $category1, _,_), $category1 = "outerwear",
Clothing($cpid, $citem, $category2, _,_), $category2 =
"outerwear",
Clothing($cpid, _, $category3, _, $cprice), $category3 =
"outerwear",
Clothing($cpid, _, $category4, _,_), $category4 = "outerwear",
SalePrice($spid,
$sprice), $cpid = $spid,
$sprice < 0.5 * cprice
The last line is S.sub.0 (W). Minimizing Q, the following equivalent query is obtained:
Q($citem) :- Clothing($cpid, _, $category3, _, $cprice),$category3 =
"outerwear",
SalePrice($spid, $sprice), $cpid = $spid, $sprice < 0.5
* cprice
Finally, the rule can be converted into from and where clauses, and a construct clause can be added whose template is S.sub.0 (T):
from Clothing($cpid, _, $category3, _, $cprice),
SalePrice($spid, $sprice)
where $category3 = "outerwear", $cpid = $spid, $sprice < 0.5 *
cprice
construct <result ID= Result("Acme Clothing")>
<supplier>Acme Clothing </supplier>
<name> $citem </name>
</result>
Lastly, column variables are replaced by tuple variables, and the single-block query C shown in section I, part C can be obtained. C. Other Exemplary Implementations The above-described exemplary implementation of the present invention illustrates a simple example of query composition. Below, several other exemplary implementations of the present invention that illustrate more complex cases are described. 1. View Tree for Multi-block Query Consider the two block RXL query in section I, part A. Below, on the left side is the view tree template for the two block query and on the right side is the datalog for the two block query.
<view ID=View( )> View( ) :- true
<product ID=Prod($name)> Prod($name) :- Clothing($name,
_)
Prod($name) :- Clearance($name,
$ddisc), $ddisc > 50
<name ID=Name($name)> Name($name) :- Clothing($name,
_)
$name Name($name) :-
Clearance($name, $ddisc), $ddisc > 50
</name>
<price ID=Price($name, $cprice)> Price($name,$cprice) :-
Clothing($name, $cprice)
$cprice
</price>
<discount ID=Discount($name, $ddisc)> Discount($name, $ddisc)
:- Clearance($name, $ddisc),
$ddisc
> 50
$ddisc
</discount>
</product>
</view>
In the RXL query, the Skolem functions Prod and Name occur twice. In the view tree, each function has two corresponding datalog rules, but in the template, they occur once. 2. Multiple Rows In general, R may contain multiple rows. To illustrate R with multiple rows, the query V described and shown in section II, part A(1) is employed. R is composed with the following XML-QL user query U':
construct <results> {
where <supplier.product.(retail .vertline. sale)>$val</> in
"http://acme.com/products.xml"
construct <price>$val</price>
} </results>
The regular expression supplier.product.(retail.vertline.sale) matches a retail or a sale element nested within a supplier and a product element. It is analogous to the XPath expression /supplier/product/retail.vertline.sale. There are two matches of U with V, which produce two rows in R:
$tl $t2 $t3 $val
Supp( ) Prod($cpid) Retail($cpid, $cprice) $cprice
Supp( ) Prod($cpid) Sale($cpid, $spid, $sprice) $sprice
The temporary variables $t1, $t2, $t3 are for supplier, product, and retail .vertline. price, respectively. The composed query C has two blocks: C=construct<results>{B1}{B2}</results> The relevant datalog rules for the first row are those for Supp, Prod and Retail of view query described in section II, part A(1). No variables are renamed, because $t2 is the parent of $t3. The generated datalog rule after minimization is: Q($cprice):-Clothing($cpid, _, $category, _, $cprice), $category="outerwear" and it produces C's first block B1:
B1 = from Clothing $c
where $c.category="outerwear"
construct <price>$c.price</price>
The relevant datalog rules for the second row are those for Supp, Prod, and Sale. As before, no variables are renamed, and the datalog rule is: Q($cprice):-Clothing($cpid, _, $category, _, _), $category="outerwear", SalePrice($spid, $sprice), $cpid=$spid which produces C's second block B2:
B2 = from Clothing $c, SalePrice $s
where $c.category="outerwear", $c.pid = $s.pid
construct <price> $c.pid </price>
3. Adding Template Variables The temporary variables $t1, $t2, etc. added to U's patterns play an important role, as revealed by the next example. The query V, written directly with column variables, is:
V = construct <v ID=H( )> { from T($x, $y)
construct <a ID = F($x)>
<b ID = G($x, $y)> $y </b>
</a>
} </V>
and the two XML-QL queries U, U' can be considered:
U = construct <results> { where
<v><a><b>$z1</b>
<b>$z2</b></a></v>
construct
<result><z1>$z1</z1>
<z2>$z2</z2></result>
} </results >
U'= construct <results>{ where
<v><a><b>$z1</b></a>
<a><b>$z2</b></a> </v>
construct
<result><z1>$z1</z1>
<z2>$z2</z2></result>
} </results>
Both return pairs of <b> values, but the first query returns pairs where both <b>'s are in the same <a> element. Without temporary variables in U's patterns, the relation R would be the same for U and U'. After introducing the new variables, the two relations R have different column names, and as expected, they produce two distinct composed queries. 4. Renaming Variables in Datalog Rules Continuing with the previous example, the need for the substitutions S.sub.1, S.sub.2, . . . . First, V's view tree is constructed:
<v ID=H( )> H( ) :- true
<a ID = F($x)> F($x) :- T($x, _)
<b ID = G($x, $y)>$y</b> G($x, $y) :- T($x, $y)
</v>
Next, the composition with U' is illustrated. Five temporary variables are added and U's pattern becomes:
<v ID=$t1> <a ID=$t2><b
ID=$t3>$z1</b></a>
<a ID=$t4><b
ID=$t5>$z2</b></a>
</v>
Matching the pattern with the template produces one row in R:
$t1 $t2 $t3 $z1 $t4 $t5 $z2
H( ) F($x) G($x, $y) $y F($x) G($x, $y) $y
Intuitively the variable $y in the $z1 column is different from $y in the $z2 column, because they match different <b> elements, possibly in different <a> elements. This distinction is made precise by the renaming step. Thus, after variable renamings, the five relevant datalog rules become:
H( ) :- true
F($x1) :- T($x1, _)
G($x1, $y1) :- T($x1, $y1)
F($x2) :- T($x2, _)
G($x2, $y2) :- T($x2, $y2)
and the composed query C, after query minimization, is:
construct <results> from T($x1, $y1), T($x2, $y2)
consruct <result><z1>$y1</z1>
<z2>$y2</z2></result>
</results>
5. XML-QL Queries with Block Structure. In general, U may have several blocks, both nested and parallel. For multi-block user queries, a different table R for each block in U can be constructed, in the same way in which the XML-QL query processor handles multiple blocks. Tables corresponding to parallel blocks are independent; for nested blocks, there is a distinct inner table that corresponds to each row in the outer table. The composed query C follows the same block structure, except that one block in U may generate multiple parallel blocks in C, as described above early in this section. 6. Query Minimization Query minimization eliminates redundancies in queries, such as duplicate conditions. Query minimization can be expensive, because it is NP-complete. That is, the complexity of query minimization is exponential in the number of variable in the query. Commercial database systems often do not perform minimization, because users typically do not write redundant queries. In SilkRoute, the composed query C can be generated automatically. One condition in a view query V may appear in multiple datalog rules, and, hence be propagated as multiple copies in the generated query C. To avoid query minimization, one could trace these repetitions to the original RXL query, but care is needed to deal with variable renamings. For RXL queries with large parallel blocks, however, query minimization may be unavoidable. D. Aggregation Queries Briefly, it is described how aggregations in XML-QL queries can be "pushed" into composed RXL views and evaluated by the target RDBMS according to the present invention. In both XML-QL and RXL, Skolem terms can be used to specify the values by which aggregate expressions are grouped. Suppose a reseller wants to count the total number of reports for each defective product. This can be expressed in XML-QL as follows:
where <supplier.product ID=$pid>
<name>$n</>
<report>$r</>
</supplier> in "http://acme.com/products.xml"
construct <product ID=F($pid)>
<name ID=G($pid,$n)>$n</>
<totaldefects ID=H($pid)>count(*)</>
</product>
The Skolem term F($pid) in <product ID=F($pid)> asserts that all bindings for the variables $pid, $n and $r are grouped by $pid's value. Similarly, the Skolem term H($pid) specifies the grouping attributes for the aggregate function count(*), which counts the total number of bindings. This idea is similar to the GROUP BY construct in SQL. XML-QL and RXL's semantics guarantee that only one element is produced for each value of a Skolem term, e.g., one name element is emitted for each value of $n. A simple extension to datalog that accommodates aggregate functions can be used. An example of a datalog rule that can use a "generator" to count values is: C(p, q, COUNT(*)):-R(p, q) Only the last argument in the head can be an aggregate function; the other arguments specify the grouping attributes. The meaning is that C contains the set of triples (p, q, r) where r is the number of tuples in the group corresponding to values (p, q) in the relation R. Using this composition algorithm, the XML-QL query above can be rewritten as:
from Clothing $c, Problems $p
where $c.pid = $p.pid
construct <product ID=F($c.pid)>
<name ID=G($c.pid, $c.item)>$c.item</>
<totaldefects ID=H($c.pid)>count(*)</>
</product>
Note that the aggregate function can be "pushed" into the RXL view. When this view is materialized, the aggregation can be evaluated by the relational engine. Significantly, this query can be evaluated efficiently, because commercial database systems are often highly optimized for aggregation queries. III. A Composition Algorithm--Pseudocode In the formal description of the algorithm above, a notation for describing the types of values that are manipulated, e.g., view trees, XML-QL blocks, are needed. Types are denoted by grammar rules, such as the following:
Node :- Tag, Rule, [ Node ]
Rule :- SkolemTerm, [ Condition ]
Condition :- TableExpr(String, [ Var ])
.vertline. Filter(BoolExpr)
.vertline. Or([ Condition ], [ Condition ])
These rules specify that a view tree Node is composed of a tag, a rule, and a list of children nodes. A Rule is composed of a Skolem term (its head) and a conjunctive list of conditions (its body). A Condition is either a table expression, a filter expression, or the disjunction of two lists of conjuncts. An XML-QL block is represented by a list of patterns, a list of filters, and a template. An RXL block is represented by a list of conditions and a template:
XMLQL :- [ Pattern ], [ Filter ], Template
RXL :- [ Condition ], Template
A template is either: a constant string; a variable; an element, which includes a tag and list of nested templates; or a nested query. To simplify presentation, templates are polymorphic, i.e., an XML-QL template contains only a nested XML-QL block and similarly, for an RXL template.
Template :- Const(String)
.vertline. Var(String)
.vertline. Element(Tag, [ Template ])
.vertline. NestedQuery(XMLQL)
.vertline. NestedQuery(RXL)
Finally, a canonical pattern is represented by the head variable (that occurs on the right-hand side of it), a regular-path expression over strings, and the target variable (that occurs in the body of an element):
Pattern :- Var, RegPathExpr, Var
In this example, each regular-path expression is one string atom, but in general, strings can be combined with the alternation (.vertline.), concatenation (.), and Kleene-star (*) operators, similar to those used in regular expressions. The composition function compose, as shown below, takes two environments, which are lists of (variable, value) pairs. Shown below is an illustrative top-level compose function for a composition algorithm:
1. // Top-level invocation of compose function
2. X_env = new [("$viewtree", Root( )]
3. S = new[ ]
4. R_block_list = compose(X_env, S, X_block)
5.
6. fun compose(Env X_env, VarMap S, XMLQL X_block) : [ RXL ] {
7. (X_patterns, X_Filters, X_template) = decompose(X_block);
8.
9. // Get pairs of(parent, child) variables from XML-QL patterns
10. X_parent_child_vars = getHeadTargetMap(X_patterns);
11.
12. // Evaluate pattern on view tree
13. R = evalPattern(X_patterns, X_env);
14.
15. // Consider each potential solution
16. R_blocks = new [ ]
17. for each r_i in R {
18. // Extend current environment with new variable bindings
19. X_env' = appendList(X_env, r_i);
20.
21. // Compute new S variable substitution from X_nodemap
22. S' = newVariables(X_envl', X_parent_child_vars, S);
23.
24. //Compute RXL block for potential solution
25. R_blocks = listAppend(oneSolution(X_env', S', X_block, r_i),
R_blocks)
26. }
27. return R_blocks
28. }
The initial environment (X_env) maps the distinguished variable $viewtree to the root of the view tree referenced by the query. The initial variable-substitution S that maps XML-QL variables to RXL expressions is empty, and X_block is the top-level XML-QL block (lines 1-3). In this example, $viewtree is bound to the root of the tree defined in the RXL query in section II, part A(1). The result of compose is a list of RXL blocks. In the pseudo-code, XML-QL expressions are prefixed by X_ and RXL expressions by R_. Function compose (line 7) decomposes X_block into its patterns, filters, and template, and rewrites each nested pattern in a canonical form as a list of unnested patterns. New temporary variables are introduced to represent the intermediate nodes in the nested pattern. On line 13, the patterns are evaluated in the current environment, producing R, which maps XML-QL variables to nodes and constants in the view tree. Each tuple in R represents one possible rewriting of the XML-QL query over the view. For each tuple r_i, the current environment is extended with the new variable bindings (line 19). Function newVariables (line 22) computes the new mappings of XML-QL and RXL variables to common RXL variables. In summary, newVariables recovers the correspondence between Skolem terms that share a common ancestor in the XML-QL pattern; this correspondence determines the mappings for RXL variables. For XML-QL variables, the mapping is simple. If the corresponding value is a leaf node or constant value, the variable is replaced by its value in the substitution mapping S described in section II, part B. If the corresponding value is an internal node, the variable is replaced by the complete RXL expression that computes that element under the substitution S. Lastly, function oneSolution (line 25) takes the new environment and computes the new RXL blocks, which are appended to the list of other potential solutions. The composition algorithm for the oneSolution function in the following code constructs the RXL block in three steps.
1. // Return new RXL block for potential solution in r_i
2. fun oneSolution(Env X_env, VarMap S, XMLQL X_block, Env r_i) : [ RXL ]
{
3. R_conditions = new [ ]
4. // For each XML-QL variable X_v in X_block
5. foreach X_v in getVariables(X_block) {
6. // Get view-tree node bound to X_v
7. R_node = project(r_i, X_v);
8. // Get rule associated with view-tree node
9. (R_tag, R_rule, R_children) = R_node
10. // Get body of rule
11. (R_head, R_body) = R_node;
12. foreach R_condition in R_body {
13. R_condition' = makeCopy(R_condition)
14. // Rename head variables in R_condition' and add to
R_conditions
15. R_conditions = cons(rewriteR(S, R_condition'),
R_conditions)
16. }
17. }
18. // Rename variables in X_filters and add to R_conditions
19. foreach X_filter in X_filters
20. R_conditions = cons(rewriteX(X_env, S, X_filter), R_conditions)
21.
22. // Put conditions in disjunctive normal form, i.e., [[ Condition ]]
23. R_disjuncts = to_DNF(R_conditions)
24.
25. / Rename variables in X_template
26. R_template = rewriteX(X_env, S, X_template)
27.
28. R_blocks = [ ]
29. // Construct new RXL block: solution conditions + RXL template
30. foreach R_conjunct in R_disjuncts
31. R_blocks = cons(new RXL(R_conjunct, R_template), R_blocks)
32.
33. return RXL_blocks
34. }
First, for each XML-QL variable X_v in X_block, it projects X_v's value from the solution tuple r_i. Its value is a view-tree element and an associated rule, whose head and body are projected in R_head and R_body, a list of conditions. Function makeCopy (line 13) assigns fresh variable names to all free variables in R_condition, i.e., those that do not occur in the rule's head. Function rewriteR (line 15) rewrites the new rule, using the variable mapping S. The new condition is added to the conjunctive conditions in R_conditions. Second, the function rewriteX (line 20) rewrites the XML-QL filters in X_filters and adds those to R_conditions. Third, the function to_DNF (line 22) puts the new conditions in disjunctive normal form. On line 23, rewriteX rewrites the XML-QL template to produce the new RXL template. Finally, one new RXL block is created for each list of conjuncts in R_disjuncts, and the union of all these blocks is returned. The rewriteX and rewriteR functions in the composition algorithm of the rewrite function below line 14 replace XML-QL and RXL variables by their new names in S.
1. // rewriteX rewrites XML-QL expression as RXL expression
2. fun rewriteX(Env X_env, VarMap S, X_Expr E) {
3. fun substX(E) {
4. case E of
5. Var(v) = lookupMap(S, v)
6. Const(c) = new Const(c)
7. Element(T, X) = new Element(T, mapList(substX, X))
8. Relop(op, E1, E2) = new Relop(op,
substX(E1), substX(E2))
9. // Cases for all types of BoolExprs . . .
10. // Recursively compose and rewrite nested
XML-QL query
11. NestedQuery(X-block) = new
NestedQuery(compose(X_env, S, X_block))
12. }
13. return substX(E)
14. }
15. // rewriteR renames RXL variables.
16. fun rewriteR(S varmap, R_Expr E) {
17. fun substR(E) {
18. case E of
19. Var(v) = lookupMap(S, v)
20. TableExpr(name, vars) = new
TableExpr(name, mapList(substR, vars))
21. Filter(b) = new Filter(substR(b))
22. Or(11, 12) = newOr(mapList(subst, 11),
mapList(substR, 12))
23. // Cases for all types of BoolExprs . . .
24. NestedQuery(RXL(conditions, template)) =
25. new NestedQuery(new RXL(mapList(substR, conditions),
substR template))
26. }
27. return substR(E)
28. }
The "helper" functions substX and substR perform the variable substitutions. Note that rewriteX calls compose recursively to rewrite a nested XML-QL block into an equivalent nested RXL block. IV. Related Systems BM's DB2 XML Extender provides a Data Access Definition (DAD) language that supports both composition of relational data in XML and decomposition of XML data into relational tables. DAD's composition feature, like RXL, supports generation of arbitrary XML from relational data. Unlike RXL, the criteria for grouping elements is implicit in the DAD and DAD specifications cannot be nested arbitrarily. More significantly, XML Extender does not support query composition, however, DAD could be used as a view-definition language in a SilkRoute architecture. V. General Discussion SilkRoute is a general, dynamic, and efficient framework for viewing and querying relational data in XML. SilkRoute is an XML-export tool that can support arbitrarily complex, virtual views of relational data and support XML user queries over virtual views. The ability to support arbitrary views is critical in data exchange between inter-enterprise applications, which must abide by public XML schemas and cannot reveal the underlying schemas of their proprietary data. SilkRoute has many benefits. For example, the fragment of the relational data requested by a user query need only be materialized; that requested data can be produced on demand; and the relational engine can perform most of the computation efficiently. SilkRoute has one translation strategy, which generates one SQL query for each RXL sub query, which must be in disjunctive-normal form (DNF). In practice, RXL view queries can be arbitrary boolean combinations of table and filter expressions; for example, parallel RXL blocks often construct parts of complex elements independently, i.e., they express unions. User queries over such views often produce composed queries with many unions. Any RXL sub-query can be normalized into multiple sub-queries in DNF, which can result in a quadratic increase in the number of sub-queries to evaluate. In practice, multiple queries in DNF can be translated directly into SQL, for example, by using SQL's union-join constructs. Similarly, nested RXL queries often express left outer joins, e.g., the parent sub-query is the left relation and the child sub-query is the right relation. Two SQL queries can be generated, one for parent and child, but one SQL query suffices. In addition to reducing the number of SQL queries, each individual RXL sub-query can be minimized, i.e., redundant expressions can be eliminated, so that the resulting SQL query is also minimal. Techniques exist for query minimization, but general algorithms are NP-complete. Heuristic algorithms are projected to be effective for RXL queries, because RXL's nested block structure can help identify those expressions that most likely are redundant. XML-QL and SilkRoute can be implemented in Java. SilkRoute has drivers for Oracle and MySQL database servers. VI. XML View-evaluation Several illustrative embodiments of the present invention address the problem of evaluating efficiently an XML view in the context of SilkRoute, a relational to XML middleware system in SilkRoute. A relational to XML view can be specified in the declarative query language RXL. An RXL query has constructs for data extraction and for XML generation. One aspect of the present invention involves materializing large RXL views. In practice, large, materialized views may be atypical: often the XML view is kept virtual, and users' queries extract small fragments of the entire XML view as described above in section I. In one implementation of the present invention, data-export or warehousing applications, which require a unique, large XML view of the entire database can be supported. In this case, computing the XML view may be costly, ranging from minutes to several hours, and query optimization can yield dramatic improvements. In the article "Efficiently Publishing Relational Data as XML Documents", VLDB 2000, pp. 65-76, by Shanmugasudaram et al., the authors evaluate experimentally a variety of approaches for publishing XML data in a relational query engine. In a data-warehousing scenario, the XML document defined by an RXL view typically exceeds the size of main memory. Therefore, the sorted, outer union approach described by Shanmugasudaram et al. is suitable for the data-warehousing scenario. The sorted, outer union approach constructs one large, SQL query from the view query; reads the SQL query's resulting tuple stream; and then adds XML tags. The SQL query includes several left-outer joins, which construct the atomic data values of the XML document. The left-outer join expressions are combined in outer unions. The resulting tuples are sorted by the XML element in which they occur, so that the XML tagging algorithm can execute in constant space. SilkRoute described above uses an approach, in which the view query is decomposed into multiple SQL queries that do not contain outer joins or outer unions. Each result is sorted to permit merging and tagging of the tuples in constant space. Such a scheme is referred to as a fully partitioned strategy. Neither the sorted, outer union approach nor the fully partitioned strategy is optimal. This is surprising because the sorted outer-union approach generates only one SQL query, and therefore, has the greatest potential for optimization by the RDBMS. However, for complex RXL queries, the outer-union query is too large and complex for the target RDBMS to optimize effectively. The sorted, outer-union strategy produces a query that is slower than the queries produced by the fully partitioned strategy due to the inability of the RDBMS optimizer to handle a complex query. An optimal strategy generates multiple SQL queries, but less queries than the fully partitioned strategy. Thus, the optimal SQL queries may contain outer joins and outer unions. XML tagging still uses constant space, because it merges sorted tuple streams. A strategy generating less multiple SQL queries than the fully partitioned strategy executes 3 to 20 times faster than the sorted, outer-union and fully partitioned strategies. Recognizing that the optimal strategy executed much faster than the other alternatives, illustrative embodiments of the present invention employ an algorithm for decomposing an RXL view query into an optimal set of SQL queries. In developing an algorithm, two issues were considered. First, the RXL view query can be very large, because it constructs an XML document and, therefore, it may be as complex as the output schema. Public XML schemas have up to several hundreds elements and several thousand attributes, therefore any program or query generating XML documents for those XML schemas will have a comparable complexity. Consequently, this rules out exhaustive-search strategies such as the dynamic-programming algorithm disclosed by P. Selinger et al. in "Access Path Selection in a Relational Database Management System", Proceedings of ACM SIGMOD Int'l Conf. on Management of Data, pp. 23-34. Second, the algorithm needs to function in a middle-ware system, and, therefore cannot rely on RDBMS-specific heuristics. The present invention can provide a greedy optimization algorithm to address the XML view-evaluation problem. An exemplary algorithm according to the present invention decomposes a large query into a set of small queries, and for example, can decompose an RXL query into a set of SQL queries. The search algorithm is generally guided by query-cost and data-size estimates provided by the RDBMS. The algorithm can facilitate obtaining an optimal strategy on, for example, two RXL views of a TPC-H (Transaction Processing Performance Council ad-hoc decision support benchmark) database. While an illustrative greedy optimization algorithm according to the present invention is primarily described herein as being implemented in a middleware system, it will be appreciated that the associated techniques can be applied in RDBMS engines that generate XML internally. Generating XML in an RDBMS engine is generally more efficient than external generation in a middleware system, because the binding cost, i.e., the cost of binding application variables to the tuples, dominates execution time. In an illustrative implementation of the present invention, a greedy optimization algorithm may be used as a preprocessing step, to split the XML-view query into multiple SQL queries of manageable size that can be optimized by the RDBMS. A greedy optimization algorithm according to the present invention is best adapted to scenarios where publishing large XML documents is necessary. In other scenarios, a user query may request only a subset of the XML view, where the resulting document is small. For example, a user may ask for all orders of customer "Smith" placed in "October, 2000". In this instance, the resulting XML document is much smaller than the XML view containing, for example, all customers and their orders. The scenario described above in section I can effectively handle subsets of the XML view. According to the scenario described above, the XML view of the database is virtual, and the user query being employed is an XML-QL user query. In forming an algorithm, it is necessary to consider the search space for the XML view definition. In a large scale XML publishing scenario, the query strategy should scale to arbitrary large XML outputs, and it should be decoupled from a relational engine's query optimizer. Shanmgasundaram et al. considered strategies without these restrictions, and found two to be effective, the unordered outer union strategy and the CLOB De-correlated queries. In the unordered outer union strategy, the tagger uses a main memory hash table to assemble the XML objects, which requires the XML view fit in main memory. In CLOB De-correlated queries, the XML result is constructed by the relational engine, which is also effective when the XML view fits in main memory. Despite the effectiveness of these two strategies, it has been discovered in connection with the present invention that the sorted outer union strategy is more effective as query complexity and result size increase. VII. RXL Query Example An illustrative implementation of the RXL query language is provided in this section. As an illustrative database, the TPC Benchmark.TM. H (TPC-H) database (see www.tpc.org), which contains information about parts, the suppliers of those parts, customers, and their part orders, will be used. An illustrative fragment of the database's schema specified in datalog syntax is provided.
Supplier(*suppkey, name, address, nationkey)
Partsupp(*partkey, suppkey, availqty)
Part(*partkey, name, mfgr, brand, size, retailprice)
Customer (*custkey, name, address, nationkey, phone)
LineItem(*orderkey, partkey, suppkey, lineno, qty, price)
Orders(*orderkey, custkey, status, price, date)
Nation(*nationkey, name, regionkey)
Region(*regionkey, name)
Key attributes are denoted by the `*` prefix. For example, the Supplier relation has four attributes and its key is the suppkey attribute. It is assumed assume that information in the TPC database needs to be exported in the format determined by the DTD below.
<?xml encoding="US-ASCII"?>
<!ELEMENT suppliers (supplier*)>
<!ELEMENT supplier (name, nation, region, part*)>
<!ATTLIST supplier ID ID>
<!ELEMENT name (#PCDATA)>
<!ELEMENT nation (#PCDATA)>
<!ELEMENT region (#PCDATA)>
<!ELEMENT part (name, order*)>
<!ATTLIST part ID ID>
<!ELEMENT order (orderkey, customer, cnation)>
<!ATTLIST order ID ID>
<!ELEMENT orderkey (#PCDATA)>
<!ELEMENT customer (#PCDATA)>
This DTD specifies the XML format for the entire contents of the TPC database for the purpose of, for example, data warehousing. Each supplier element includes its name, its nation, the geographical region of the nation, and a list of the supplier's parts. Each part element includes a part name and a list of orders pending for the part. Each order element includes an orderkey, the associated customer, and the customer's nation. The name, nation, region, and customer elements all contain strings. To keep the example simple, a DTD that follows naturally from the relational schema has been designed. Although it should be understood that in practice, this may not be the case. DTDs are created by agreement between partners, for the purpose of data exchange, and generally do not match each partners relational schema exactly. The DTD is also not unique. For example, a different DTD might be specified by a public consortium of parts suppliers to provide access to order information for their customers. These requirements rule out automatic generation of the DTD or of the mapping between the relational schema and the DTD. An RXL query mapping the relational data to an XML output that is valid with respect to the DTD is shown below, and more particularly, an RXL view query of TPC-H Database, which may also be referred to as Query 1 herein.
from Supplier $s
construct
<supplier><name>$s.name</name>
{ from Nation $n
where $s.nationkey = $n.nationkey
construct <nation>$n.name</nation>
{ from Region $r
where $n.regionkey = $r.regionkey
construct <region>$r.name</region>
}
}
{ from Partsupp $ps, Part $p
where $s.suppkey $ps.suppkey,
$ps.partkey = $p.partkey
construct
<part><name>$p.name</name>
{ from LineItem $1, orders $o
where $ps.partkey = $1.partkey,
$ps.suppkey = $1.suppkey,
$1.orderkey = $o.orderkey
construct
<order><orderkey>$o.orderkey</orderkey>
{ from Customer $c
where $o.custkey = $c.custkey
construct
<customer>$c.name</customer>
.thrfore. from Nation $n2
where $c.nationkey =
$n2.nationkey
construct
<cnation>$n2#name</cnation>
}
}
</order>
}
</part>
}
</supplier>
As in SQL, the from clause declares tuple variables that iterate over tables. In this example, $s is a tuple variable that iterates over the Supplier table. The where clause contains conditions over these variables: for example $s.nationkey=$n.nationkey is a join condition. The construct clause specifies an XML fragment that may contain expressions over the tuple variables. Three features in RXL make it possible to create arbitrarily complex XML structures: nested queries, Skolem functions, and block structure. Nested queries occur inside construct clauses to construct sets of sub-elements. The block structure permits independent sub-queries to construct different sets of elements, i.e., parallel blocks express unions. For example, the outermost query above has two sub-queries delimited by block boundaries {. . . }, each constructing a different set of elements. Skolem functions can be used to fuse objects constructed by different queries, which is especially useful in data integration. To evaluate the RXL query computing the XML view, one or more SQL queries need to be computed to extract and group the data for the XML view and then add the XML tags. Each sub-query in the view definition corresponds to an SQL query, but they are correlated, and it is unclear how to put them together. To illustrate, the simpler RXL query used in FIG. 4 shows a fragment of the above-defined query.
from Supplier $s
construct
<supplier>
{
from Nation $n
where $s.nationkey = $n.nationkey
construct <nation>$n.name</nation>
}
{
from Partsupp $ps, Part $p
where $s.suppkey = $ps.suppkey, $ps.partkey = $p.partkey
construct <part>name=$p.name/>
}
</supplier>
The set of all possible choices are best visualized on the intermediate representation for RXL queries, which is called a view tree. FIG. 3 depicts a view tree for the above-simplified RXL query. Each node corresponds to an element in one of the construct clauses in the RXL query, and is annotated by a non-recursive datalog query that computes all instances of that node in the output XML. From the queries, it is possible to derive the multiplicities of the parent/child relationships, which are indicated by the labels 1 and *, with obvious meaning. For example, in FIG. 3, the 1 between <supplier> and <nation> indicates that each <supplier> element in the output XML document will have exactly one child of type <nation>, and the * between <supplier> and <part> means that <supplier> may have arbitrarily many children of type <part>. The view tree makes it clear how to generate queries. A `1`-labeled edge requires an inner join, while a * requires a left outer join. Hence, the view tree leads to the following SQL query:
select s.suppkey, n.name, SubQuery.partkey, SubQuery.name
from Supplier s, Nation n
where s.nationkey = n.nationkey
left outer join
(select ps.suppkey as suppkey, p.name as pname
from PartSupp ps, Part p
where ps.partkey = p.partkey)
as SubQuery
on s.suppkey = SubQuery.suppkey
order by s.suppkey
An outer join is needed because there could be suppliers without parts, and they need to appear in the XML document. The order by clause groups tuples from the same supplier together and allows the tagger to construct the <supplier> element using little main memory. The above query may be referred to as a "unified" translation, because it corresponds to the entire view tree and produces one relation. It is equivalent to a sorted outer union query described in J. Shanmugasundaram et al., "Efficiently Publishing Relational Data as XML Documents" VLDB 2000, pp. 65-76. Also, the view tree can be split into connected components, and generate a separate SQL query for each such component. FIG. 4 provides an illustrative execution for systematically splitting the view tree into connected components. Execution plan (a) corresponds to the query above, while execution plans (b), (c), and (d) are three alternative ways to partition the view tree into connected components. Each execution plan produces a set of SQL queries. For example, execution plan (b) results in the two SQL queries:
select s.suppkey, n.name
from Supplier s, Nation n
where s.nationkey n.nationkey
order by s.suppkey
select s.suppkey, p.name
from Supplier s, Part p, Partsupp ps
where s.suppkey ps.suppkey and ps.partkey = p.partkey
order by s.suppkey
Notably, no outer join is needed, because the first query produces all the values for Supplier. The tagger must merge the two sorted tuple streams to produce the XML elements. For execution plan(c) in FIG. 4, the queries are:
select s.suppkey, n.name
from Supplier s, Nation n
where s.nationkey = n.nationkey
order by s.suppkey
select s.suppkey, SubQuery.partkey, SubQuery.pname
from Supplier s
left outerjoin
(select ps.suppkey as suppkey, p.name as pname
from PartSupp ps, Part p
where ps.partkey = p.partkey)
as SubQuery
on s.suppkey = SubQuery.suppkey
order by s.suppkey
Execution plan (d) in FIG. 4 corresponds to three SQL queries, which have been omitted for convenience, but will be readily apparent to those skilled in the art. FIG. 5 depicts the view tree for the large RXL query (Query 1) described above. In this view tree, there are nine edges and 2.sup.9 or 512 subsets of edges, each of which corresponds to a partition of the tree. Therefore there are 512 possible plans for splitting the tree into a collection of SQL queries; each plan including between 1 and 10 tuple streams. On a TPC/H database of 100 MB, some running times were tested:
10 queries: 1794s (569s)
5 queries: 589s (244s)
1 query: timed out after 1000 seconds
The first number is total execution time, which includes the time to execute the query at the server and to bind and transfer the data to RXL; the number in parentheses includes only the query time. The first line represents a plan that splits the query into ten small SQL queries, having sorted tuple streams that are merged by the tagger. The second line shows the best plan: it includes five SQL queries. In this case, the tagger has to merge five tuple streams. The plan on the last line is unified translation, i.e., a single SQL query. The two "extreme" plans performed poorly, but the optimal plan is order of magnitudes better than the fully partitioned plan, which one might expect to perform well. Also, several other plans, including 3, 4, and 6 SQL queries respectively, performed almost as well as the optimal plan: under 600 s (246 s). In general, there are 2.sup..vertline.E.vertline. possible translations of an RXL query into one or more SQL queries, where .vertline.E.vertline. is the number of edges in the query's corresponding view tree. Given the exponential number of potential plans, SilkRoute uses heuristics to choose a good plan. Those heuristics are described later herein. In commercial XML middle-ware products, the user typically must write these SQL queries, which effectively "hard wires" the evaluation plan into the definition of the XML view. This may seem like a reasonable requirement, but in practice, it is difficult to choose a good plan. The simplest choices are to always produce one unified relation as in execution plan (a) in FIG. 4 or fully partitioned relations as in execution plan (d) in FIG. 4. However, as will be described later here, the unified and fully partitioned plans are often substantially slower than the optimal plans. VIII. Plan Generation Below, a formal definition of a view tree is provided and the algorithm for translating a partitioned view tree into one or more SQL queries is described. FIG. 6 depicts the architecture of an illustrative query planner and translator. The planner partitions a view tree into one or more subtrees; for each subtree, one SQL query is generated. The translator submits the SQL queries to the underlying RDBMS, reads in the result relations, and constructs one integrated (logical) relation. A tuple in the integrated relation represents a path from the root element to a leaf element in the result XML document. The XML document is constructed by re-nesting the tuples in the result relation and tagging each element. A. View Tree An RXL view query V is represented by a view tree, which includes a global XML template and a set of datalog rules. The global XML template can be obtained by merging all the view query V's XML templates from all its construct clauses. Every XML template has an associated Skolem term that uniquely identifies the XML template in an RXL view. The user may assign a Skolem term explicitly to a template in the view query, or if absent, SilkRoute assigns a term. Elements from two different XML templates are merged if and only if they have the same Skolem function, hence each Skolem function occurs exactly once in the view tree. For example, the tree in FIG. 3 represents the global XML template for the RXL view query fragment described above. The Skolem terms S1(suppkey.sub.(11)), S1.1(suppkey.sub.(11), name.sub.(2,1)), S1.2(suppkey.sub.(1,1), pname.sub.(2,2)) uniquely identify the supplier, nation, and part elements, respectively. The XML generator for SilkRoute uses the XML template to instantiate the result document. A view tree's datalog rules are non-recursive. Their heads are Skolem terms, and their bodies include relation names and filters. The datalog rules are constructed as follows. For each occurrence of a Skolem function F in view query V, one rule of the form: F(x, y, . . . ):-body, where body is the conjunction of all from and where clauses in the scope where F occurs can be constructed. When a rule is associated with a Skolem term, the rule guards that Skolem term and its corresponding XML element. In both the XML template and in the datalog rules, the tuple variables used in RXL by column variables can be replaced. The head of a datalog rule corresponds to an element in the global XML template, and the body of a rule defines the conditions under which the element is created. When assigning a Skolem term to a node, a Skolem-function index is associated with each Skolem function and a Skolem-term variable index is associated with each Skolem variable. A Skolem-function index uniquely defines the tag and location of a node. These indices are used to sort the tuples of partitioned relations during tagging of the XML document. A Skolem-function index (l.sub.1,l.sub.2 . . . ) is assigned to each node in breadth-first order. For example, the Skolem function S1 is assigned to the root, and S1.1 is assigned to the root's first child. Each Skolem-term variable v is assigned a Skolem-term variable index (p, q) as follows. Let n.sub.v be the node closest to the root that has v in its Skolem term. Then, p is equal to the level of n.sub.v in the view tree, and q is the first integer such that (p, q) is unique for all variables in the tree. For example, the variable suppkey .sub.(1,1) is assigned index (1,1), because its containing element is at level one, and it is the first variable in the term. Similarly, the variable name..sub.(2,1) is assigned index (2,1), because it is the first variable that appears in a term at level two. Finally, the variable pname.sub.(2,2) is assigned (2,2), because it is the second variable that appears in a term at level two. B. View-tree Partitioning As described in section VII, the planner produces one plan for each spanning forest of the view tree, so it produces 2.sup..vertline.E.vertline. plans, where .vertline.E.vertline. is the number of edges in the view tree. For example, given the view tree for Fragment 1 in FIG. 3, possible plans are shown in FIG. 4. The planner produces one SQL query for each tree in a spanning forest. In section V, a greedy algorithm is presented that heuristically chooses a small subset of the 2.sup..vertline.E.vertline. plans. For each tree in a spanning forest, the schema of the relational relation that computes the nodes in the tree needs to be defined. To illustrate, consider the unified execution plan (a) in FIG. 4 that corresponds to the entire view tree. Given the example fragment of TPC-H database instance,
Supplier(supp#1, "USA Metalworks", "New York", usa#24)
Supplier(supp#2, "Romana Fspanola", "Madrid", spain#3)
Supplier(supp#3, "Fonderie Francais", "Paris", france#19)
Nation(usa#24, "USA", reg#1)
Nation(japan#3, "Spain", reg#2)
Nation(rom#19, "France", reg#3)
PartSupp(part#4, supp#1, 100)
PartSupp(part#12, supp#1, 320)
PartSupp(part#20, supp#3, 64)
Part(part#4, "plated brass", mfgr#3, "Brand1", "S", 904.00)
Part(part#12, "anodized steel", mfgr#4, "Brand2", "M", 912.01)
Part(part#20, "polished nickel", mfgr#1 , "Brand3", "L", 920.02)
the corresponding query produces a fragment of an XML document such as:
<supplier key="supp#1">
<nation>USA</nation>
<part>plated brass</part>
<part>anodized steel</part>
</supplier>
<supplier key="supp#2">
<nation>Spain</nation>
</supplier>
<supplier key="supp#3">
<nation>France</nation>
<part>polished nickel</part>
</supplier>
The result of the SQL query is the relation for plan (a) in FIG. 4 below:
L.sub.1, L.sub.2 s.suppkey.sub.(1,1) n.name.sub.(2,1) p.name.sub.(2,2)
1 1 supp#1 USA
1 2 supp#1 plated brass
1 2 supp#1 anodized steel
1 1 supp#2 Spain
1 1 supp#3 France
1 2 supp#3 polished nickel
In general, let T.sub.i be one spanning tree in a partitioned view tree T, and let SFI_maxlen(T.sub.i) be the maximum length of the Skolem-function indices in T.sub.i. Let R.sub.i be the partitioned relation that corresponds to T.sub.i. Then, the relational schema of R.sub.i is defined as attrs(R.sub.i)=SFI_attrs.sub.i U STV_attrs.sub.i ; where Skolem-function index attributes: SFI_attrs.sub.i ={"L.sub.j ".vertline.l=j=SFI_maxlen(Ti)}, and Skolem-term variable attributes: STV_attr.sub.j ={v.vertline.v is a Skolem-term variable in T.sub.i }. To illustrate, FIG. 7 contains partitioned relations for all the execution plans in FIG. 4. The upper-left relation corresponds to the tree containing only the supplier node. Its Skolem-function index contains only one label L.sub.1 and one Skolem-term variable suppkeyi.sub.(1,1). An instance of a partitioned relation I(R.sub.i) as follows. Let V.sub.(p,q) be a Skolem-term variable. Then, (L.sub.1 : l.sub.1, . . . , L.sub.m : l.sub.m, L.sub.m+1 : l.sub.m+1, . . . L.sub.SFI.sub..sub.-- .sub.maxlen(Ti) : l.sub.SFI.sub..sub.-- .sub.maxlen(Ti), V.sub.(p1, q1) : v.sub.(p1, q1), . . . , V.sub.(pk,qk) : V.sub.(pk,qk))? I(R.sub.i) if and only if there is an element E in the result XML document that corresponds to a node in T.sub.i, where (l.sub.1 l.sub.2 . . . l.sub.m) is the Skolem-function index for E, and the Skolem-term variables for E are included in STV.sub.attrj and have non-null values for E. The tuples in I(R.sub.i) are sorted by L.sub.1, V.sub.(1,1) . . . , V.sub.(1,n1), L.sub.2, V.sub.(2,1) . . . , V.sub.(2,n2), etc. This order is consistent with the structural relationship between the elements in the result XML document. C. Integration and Tagging In SilkRoute, the integrated relation is logical, namely, SilkRoute does not materialize the relation. Instead, the result XML document is constructed directly from the partitioned relations. An exemplary XML generation algorithm containing the integration and tagging algorithm is shown below:
Types:
Relation A partitioned relation
Tuple Tuple in the integrated relation (L.sub.1, V.sub.(1.1) . . . ,
V.sub.(1,n1), . . . L.sub.m, V.sub.(m,1). . . , V.sub.(m,nm))
Tag Set of tags
SFI A Skolem-function index (l.sub.1, . . . , l.sub.m)
STV A Skolem-term variable value (v.sub.(1,1), . . . , v.sub.(1,n1),
. . . , v.sub.(m,1), . . . v.sub.(m,nm))
Functions and procedures:
getTuple: {Relation} ? Tuple Returns the next tuple from the integrated
relation
getTag: SFI ? Tag Returns the tag associated with a
Skolem-function index
getSFI: Tuple ? SFI Projects the Skolem-function index values
from a tuple
getValues: Tuple ? STV Projects the Skolem-term variable values
from a tuple
getLeaf: Tuple .times. SFI ? String.vertline.null Returns the leaf (atomic)
string value associated with a
Skolem-term, or null if it has no atomic
value
SAXWriter An implementation of a SAX Writer
EmitXML Emits tags and values for a given tuple
generateXML Given partitioned relations, generates
result XML Output:
An XML document
procedure generateXML(Relations: {Relation}) {
SAXWriter.startDocument( )
// Initialize all Skolem-function indices and Skolem-term values to
null
sfi' = (L.sub.1 : null, . . . , L.sub.m : null)
values' = (V.sub.(1,1) : null, . . . , V.sub.(m,n) : null)
// Get next tuple from Relations in order (L.sub.1, V.sub.(1,1) . . . ,
V.sub.(1,n1), . . . L.sub.m, V.sub.(m,1) . . . , V.sub.(m,n))
while ((tuple = getTuple(Relations)) != EOF) {
sfi = getSFI(tuple)
values = getValues(tuple)
if(sfi' != sfi or values' != values) {
// Get maximum index where new tuple and old tuple differ
let l.sub.1, . . . , l'.sub.m = sfi'
l.sub.1, . . . , l.sub.m = sfi
n.sub.1 = max{i.vertline.sfi.L.sub.i = sfi'.L},
n.sub.2 = max{i.vertline.values.V.sub.(ij) =
values'.V.sub.(ij) },
in emitXML(m', min.sub.(n1,n2) +1, m, tuple)
}
sfi' = sfi
values' = values
}
SAXWriter.endDocument ( )
}
procedure emitXML(m', n, m, tuple)
sfi = getSFI(tuple)
// Close all open elements upto new element
for (i=m'; i=n; i=i-1)
SAXWriter.endElement(getTag(sfi.L.sub.1 . . . sfi.L.sub.i))
// Open all containing elements upto new element
for (i=n; i=m; i=i+1)
SAXWriter.startElement(getTag(sfi.L.sub.1, . . . , sfi.L.sub.i))
leafValue = getLeaf(sfi.L.sub.1, . . . , sfi.L.sub.i, tuple)
if (leafValue != null) SAXWriter.characters (leaf Value)
}
}
Intuitively, the integration and tagging algorithm merges the partitioned tuple streams into one tuple stream, nests the tuples, and tags their values. An illustrative implementation of the integration and tagging algorithm according to the present invention includes several steps. These steps may include receiving one or more tuple streams each containing multiple tuples such that each tuple has a corresponding node index (e.g., Skolem term); comparing node indices of two tuples; and emitting an XML tag based on the result of the comparison. Tuple streams may vary in width (i.e., contain differing numbers of fields). Therefore, the tuple streams may be logically integrated before the streams are compared. For example, if one tuple stream contains three fields, and another tuple stream contains 20 fields, the result of logically integrating the two tuple streams is one tuple stream that contains 23 fields and that is ordered by the tuple's node indices in document order. Each tuple in the integrated tuple stream is processed in order. The node index of the current tuple is then compared to the node index of the previous tuple to determine where the tuple's data belongs in the XML output document and what XML tag should be emitted. Node indices specify uniquely the level at which the tuple's data should appear in the XML output document. The algorithm compares the current node index to the previous node index, and an XML open or close tag is then emitted according to the difference in levels of the view tree between the two node indices. If the difference between the node indices is greater than one level of the view tree, more than one XML tag may be emitted. For example, if a tuple has the node index of 1.2 and a previous node index is 1.1.1, two XML close tags and one open tag would be emitted. The integration and tagging algorithm can compare two node indices at a time. Once a node index is compared to the previous node index, the algorithm does not need to refer to the previous node again. Therefore, the required memory size of the algorithm can depend only on the number of nodes and Skolem-term variables in the view tree. It need not depend on the size of the database instance; therefore the algorithm scales well as the size of the underlying database, and corresponding XML document, can increase. D. SQL Generation SilkRoute uses outer-union plans, as described by J. Shanmugasundaram et al. in "Efficiently Publishing Relational Data as XML Documents", VLDB 2000, pp.65-76, to construct SQL queries for partitioned relations. The outer-union plans can be implemented using the `with` clause and the outer-join and union operators of SQL. For example, one possible SQL query for execution plan (a) of FIG. 4) uses a left-outer join to combine the root (supplier) node with its children nodes, and it uses an outer union to combine the children nodes (the nation and part elements).
select 1 as L1, L2, s.suppkey, SubQuery.name, SubQuery.pname
from Supplier s
left outerjoin
((select 1 as L2, n.nationkey as nationkey, n.name as name, null as
suppkey, nulI as pname
from Nation n)
union
(select 2 as L2, null as nationkey, null as name, ps.suppkey as
suppkey, p.name as pname
from Partsupp ps, Part p
where ps.partkey = p.partkey))
as SubQuery
on (L2=1 and s.nationkey = SubQuery.nationkey) or (L2=2 and s.suppkey =
SubQuery.suppkey)
sort by L1, s.suppkey, L2, SubQuery.nationkey, SubQuery.name,
SubQuery.pname
The structure of outer-union plans using left-outer joins and unions corresponds closely to the structure of subtrees. The sub-query for a node n in a view tree and the sub-queries of n's children can be combined with an outer join. The sub-queries for n's children (siblings) can be combined with an outer union. The outer union is necessary because sibling nodes have different relational structures: in the relation that computes a node m, the attributes of m's siblings are null values. The SQL query above can be simplified further by view-tree reduction described in the next section. Note that some of the plans SilkRoute produces do not require outer union, outer join, or the with clause. For example, a fully partitioned plan (i.e., with no edges) does not require any of these constructs. Plans with no branches (i.e., no sibling nodes) do not require the union operator. This characteristic is especially useful in a middle-ware system, because all SQL engines do not necessarily support all these constructs. In those cases, SilkRoute can choose permissible plans based on the source description of the underlying RDBMS. E. View-tree Reduction The view tree provides a flexible intermediate representation, because it supports generation of multiple execution plans. Its flexibility, however, can introduce redundant queries in the view tree and in corresponding execution plans. A single condition in an RXL query often guards the creation of multiple elements. For example, the part element and its sub-element name in the RXL view query of TPC-H database (Query 1) are both guarded by the condition $s.suppkey=$ps.suppkey, $ps.p | ||||||