OSLC Core Query Syntax v.Next DRAFT

By: The OSLC Core Workgroup

Overview

This document includes query syntax features that will be in a later version of OSLC Core Query Syntax. It is a copy of the OSLC Core Query Syntax v2 specification taken in May 25, 2010 -- one which still includes oslc.from, oslc.limit and oslc.offset, all of which were removed from the OSLC Core Query Syntax v2 specification during convergence.

This document defines a standard set of OSLC query parameters that other OSLC domain specifications MAY use to query resources.

Graph Patterns

The OSLC standard query parameters define graph patterns that are matched against the resources managed by the service, which we view as comprising an RDF graph. The pattern matching process begins at some starting subject resource and matches a set of triples in the RDF graph. The starting subject resources used for pattern matching are determined by the the Query Capability. A Query Capability defines a starting subject resource, namely the base URI itself. The result of the query is the set of all triples matched by the pattern, beginning the matching processing with the starting subject resource defined by the Query Capability.

The full graph pattern defined by the query parameters is composed of two kinds of graph pattern, which we refer to as property tree patterns and member list patterns. These patterns MAY be used individually or MAY be combined in a single query, but one or the other MUST be present. The result of matching the full graph pattern is the union of the results of matching each kind of pattern.

Property Tree Patterns

A property tree pattern is defined by the presence of the oslc.properties query parameter. In RDF, the data model of a resource is viewed as consisting of a set of properties of the resource. The values of these properties are either literal values, such as numbers, strings, and dates, or links to other resources, which may have their own properties. Linked resources may link to other resources, and so on, to arbitrary depths. However, the scope of the query is limited to the boundaries of the service as defined by the appropriate domain specification. Properties of linked resources may be viewed as nested properties of the initial resource. The oslc.properties query parameter lets you match both the immediate and nested properties of a starting subject resource.

Member List Patterns

A member list pattern is defined by the presence of at least one of the following query parameters: oslc.from, oslc.where, oslc.searchTerms, oslc.orderBy, oslc.offset, oslc.limit, and oslc.select. In this pattern, the starting subject resource is viewed as a container of other resources. The members of the container are defined by one or more multi-valued membership properties of the starting subject resource. The membership properties are either given by the oslc.from query parameter, or default to the properties described as membership properties (via oslc:isMemberProperty) in the Resource Shape of the starting subject resource. The pattern matching process is defined by the following sequence of steps. Default processing is done in steps where the corresponding query parameter is not present:

1. oslc.from - create the member list defined by membership properties of the starting subject resource, and arrange it in some order
2. oslc.where - filter the member list, keeping only those member resources that satisfy the boolean test on the member resource properties
3. oslc.searchTerm - score each member resource using a full text search on it text-valued properties, and sort them in descending order of score
4. oslc.orderBy - sort the members using the sort keys, using the full text search score as the primary key
5. oslc.offset - discard the specified number of member resources from the start of the list
6. oslc.limit - discard any member resources beyond the specified number from the end of the list
7. oslc.select - match the immediate and nested properties of the member resources (like in oslc.properties)

Prefixed Names

In addition to the preceeding query parameters, we also define oslc.prefix which is used to define the prefixes for prefixed names that appear in the other query parameters. The oslc.prefix query parameter is not itself part of the graph pattern. Other specifications MAY use oslc.prefix to define prefixes for prefixed name used other query parameters.

Query Parameters

The query parameters defined here are intended to satsify simple query requirements, and be easy to both use and implement. More complex query requirements should be satisfied using a full query language, e.g. SQL or SPARQL.

This specification formally defines the syntax of the query parameters using Backus-Naur Form (BNF) and informally illustrates their semantics with simple examples. For a more complete definition of the semantics see OSLC Simple Query Semantics V1.

The query parameters MAY contain characters that MUST be URL encoded when transmitted in an HTTP request. The examples below show unencoded query parameters for clarity. See URL Encoding for a further discussion of this topic.

This specification defines the syntax of a set of related query parameters that MAY be used to perform queries. We have adopted a consistent naming convention for these query parameters to identify them as common across all OSLC specifications, namely all query parameter names are prefixed with the characters " oslc. " The following sections define these query parameters.

An OSLC domain specification MAY use some or all of these query parameters, and SHOULD use these rather than defining new query parameters that have the same or very similar meanings.

Each of these query parameters SHOULD appear at most once in a query. The behavior is undefined when a query parameter appears more than once.

In the following sections, syntax is formally defined using a common extended form of BNF. Informal definitions and comments are delimited by /* and */.

oslc.prefix

This query parameter does NOT define a graph pattern. It is used by other query parameters that contain prefixed names.

In the following examples, property names include a URI prefix, i.e. dc: or foaf:. An OSLC service SHOULD predefine URI prefixes for its properties. Here we assume that OSLC has predefined the Dublin Core ( dc:) and Friend of a Friend ( foaf:) prefixes. However, OSLC resources SHOULD also be open to new content, which means that new properties may not have predefined URI prefixes. We therefore need a way to define new URI prefixes in queries. The oslc.prefix query parameter lets you specify URI prefixes used in property names. For example, suppose the foaf: prefix was not predefined. The following URL illustrates the use of the oslc.prefix query parameter to define it:

http://example.com/bugs/4242?oslc.prefix=foaf=<http://xmlns.com/foaf/0.1/>&oslc.properties=foaf:lastName,...

Syntax

The syntax of the oslc.prefix query parameter is defined by the oslc_prefix term in the following BNF grammar:

oslc_prefix ::= "oslc.prefix=" prefix_defs
prefix_defs ::= prefix_def ("," prefix_def)*
prefix_def  ::= prefix "=" uri_ref_esc
prefix      ::= PN_PREFIX
PN_PREFIX   ::= /* see "SPARQL Query Lanaguage for RDF", http://www.w3.org/TR/rdf-sparql-query/#rPN_PREFIX */
uri_ref_esc ::= /* an angle bracket-delimited URI reference in which > and \ are \-escaped. */

oslc.properties

This query parameter defines a property tree pattern.

The oslc.properties query parameter lets you specify the set of properties to match in a property tree pattern. Both immediate and nested properties may be specified. A nested property is a property that belongs to the resource referenced by another property. Nested properties are enclosed in brace brackets, and this nesting may be done recursively, i.e. a nested property may contain other nested properties.

For example, suppose we have a bug report resource at the following URL:

 http://example.com/bugs/4242

Suppose this bug resource has properties such as dc:title, dc:description, and dc:creator, and that dc:creator refers to a person resource that has properties such as foaf:givenName and foaf:familyName. Suppose you want a representation of the bug report that includes its dc:title and the foaf:givenName and foaf:familyName of the person refered to by its dc:creator. The following URL illustrates the use of the oslc.properties query parameter to include those properties:

http://example.com/bugs/4242?oslc.properties=dc:title,dc:creator{foaf:givenName,foaf:familyName} 

Syntax

The oslc.properties query parameter is defined by the oslc_properties term in the following BNF grammar:

oslc_properties ::= "oslc.properties=" properties
properties      ::= property ("," property)*
property        ::= identifier | nested_prop | wildcard
identifier      ::= PrefixedName 
PrefixedName    ::= /* see "SPARQL Query Lanaguage for RDF", http://www.w3.org/TR/rdf-sparql-query/#rPrefixedName */
nested_prop     ::= property "{" properties "}"
wildcard        ::= "*"

oslc.from

This query parameter defines part of a member list pattern. If a member list pattern is defined but this query parameter is not present then it defaults to the set of all membership properties defined in the Resource Shape of the starting subject resource, i.e. the member list is the set of all resources that are related to the starting subject resource by one or more of the membership properties.

The OSLC resource data model is defined in terms of the properties that a given type of resource may have. Some properties define attributes of resources and some define links to other related resources. Membership is an important kind of link property. Membership properties are in general multi-valued and relate a given subject resource to zero or more member resources. A resource may have more than one kind of membership property, in which case the members are all resources that are linked to the subject resource by some membership property.

In general, there may be a very large number of member resources associated with a given subject resource, and it is often useful to search this set for member resources that satify some conditions. The oslc.from query parameter lets you specify which membership properties to use for this search. It is like the FROM clause in a SQL statement. The set of member resources thus defined may be further refined using boolean conditions specified in the oslc.where query parameter.

The oslc.from query parameter is not restricted to membership properties. It MAY contain any property.

For example, suppose the following URL represents the set of all change requests managed by a service:

http://example.com/requests

Suppose that this service manages two kinds of change requests, namely bugs and features, and that these are related to the base URL via the membership properties cm:memberBug and cm:memberFeature. An HTTP GET request on the base URL would return all the bugs and features.

The following query would also return all the bugs and features since the base resource has only the two specified membership properties:

http://example.com/requests?oslc.from=cm.memberBug,cm:memberFeature 

However, the following query would return only the bugs:

http://example.com/requests?oslc.from=cm:memberBug

Syntax

The syntax of the oslc.from query parameter is defined by the oslc_from term in the following BNF grammar:

oslc_from   ::= "oslc.from=" identifiers
identifiers ::= identifier ("," identifier)* 

oslc.where

This query parameter defines part of a member list pattern. If a member list pattern is define but this query parameter is not present then it defaults to the boolean condition true, i.e. it matches all members of the member list.

As discussed in the section on oslc.from, a resource may have an associated set of related resources defined by one or more membership properties. Service consumers often need to search for subsets of these related resources that satisfy certain conditions. The oslc.where query parameter lets you specify the conditions that these related resources must satisfy. It is like the WHERE clause of a SQL statement.

For example, suppose that the following URL represents the set of all bug resources managed by a service, and that is has a single membership property, cm:memberBug:

http://example.com/bugs 

Suppose bug reports have a dc:identifier property that gives their bug number, e.g. "4242". The following URL filters this set and selects only those bugs that have a dc:identifier property whose value is "4242":

http://example.com/bugs?oslc.where=dc:identifier="4242" 

Note that in the preceding query, the oslc.from query parameter is absent so it defaults to all the membership properties, and in this case there is only one, namely cm:memberBug.

Conditions may use the usual binary comparison operators and be combined using the boolean conjunction operator, " and ". For example, suppose bugs have cm:severity and dc:created properties. The following example finds high severity bugs created after April 1, 2010:

http://example.com/bugs?oslc.where=cm:severity="high" and dc:created>"2010-04-01"

The following URL illustrates the use of nested properties. It finds the set of bugs created by John Smith:

http://example.com/bugs?oslc.where=dc:creator{foaf:givenName="John" and foaf:familyName="Smith"}

Syntax

The syntax of the oslc.where query parameter is defined by the oslc_where term in the following BNF grammar:

oslc_where    ::= "oslc.where=" compound_term
compound_term ::= simple_term (space? boolean_op space? simple_term)*
simple_term   ::= term | scoped_term
space         ::= " " /* a space character */
boolean_op    ::= "and"
term          ::= identifier_wc comparison_op value | identifier_wc space in_op space? in_val
scoped_term   ::= identifier_wc "{" compound_term "}"
identifier_wc ::= identifier | wildcard
comparison_op ::= "=" | "!=" | "<" | ">" | "<=" | ">="
in_op         ::= "in"
in_val        ::= "[" value ("," value)* "]"
value         ::= uri_ref_esc | literal_value
literal_value ::= boolean | decimal | string_esc (LANGTAG | ("^^" PrefixedName))?
boolean       ::= "true" | "false"
decimal       ::= /* see "XML Schema Part 2: Datatypes Second Edition", http://www.w3.org/TR/xmlschema-2/ */
string_esc    ::= /* a string enclosed in double quotes, with certain characters escaped. See below. */
LANGTAG       ::= /* see "SPARQL Query Lanaguage for RDF", http://www.w3.org/TR/rdf-sparql-query/#rLANGTAG */

boolean_op

The boolean_op term represents a boolean operation that lets you combine simple boolean expressions to form a compound boolean expression.

The only boolean operation allowed is " and " which represents conjunction. The boolean operator " or " for disjunction is not allowed in the interests of keeping the syntax simple. The effect of " or " in the simple case of testing a property for equality with one of several values can be achieved through the use of the " in " operator. For example, the following query finds bugs with severity " high " or " medium ":

 http://example.com/bugs?oslc.where=cm:severity in ["high","medium"]

space

The space term represents a single space character. A space character MAY be used to delimit the binary_op term in the compound_term term to improve readability.

comparison_op

The comparison_op term represents one of the following binary comparison operators:

in_op

The in_op term represents the operator " in " which is a test for equality to any of the values in a list. The list is a comma-separated sequence of values, enclosed in square brackets, whose syntax is defined by the term in_val.

value

The value term represents either a URI reference ( uri_ref_esc) or a literal value ( literal_value).

literal_value

The literal_value term represents either a plain literal or a typed literal.

A plain literal is a quoted string ( string_esc), optionally followed by a language tag ( LANGTAG). For example, "Bonjour"@fr is a plain literal with a language tag for French. If the range of a property is National Language strings and no language tag is provided in a literal value, then the service SHOULD infer the language tag from the HTTP request or provide a default value.

A typed literal is formed by suffixing a quoted string ( string_esc) with "^^" followed by the prefixed name ( PrefixedName) of a datatype URI. If the range of a property includes literal values from more than one datatype, then a typed literal MUST be used in order to avoid ambiguity. Otherwise a plain literal MAY be used and the service SHOULD infer the datatype.

The terms boolean and decimal are short forms for typed literals. For example, true is a short form for "true"^^xsd:boolean, 42 is a short form for "42"^^xsd:integer and 3.14159 is a short form for "3.14159"^^xsd:decimal.

decimal

The decimal term represents a decimal number as defined in XML Schema Part 2: Datatypes Second Edition. As mentioned above, this term is a short form for typed literals whose datatype URIs are either xsd:integer or xsd:decimal. An integer literal value is a special case of a decimal literal value, namely one in which the decimal point is omitted from the lexical representation. For example, 42 is a valid decimal number which happens also to be a valid integer and so it is a short form for the typed literal "42"^^xsd:integer.

string_esc

The string_esc term represents an arbritrary sequence of characters. The sequence of characters is enclosed in double quote (") characters. Therefore, the double quote character itself MUST be escaped. All occurances of the double quote character in the string MUST be replaced by the sequence \" and all occurances of the backslash character (\) MUST be replaced by the sequence \\. This escaping MUST be undone to give the actual value of the string.

oslc.searchTerms

This query parameter defines part of a member list pattern. If a member list pattern is defined but this query parameter is not present then the matching process assigns an effective score of 0 to each member of the list and does NOT change the order of the member list

Resource properties often contain text so it is useful to search for resources that contain specified terms. The oslc.searchTerms query parameter lets you perform a full text search on a set of resources. In a full text search, each resource is matched against the list of search terms and assigned a numeric score. A high score indicates a good match. The matching resources are returned in the response, sorted by the score in descending order. Each resource that is returned in the response is annoted with a special property, oslc:score, that gives its match score.

An OSLC domain specification that supports full text search SHOULD specify which resource properties are indexed so that search results are consistent across implementations.

When oslc.searchTerms is used in the request, each matching resource (hit) in the response MAY contain an oslc:score property. Note that oslc:score is not purely a property of the resource since it also depends on the search terms. It is therefore a pseudo-property whose validity is limited to the HTTP response.

The oslc:score property MUST be a non-negative number and SHOULD be in the range from 0-100. Results MUST be ordered with the entry with the largest oslc:score occuring first.

The oslc.orderBy query parameter MAY be used with oslc.searchTerms. When oslc.orderBy is used with oslc.searchTerms the result MUST be first sorted in descending order by the oslc:score pseudo-property, and then by the other sort keys specified in oslc.orderBy. This behavior is like prepending the list of sort keys specified in oslc.orderBy with the key -oslc:score. However, the pseudo-property oslc:score MUST NOT appear explicitly in oslc.orderBy.

The oslc.limit and oslc.offset query parameters MAY be used in combination with oslc.searchTerms. For example, the following query returns the top 10 bugs that deal with database performance:

http://example.com/bugs?oslc.limit=10&oslc.searchTerms="database","performance" 

The oslc.where query parameter MAY be used with oslc.searchTerms. When oslc.where is used with oslc.searchTerms then the set of resources searched for matches MUST be restricted to only those resources that satisfy the conditions in oslc.where. For example, the following query returns the top 10 high severity bugs that deal with database performance:

http://example.com/bugs?oslc.where=cm:severity="high"&oslc.limit=10&oslc.searchTerms="database","performance" 

Syntax

The syntax of the oslc.searchTerms query parameter is defined by the oslc_searchTerms terms in the following BNF grammar:

oslc_searchTerms ::= "oslc.searchTerms=" search_terms
search_terms     ::= string_esc (, string_esc)* 

oslc.orderBy

This query parameter defines part of a member list pattern. If a member list pattern is define but this query parameter is not present then there are no sort keys and the matching process does NOT change the order of the member list.

The oslc.orderBy query parameter lets you sort the result set. It is like the ORDER BY clause of a SQL statement.

You can specify a list of one or more immediate or nested properties of the resources in the member list, and a sort direction for each where " + " means ascending order and " -" means descending order. The following example sorts the high severity bugs by the family and given names of the creator, with most recently created first:

http://example.com/bugs?oslc.orderBy=dc:creator{+foaf:familyName,+foaf:givenName},-dc:created
    &oslc.where=cm:severity="high"

The properties in the oslc.orderBy list are sort keys. The member list is sorted by the sort keys, in the indicated direction. The sorting order of the property values MUST be the same as that used for evaluating the binary comparison operators in the oslc.where query parameter.

Each sort key SHOULD be the name of a single-valued property of the each resource in the result set. The sorting behavior is undefined if the sort key properties are not single-valued.

Syntax

The syntax of the oslc.orderBy query parameter is defined by the oslc_orderBy term in the following BNF grammar:

oslc_orderBy        ::= "oslc.orderBy=" sort_terms
sort_terms          ::= sort_term ("," sort_term)* 
sort_term           ::= scoped_sort_terms | ("+" | "-") identifier
scoped_sort_terms   ::= identifier "{" sort_terms "}" 

oslc.offset

This query parameter defined part of a member list query. If a member list query is defined but this query parameter is not present then it has an effective value of 0 and the matching process does NOT discard any resources from the start of the member list.

The oslc.offset query parameter lets you specify where in result set to begin the contents of the HTTP response. An oslc.offset of 0 omits no members of the result set. When you use oslc.offset used in conjunction with oslc.limit, you can break up the member list into chunks that will be returned in the HTTP response.

For example, the following query returns the next 100 most recent high severity bugs:

http://example.com/bugs?oslc.where=cm:severity="high"&oslc.orderBy=-dc:created&oslc.limit=100&oslc.offset=100

Syntax

The syntax of the oslc.offset query parameter is defined by the oslc_offset term in the following BNF grammar:

oslc_offset        ::= "oslc.offset=" nonNegativeInteger
nonNegativeInteger ::=  /* see "XML Schema Part 2: Datatypes Second Edition", http://www.w3.org/TR/xmlschema-2/ */

nonNegativeInteger

The nonNegativeInteger term represents a non-negative integer value (reference: XSD Datatypes).

oslc.limit

This query parameter defines part of a member list pattern. If a member list pattern is defined and this query is not present then it has an effective value of infinity and the matching process does NOT discard any resources from the end of the member list.

In practice, a result set may contain a large number of resources, so it is useful to limit the number of resources that are returned in the HTTP response. The oslc.limit query parameter lets you limit the number of resources included in the response.

For example, the following query returns the 100 most recently created high severity bugs:

http://example.com/bugs?oslc.where=cm:severity="high"&oslc.orderBy=-dc:created&oslc.limit=100

Syntax

The syntax of the oslc.limit query parameter is defined by the oslc_limit term in the following BNF grammar:

oslc_limit      ::= "oslc.limit=" positiveInteger
positiveInteger ::=  /* see "XML Schema Part 2: Datatypes Second Edition", http://www.w3.org/TR/xmlschema-2/ */

positiveInteger

The positiveInteger term represents a positive integer value (reference: XSD Datatypes).

oslc.select

This query parameter defines part of a member list pattern. If a member list pattern is defined but this query parameter is not present then the matching process does NOT match any properties of the resources in the member list.

The oslc.select query parameter lets you specify which immediate and nested properties of the resources in the member list to match. It is like the SELECT clause of a SQL statement.

The syntax of the oslc.select query parameter is the same as that of the oslc.properties query parameter. However, the property names that appear in the oslc.select query parameter are those that belong to the resources in the member list as opposed to those that belong to the starting subject resource.

For example, the following URL finds the high severity bugs and includes their creation date and the family name of the creator in the HTTP response:

 http://example.com/bugs?oslc.select=dc:created,dc:creator{foaf:familyName}&oslc.where=cm:severity="high"

Syntax

The oslc.select query parameter is defined by the term oslc_select in the following BNF grammar:

oslc_select ::= "oslc.select=" properties

The oslc.select query parameter uses the same syntax as the oslc.properties query parameter. The difference between them is in the meaning of the identifers. In oslc.properties, the identifiers are the names of properties of the starting subject resource. In oslc.select, the starting subject resource has an associated list of member resources, and the identifers are the names of properties of the resources that are contained in the member list

For example, suppose that a bug resource has the membership property cm:comment that relates it to comment resources, and that the bug and each comment have dc:modified properties that give their last modification dates. The following query specifes that the response should include the modification date of bug 4242 and the modification date of each of its comments:

 http://example.com/bugs/4242?oslc.properties=dc:modified&oslc.from=cm:comment&oslc.select=dc:modified

wildcard

As in the case of oslc.properties, the wildcard is equivalent to the list of all properties of the member resources. The query parameter oslc.select=* specifies that pattern MUST match all the properties of the resources in the member list.

URL Encoding

The query parameter syntax defined in this specification permits the use of characters that MUST be properly encoded when transmitted in HTTP requests. For example:

Not encoded:

?oslc.where=dc:title="test case 1" and dc:modified>="2008-12-02T18:42:30"

?oslc.where=dc%3Atitle%3D%22test%20case%201%22%20and%20dc%3Amodified%3E%3D%222008-12-02T18%3A42%3A30%22

Appendix A: References

These are the specifications referenced by the OSLC Core Query Guidance.

• BNF Backus-Naur Form
• CSS - Cascading Style Sheets Level 2 Revision 2 v2.1
• Dublin Core - Dublin Core Metadata Element Set, Version 1.1
• FOAF - Friend of a Friend (FOAF)
• HTML5 Posting Message- W3C HTML5 postMessage
• HTTP - Hyper-text Transfer Protocol (HTTP/1.1)
• RDF/XML Concepts - RDF/XML Concepts and Abstract Syntax
• RDF/XML Syntax - RDF / XML Syntax Specification (Revised)
• URI Syntax - URI Generic Syntax
• Window Object - Window Object 1.0 Window.name
• XML Namespaces - Namespaces in XML 1.0 (Third Edition)
• XML Base - XML Base (Second Edition)
• XSD Datatypes - XML Schema Part 2: Datatypes Second Edition