|Type of format||markup language|
SXML is an alternative syntax for writing XML data, using the form of S-expressions. It is also a set of implementations that provide typical XML-processing functionalities that operate on the SXML syntax.
Textual correspondence between SXML and XML for a sample XML snippet is shown below:
<tag attr1="value1" attr2="value2"> <nested>Text node</nested> <empty/> </tag>
(tag (@ (attr1 "value1") (attr2 "value2")) (nested "Text node") (empty))
The following two observation can be drawn from the above example:
- Textual notations for XML and SXML are much alike; informally, SXML textually differs from XML in relying on round brackets instead of angular braces.
- Additionally, SXML is not only a straightforward textual notation for XML data, but also has a directly-corresponding primary data structure for the LISP family of functional programming languages, thus providing an illustrative approach for processing XML data with a general-purpose programming language.
Similarity between XML and S-expressions reified in SXML allows achieving close integration between XML data and programming language expressions, resulting in illustrativeness and simplicity of XML data processing for an application programmer.
- 1 Motivation
- 2 XML, XML Information Set and SXML
- 3 SXML Specification
- 4 Example
- 5 SXML features
- 6 SXML shortcomings
- 7 Citations
- 8 External links
An attempt to combine two different languages (for example, XPath and Java) leads to a problem known as impedance mismatch. Impedance mismatch problem consists of two aspects:
- Different data models. E.g. XPath models an XML document as a tree, while most general purpose programming languages have no native data types for a tree.
- Different programming paradigms. Say, XSLT is a functional language, while Java is object-oriented, and Perl is a procedural one.
- Nested lists (S-expressions in Scheme) provide a natural representation for nested XML documents. Scheme represents its code and data as nested lists of dynamically typed elements. XML document, being a hierarchical structure of nested XML elements, can be thought of as a hierarchical nested Scheme list (so-called S-expression).
- Scheme is a functional language, as most XML-languages are (e.g. XSLT and XQuery). Scheme processes (nested) lists in a recursive manner which can be thought of as traversing/transforming the document tree.
Scheme, being a Lisp dialect, is a widely recognized scripting language. It is one of the most elegant and compact programming languages practically used: the description of Scheme standard consists of 40 pages only. Scheme is a high-level programming language, suitable for fast prototyping. Moreover, Scheme programs are generally several times more compact than the equivalent C programs.
XML and SXML textual notations are much alike: informally, SXML replaces XML start/end tags with opening/closing brackets. On the other hand, SXML is an S-expression and thus maps trivially to and from the main data structure for Scheme programming language. Consequently SXML can be easily and naturally processed via Scheme.
XML, XML Information Set and SXML
An XML document is essentially a tree structure. The start and the end tags of the root element enclose the whole content of the document, which may include other elements or arbitrary character data. Text with familiar angular brackets is an external representation of an XML document. Applications ought to deal with an internalized form: an XML Information Set, or its specializations (such as the DOM). This internalized form lets an application locate specific data or transform an XML tree into another tree.
The W3 Consortium defines the XML Information Set (Infoset) as an abstract data set that describes information available in a well-formed XML document. An XML document's information set consists of a number of information items, which denote elements, attributes, character data, processing instructions, and other components of the document. Each information item has a number of associated properties, e.g., name, namespace URI. Some properties—for example, "children" and "attributes"—are collections of other information items. Although technically Infoset is specified for XML, it largely applies to other semi-structured data formats, in particular, HTML.
XML document parsing is just one of possible ways to create an instance of XML Infoset.
It is worth a note that XML Information Set recommendation does not attempt to be exhaustive, nor does it constitute a minimum set of information items and properties. Its purpose is to provide a consistent set of definitions for use in other specifications that need to refer to the information in a well-formed XML document.
The abstract data model defined in the XML Information Set Recommendation is applicable to every XML-related specification of the W3 Consortium. Namely, the Document Object Model can be considered the application programming interface (API) for dealing with information items; the XPath data model uses the concept of nodes which can be derived from information items, etc. The DOM and the XPath data model are thus two instances of XML Information Set.
XML Information Set Recommendation itself imposes no restrictions on data structures or interfaces for accessing information items. Different interpretations are thus possible for the XML Information Set abstract data model. For example, it is convenient to consider an XML Information Set a tree structure, and the terms "information set" and "information item" are then similar in meaning to the generic terms "tree" and "node" respectively.
An information item may be also considered as a container for its properties, either text strings (e.g. name, namespace URI) or containers themselves (e.g. child elements for an XML element). The information set is thus a hierarchy of nested containers. Such a hierarchy of containers comprising text strings and other containers greatly lends itself to be described by an S-expression, because the latter is recursively defined as a list whose members are either atomic values or S-expressions themselves. S-expressions are easy to parse into an internal representation suitable for traversal; they also have a simple external notation, which is relatively easy to compose even by hand.
SXML is a concrete instance of the XML Infoset in the form of S-expressions. Infoset's goal is to present in some form all relevant pieces of data and their abstract, container-slot relationships with each other. SXML gives the nest of containers a concrete realization as S-expressions, and provides means of accessing items and their properties. SXML is a "relative" of XPath and the DOM, whose data models are two other instances of the XML Infoset. SXML is particularly suitable for Scheme-based XML/HTML authoring, XPath queries, and tree transformations.
XML and SXML can thus be considered two syntactically different representations for the XML Information Set.
As noted in the previous section, SXML is the concrete instance of the XML Infoset in the form of S-expressions. Further discussion on SXML in this section is based on the SXML specification.
<TOP> ::= ( *TOP* <PI>* <Element> )
<Element> ::= ( <name> <attributes-list>? <child-of-element>* )
<attributes-list> ::= ( @ <attribute>* )
<attribute> ::= ( <name> "value"? )
<child-of-element> ::= <Element> | "character data" | <PI>
<PI> ::= ( *PI* pi-target "processing instruction content string" )
Since an information item in the XML Infoset is a sum of its properties, a list is a particularly suitable data structure to represent an item. The head of the list, a Scheme identifier, names the item. For many items this is their (expanded) item name. For an information item that denotes an XML element, the corresponding list starts with element name, optionally followed by a collection of attributes. The rest of the element item list is an ordered sequence of element children, character data, processing instructions, and other elements in turn. Every child is unique; items never share their children even if the latter have the identical content.
The following example illustrates an XML element and its SXML form (which satisfies the <Element> production in SXML grammar).
<WEIGHT unit="pound"> <NET certified="certified">67</NET> <GROSS>95</GROSS> </WEIGHT>
(WEIGHT (@ (unit "pound")) (NET (@ (certified)) "67") (GROSS "95"))
A collection of attributes is considered an information item in its own right, tagged with a special name @. The character "@" may not occur in a well-formed XML name; therefore an <attributes-list> cannot be mistaken for a list that represents an element. An XML document renders attributes, processing instructions and other meta-data differently from the element markup. In contrast, SXML represents element content and meta-data uniformly—as tagged lists. RELAX NG—a schema language for XML—also aims to treat attributes as uniformly as possible with elements. This uniform treatment, argues James Clark, is a significant factor in simplifying the language. SXML takes advantage of the fact that every XML name is also a valid Scheme identifier, but not every Scheme identifier is a valid XML name. This observation lets us introduce administrative names such as @, *PI*, *TOP* without worrying about potential name clashes. The observation also makes the relationship between XML and SXML well-defined. An XML document converted to SXML can be reconstructed into an equivalent XML document (in terms of the Infoset). Moreover, due to the implementation freedom given by the Infoset specification, SXML itself is an instance of the Infoset.
The XML Recommendation specifies that processing instructions (PI) are distinct from elements and character data; processing instructions must be passed through to applications. In SXML, PIs are therefore represented by nodes of a dedicated type *PI*. XPath and the DOM Level 2 treat processing instructions in a similar way.
A sample XML document and its SXML representation are both shown below, thus providing an illustrative comparison between nested XML tags and nested SXML lists. Note that the SXML document is a bit more compact than its XML counterpart.
<?xml version='1.0'> <di contract="728g"> <wt refnum="345"> <delivery> <date month="6" day="01" year="2001"/> <weight>783</weight> </delivery> <vehicle type="lorry" number="A567TP99"/> </wt> <wt refnum="459"> <vehicle type="car" number="25676043"/> </wt> </di>
(*TOP* (*PI* xml "version='1.0'") (di (@ (contract "728g")) (wt (@ (refnum "345")) (delivery (date (@ (month "6") (day "1") (year "2001"))) (weight "783")) (vehicle (@ (type "lorry") (number "A567TP99")))) (wt (@ (refnum "459")) (vehicle (@ (type "car") (number "25676043"))))))
SXML can also be considered an abstract syntax tree for a parsed XML document. An XML document or a well-formed part of it can automatically be converted into the corresponding SXML form via a functional Scheme XML parsing framework SSAX.
It is worth noting that SXML represents all the information contained in XML documents, including comments, namespaces and external entities. These are omitted in this section for the sake of simplicity, but they are considered in the SXML specification.
Take the following simple XHTML page:
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en"> <head> <title>An example page</title> </head> <body> <h1 id="greeting">Hi, there!</h1> <p>This is just an >>example<< to show XHTML & SXML.</p> </body> </html>
After translating it to SXML, the same page now looks like this:
(*TOP* (@ (*NAMESPACES* (x "http://www.w3.org/1999/xhtml"))) (x:html (@ (xml:lang "en") (lang "en")) (x:head (x:title "An example page")) (x:body (x:h1 (@ (id "greeting")) "Hi, there") (x:p "This is just an >>example<< to show XHTML & SXML."))))
Each element's tag pair is replaced by a set of parentheses. The tag's name is not repeated at the end, it is simply the first symbol in the list. The element's contents follow, which are either elements themselves or strings. There is no special syntax required for XML attributes. In SXML they are simply represented as just another node, which has the special name of @. This can't cause a name clash with an actual "@" tag, because @ is not allowed as a tag name in XML. This is a common pattern in SXML: anytime a tag is used to indicate a special status or something that is not possible in XML, a name is used that does not constitute a valid XML identifier.
We can also see that there's no need to "escape" otherwise meaningful characters like & and > as & and > entities. All string content is automatically escaped because it is considered to be pure content, and has no tags or entities in it. This also means it is much easier to insert autogenerated content and that there is no danger that we might forget to escape user input when we display it to other users (which could lead to all kinds of nasty cross-site scripting attacks or other annoyances).
This section considers some important features of SXML, deductible from SXML grammar and properties of S-expressions.
The cdr of an SXML attribute list forms an association list, so that, when SXML is read into a Lisp program, any SXML attribute can be extracted from an attribute list using Lisp's built-in assoc function.
SXML elements and attributes
The uniformity of the SXML representation for elements, attributes, and processing instructions simplifies queries and transformations. For the SXML data model, attributes and processing instructions look like regular elements with a distinguished name. Therefore, query and transformation functions dedicated to attributes become redundant, because ordinary functions with distinguished names can be used.
The uniform representation for SXML elements and attributes is especially convenient for practical tasks. Differences between elements and attributes in XML are blurred. Choosing either an element or an attribute for representing concrete practical information is often a question of style, and such a choice can later be changed. Such a change in a data structure is expressed in SXML as simply an addition/removal of one hierarchy level, namely an attribute-list. This requires the minimal modification of an SXML application. For the SXML notation, the only difference between an attribute and an element is that the former is contained within the attribute-list (which is a special SXML node) and cannot have nested elements.
For example, if data restructuring requires that the weight of a delivered load, initially represented as a nested element, is to be represented as an attribute instead, the SXML element
(delivery ... (weight "789")))
will be changed to
(delivery (@ (weight "789")) ...)
Such a notation for elements and attributes simplifies SXML data restructuring and allows uniform queries to be used for data processing.
SXML document as a tree of uniform nodes
Since an SXML document is essentially a tree structure, it can be described in a more uniform way by introducing the term of an SXML node for nodes of this tree.
An SXML node can be defined on the basis of SXML grammar as a single production [N] given below. Alternatively, an SXML node can be defined as a set of two mutually recursive datatypes: [N1], [N2] and [N3]. In the latter case, a Node is constructed by adding a name to the Nodelist as its leftmost member; a Nodelist is itself a (sorted) list of Nodes.
[N] <Node> ::= <Element> | <attributes-list> | <attribute> | "character data: text string" | <TOP> | <PI> [N1] <Node> ::= ( <name1> . <Nodelist> ) | "text string" [N2] <Nodelist> ::= ( <Node> <Node>* ) [N3] <name1> ::= <name> | @ | *TOP* | *PI*
Such a consideration emphasizes SXML tree structure and the uniform representation for information items as S-expressions
SXML as a Scheme program
The syntax of LISP family programming languages, in particular, Scheme, is based on S-expressions used for both data and code representation. This makes it possible and convenient for Scheme programs to be treated as a semi-structured data and vice versa.
Since an SXML document and its nodes are S-expressions, they can be used for representing a Scheme program. For making this possible, it is sufficient that the first member of every list contained in the SXML tree is a function; the use of macros offers more possibilities. The rest of the members of the list are then the arguments, which are passed to that function. In accordance with SXML grammar, attribute and element names and special names must be bound to functions.
An SXML document or an SXML node that fulfills these requirements may be considered a Scheme program which can be evaluated, for example, by means of eval function.
For example, if para and bold are defined as functions as follows:
(define (para . x) (cons 'p x)) (define (bold . x) (cons 'b x))
then the following SXML element
(para "plain" (bold "highlighted") "plain")
can be treated as a program, and the result of its evaluation is the SXML element:
(p "plain" (b "highlighted") "plain")
Note that the result of evaluating such a program is not necessarily an SXML element. Namely, a program may return a textual representation for the source data in XML or HTML; or even have a side effect, such as saving the SXML data in a relational database.
As a data model
SXML, like XML, models documents as "ordered hierarchies of content-based objects." This has many strengths, perhaps most importantly separating formatting and other processing of documents, from their representation per se. However, this model may be a less natural fit for other purposes. For example, relational databases differ in not being (inherently) hierarchical, and in not being inherently ordered; either model can simulate the other, but at some cost in naturalness, performance, and/or other properties.
SXML's representation suggests a slight model difference from XML's with regard to processing instructions and comments. In XML, these are reserved node types, whose content is essentially text (no attributes, and no nested elements, comments, or PIs). In the SXML grammar above, processing instructions are indeed a special type, but comments do not appear at all; while in the "uniform nodes" model above, all node-types (again except comments) are treated as equivalent to elements with reserved names (for example, "*PI*").
SXML does not provide schema specifications or validators as does XML; however, insofar as SXML is an alternative representation of the same information structure as XML, that functionality can be obtained by converting to XML and then using existing specifications and tools.
As a syntactic representation, or file format
SXML may be very slightly more compact than XML (almost entirely due to specifying element names only at element starts, and not at ends. However, this has the drawback that it is harder to detect errors such as a misplaced ")" (thus, the entropy of the file is slightly higher). This change can either increase or decrease human-readability or the raw data, largely depending on how dense the markup is in a given case.
Of course SXML can be parsed by a program in any programming language, and then be represented using any desired data structure. Precisely as with XML, implementations vary: XML applications that can process data in a one-pass serial fashion typically use SAX style interfaces that stay very close to the raw input data stream, while applications that must access parts of the data in non-linear random-access fashion use DOM interfaces that mirror the hierarchical structure instead.
It has been claimed that because the underlying structure is based on singly linked lists, nodes have no default access to either the parent node and the siblings nodes, only to their child nodes. But this confuses underlying structure, with a linear representation of a structure. Any disk file is a linear sequence of bytes or characters—but that mundane fact places almost no limits on what structures can be represented.
As a simple example, saying that the following expression's "underlying structure" is either a 21-character string, or a singly-linked list of 11 nodes (4 numbers, 3 arithmetic operators, and 4 grouping delimiters), is at best a gross oversimplification:
( 1 + 2 ) * ( 3 + 4 )
Because SXML is so similar to S-expressions syntactically, it is trivial to load it into a LISP or Scheme program just as if it were a generic S-expression. Doing so is utterly trivial to program in such languages, but would lead to each parenthesized group becoming a singly-linked list: a data structure which is far from optimal for kinds of processing commonly anticipated for XML-like structures. Similarly, in any programming language it is trivial to load an entire SXML document into one long string—but it would be a poor choice for most purposes.
In reality, XML, SXML, SGML, or most any data representation is loaded into data structures that facilitate required operations. DOM and other interfaces provide methods to get from an element to its parent, preceding and following siblings, and numbered children directly, and to access attributes by name. Practical DOM implementations make likely operations very fast.
If a program does not do this, then typical operations such as getting the Nth child of an element, or the preceding element in a long list, or the element with a given ID, remain possible but are far from optimal.
- Tim Berners-Lee. "Reform of SGML." March 1993. http://www.w3.org/MarkUp/SGML/TimComments
- Joe English. "Delimited pseudoelements". Oct 1, 1996. http://lists.w3.org/Archives/Public/w3c-sgml-wg/1996Oct/0021.html
- Ora Lassila. "PICS-NG Metadata Model and Label Syntax W3C NOTE. 1997-05-14 (section 6.2). http://www.w3.org/TR/NOTE-pics-ng-metadata.html
- "CL-XML Provides Common Lisp Support for XML, XPath, and XQuery." The Cover Pages, June 9, 2001. http://xml.coverpages.org/ni2001-06-09-a.html
- Kirill Lisovskiy, Dmitry Lizorkin. "SXML: an XML document as an S-expression" // Russian Digital Libraries Journal. – Moscow: IIS, 2003. – Vol. 6, No 2. – ISSN 1562-5419. - http://modis.ispras.ru/Lizorkin/Publications/sxml-eng.pdf
- Allen Renear, Elli Mylonas, David Durand. "Refining our Notion of What Text Really Is: The Problem of Overlapping Hierarchies." Proceedings of the annual joint meeting of the Association for Computers and the Humanities and the Association for Literary and Linguistic Computing, Christ Church, Oxford University, April 1992. http://www.stg.brown.edu/resources/stg/monographs/ohco.html
- Steven DeRose. "Architecture and Speed of Common XML Operations." In Proceedings of Extreme Markup Languages. Montreal, 2005.
- SXML Tools Tutorial by Dmitry Lizorkin
- Main SSAX/SXML page
- XML Matters: Investigating SXML and SSAX: Manipulating XML in the Scheme programming language by David Mertz, Ph.D. IBM developerWorks article
Detailed introduction, motivation and real-life case-studies of SSAX, SXML, SXPath and SXSLT. The paper and the complementary talk presented at the International Lisp Conference 2002.