Numeric character reference

A numeric character reference (NCR) is a common markup construct in SGML and other SGML-based markup languages such as HTML and XML, and is also a source of much confusion among web document authors.

Example

In SGML, HTML, and XML, the following are numeric character references for the Greek capital letter Sigma ("Σ"):

&#931;
&#0931;
&#x3A3;
&#x3a3;

Discussion

Markup languages are typically defined in terms of ISO 10646 or Unicode characters. That is, a document consists, at its most fundamental level of abstraction, of a sequence of characters, which are abstract units that exist independently of any encoding.

Ideally, when the characters of a document utilizing a markup language are encoded for storage or transmission over a network as a sequence of bits, the encoding that is used will be one that supports representing each and every character in the document, if not in the whole of Unicode, directly as a particular bit sequence.

Sometimes, though, for reasons of convenience or due to technical limitations, documents are encoded with an encoding that cannot represent some characters directly. For example, the widely used encodings based on ISO 8859 can only represent, at most, 256 unique characters as one 8-bit byte each.

Documents are rarely, in practice, ever allowed to use more than one encoding internally, so the onus is usually on the markup language to provide a means for document authors to express unencodable characters in terms of encodable ones. This is generally done through some kind of "escaping" mechanism.

The SGML-based markup languages allow document authors to use special sequences of characters from the ASCII range (the first 127 code points of Unicode) to represent, or reference, any Unicode characters, regardless of whether the characters being represented are supported in the document's encoding. These special sequences are character references.

Character references that are based on the referenced character's Unicode "code point" are called numeric character references. In HTML 4 and in all versions of XHTML and XML, the code point can be expressed either as a decimal (base 10) number or as a hexadecimal (base 16) number. The syntax is as follows:

Character U+0026 (ampersand), followed by character U+0023 (number sign), followed by one of the following choices:

one or more decimal digits zero (U+0030) through nine (U+0039); or
character U+0078 ("x") followed by one or more hexadecimal digits, which are zero (U+0030) through nine (U+0039), Latin capital letter A (U+0041) through F (U+0046), and Latin small letter a (U+0061) through f (U+0066);

all followed by character U+003A (semicolon).

The characters that comprise a numeric character reference can be represented in every character encoding used in computing and telecommunications today, so there is no risk of the reference itself being unencodable.

There is another kind of character reference called a character entity reference, which is something slightly different than a numeric character reference.

Restrictions

ISO 10646 (the Universal Character Set) is the "document character set" of SGML, HTML 4, so by default, any character in such a document, and any character referenced in such a document, must be in the UCS.

While the syntax of SGML does not prohibit unassigned code points such as  from being referenced, SGML-derived markup languages such as HTML and XML can, and often do, restrict numeric character references to only code points that have been assigned to characters. Restrictions may also apply for other reasons, e.g. in HTML 4, , which is a reference to a non-printing "Form Feed" control character, is allowed, but in XML, it is not, while in XML, another control character, , is allowed, but is not allowed in HTML.