4.1.7.5 Syntax of Tokens as Character Strings

SICStus Prolog supports wide characters (up to 31 bits wide), interpreted as a superset of Unicode.

Each character in the code set has to be classified as belonging to one of the character categories, such as small-letter, digit, etc. This classification is called the character-type mapping, and it is used for defining the syntax of tokens.

Only character codes 0..255, i.e. the ISO-8859-1 (Latin 1) subset of Unicode, can be part of unquoted tokens⁵, unless the Prolog flag legacy_char_classification is set; see ref-lps-flg. This restriction may be lifted in the future.

For quoted tokens, i.e. quoted atoms and strings, almost any sequence of code points assigned to non-private abstract characters in Unicode 15 is allowed. The disallowed characters are those in the whitespace-char category except that space (character code 32) is allowed despite it being a whitespace-char.

An additional restriction is that the sequence of characters that makes up a quoted token must be in Normal Form C (NFC) http://www.unicode.org/reports/tr15/. This is currently not enforced. A future release may enforce this restriction or perform this normalization automatically.

NFC is the normalization form used on the web (http://www.w3.org/TR/charmod/) and what most software can be expected to produce by default. Any sequence consisting of only characters from Latin 1 is already in NFC.

When the Prolog flag legacy_char_classification is set, characters in the whitespace-char category are still treated as whitespace but other character codes outside the range 0..255, assigned to non-private abstract characters in Unicode 15, are treated as lower case. Such characters can therefore appear as themselves, without using escape sequences, both in quoted and unquoted tokens.

Note: Any output produced by write_term/2 with the option quoted(true) will be in NFC. This includes output from writeq/[1,2] and write_canonical/[1,2].

whitespace-char

These are character codes 0..32, 127..160, 8206..8207, and 8232..8233. This includes ASCII characters such as TAB, LFD, and SPC, as well as all characters with Unicode property “Pattern_Whitespace” including the Unicode-specific LINE SEPARATOR (8232).

small-letter

These are character codes 97..122, i.e. the letters ‘a’ through ‘z’, as well as the non-ASCII character codes 170, 186, 223..246, and 248..255.

If the Prolog flag legacy_char_classification (see ref-lps-flg) is set, then the small-letter set will also include almost every code point above 255 assigned to non-private abstract characters in Unicode 15.

capital-letter

These are character codes 65..90, i.e. the letters ‘A’ through ‘Z’, as well as the non-ASCII character codes 192..214, and 216..222.

digit

These are character codes 48..57, i.e. the digits ‘0’ through ‘9’.

symbol-char

These are character codes 35, 36, 38, 42, 43, 45..47, 58, 60..64, 92, 94, and 126, i.e. the characters:

+ - * / \ ^ < > = ~ : . ? @ # $ & 

In addition, the non-ASCII character codes 161..169, 171..185, 187..191, 215, and 247 belong to this character type⁶.

solo-char

These are character codes 33 and 59 i.e. the characters ‘!’ and ‘;’.

punctuation-char

These are character codes 37, 40, 41, 44, 91, 93, and 123..125, i.e. the characters:

% ( ) , [ ] { | }

quote-char

These are character codes 34, 39, and 96 i.e. the characters ‘"’, ‘'’, and ‘`’.

underline

This is character code 95 i.e. the character ‘_’.

Other characters are unclassified and may only appear in comments and to some extent, as discussed above, in quoted atoms and strings.