Title

Author

Table of contents

• Abstract
• Variable index
• Rationale
  • "Linear-update" operations
  • Extra-SRFI recommendations
• Specification
  • General procedures
  • Iterating over character sets
  • Creating character sets
  • Querying character sets
  • Character set algebra
  • Standard character sets
• Unicode, Latin-1 and ASCII definitions of the standard character sets
• Reference implementation
• Acknowledgements
• References & Links
• Copyright

Abstract

The ability to efficiently represent and manipulate sets of characters is an unglamorous but very useful capability for text-processing code -- one that tends to pop up in the definitions of other libraries. Hence it is useful to specify a general substrate for this functionality early. This SRFI defines a general library that provides this functionality. It is accompanied by a reference implementation for the spec. The reference implementation is fairly efficient, straightforwardly portable, and has a "free software" copyright. The implementation is tuned for "small" 7 or 8 bit character types, such as ASCII or Latin-1; the data structures and algorithms would have to be altered for larger 16 or 32 bit character types such as Unicode -- however, the specs have been carefully designed with these larger character types in mind. Several forthcoming SRFIs can be defined in terms of this one:

• string library
• delimited input procedures (e.g., read-line)
• regular expressions

Variable Index

Here is the complete set of bindings -- procedural and otherwise -- exported by this library. In a Scheme system that has a module or package system, these procedures should be contained in a module named "char-set-lib".

Rationale

The ability to efficiently manipulate sets of characters is quite useful for text-processing code. Encapsulating this functionality in a general, efficiently implemented library can assist all such code. This library defines a new data structure to represent these sets, called a "char-set." The char-set type is distinct from all other types.

This library is designed to be portable across implementations that use different character types and representations, especially ASCII, Latin-1 and Unicode. Some effort has been made to preserve compatibility with Java in the Unicode case (see the definition of char-set:whitespace for the single real deviation).

Linear-update operations

The procedures of this SRFI, by default, are "pure functional" -- they do not alter their parameters. However, this SRFI defines a set of "linear-update" procedures which have a hybrid pure-functional/side-effecting semantics: they are allowed, but not required, to side-effect one of their parameters in order to construct their result. An implementation may legally implement these procedures as pure, side-effect-free functions, or it may implement them using side effects, depending upon the details of what is the most efficient or simple to implement in terms of the underlying representation.

The linear-update routines all have names ending with "!".

Clients of these procedures may not rely upon these procedures working by side effect. For example, this is not guaranteed to work:

(let* ((cs1 (char-set #\a #\b #\c))      ; cs1 = {a,b,c}.
       (cs2 (char-set-adjoin! cs1 #\d))) ; Add d to {a,b,c}.
  cs1) ; Could be either {a,b,c} or {a,b,c,d}.

However, this is well-defined:

(let ((cs (char-set #\a #\b #\c)))
  (char-set-adjoin! cs #\d)) ; Add d to {a,b,c}.

So clients of these procedures write in a functional style, but must additionally be sure that, when the procedure is called, there are no other live pointers to the potentially-modified character set (hence the term "linear update").

There are two benefits to this convention:

• Implementations are free to provide the most efficient possible implementation, either functional or side-effecting.
• Programmers may nonetheless continue to assume that character sets are purely functional data structures: they may be reliably shared without needing to be copied, uniquified, and so forth.

Note that pure functional representations are the right thing for ASCII- or Latin-1-based Scheme implementations, since a char-set can be represented in an ASCII Scheme with 4 32-bit words. Pure set-algebra operations on such a representation are very fast and efficient. Programmers who code using linear-update operations are guaranteed the system will provide the best implementation across multiple platforms.

In practice, these procedures are most useful for efficiently constructing character sets in a side-effecting manner, in some limited local context, before passing the character set outside the local construction scope to be used in a functional manner.

Scheme provides no assistance in checking the linearity of the potentially side-effected parameters passed to these functions --- there's no linear type checker or run-time mechanism for detecting violations. (But sophisticated programming environments, such as DrScheme, might help.)

Extra-SRFI recommendations

Users are cautioned that the R5RS predicates

may or may not be in agreement with the SRFI 14 base character sets

Implementors are strongly encouraged to bring these predicates into agreement with the base character sets of this SRFI; not to do so risks major confusion.

Specification

In the following procedure specifications:

• A cs parameter is a character set.
• An s parameter is a string.
• A char parameter is a character.
• A char-list parameter is a list of characters.
• A pred parameter is a unary character predicate procedure, returning a true/false value when applied to a character.
• An obj parameter may be any value at all.

Passing values to procedures with these parameters that do not satisfy these types is an error.

Unless otherwise noted in the specification of a procedure, procedures always return character sets that are distinct (from the point of view of the linear-update operations) from the parameter character sets. For example, char-set-adjoin is guaranteed to provide a fresh character set, even if it is not given any character parameters.

Parameters given in square brackets are optional. Unless otherwise noted in the text describing the procedure, any prefix of these optional parameters may be supplied, from zero arguments to the full list. When a procedure returns multiple values, this is shown by listing the return values in square brackets, as well. So, for example, the procedure with signature

halts? f [x init-store] -> [boolean integer]

A parameter followed by "..." means zero-or-more elements. So the procedure with the signature

sum-squares x ...  -> number

spell-check doc dict1 dict2 ... -> string-list

General procedures

char-set? obj -> boolean

Is the object obj a character set?

char-set= cs₁ ... -> boolean

Are the character sets equal?

Boundary cases:

(char-set=) => true
(char-set= cs) => true

Rationale: transitive binary relations are generally extended to n-ary relations in Scheme, which enables clearer, more concise code to be written. While the zero-argument and one-argument cases will almost certainly not arise in first-order uses of such relations, they may well arise in higher-order cases or macro-generated code. E.g., consider

(apply char-set= cset-list)

This is well-defined if the list is empty or a singleton list. Hence we extend these relations to any number of arguments. Implementors have reported actual uses of n-ary relations in higher-order cases allowing for fewer than two arguments. The way of Scheme is to handle the general case; we provide the fully general extension.

A counter-argument to this extension is that R5RS's transitive binary arithmetic relations (=, <, etc.) require at least two arguments, hence this decision is a break with the prior convention -- although it is at least one that is backwards-compatible.

char-set<= cs₁ ... -> boolean

Returns true if every character set cs_i is a subset of character set cs_i+1.

Boundary cases:

(char-set<=) => true
(char-set<= cs) => true

Rationale: See char-set= for discussion of zero- and one-argument applications. Consider testing a list of char-sets for monotonicity with

(apply char-set<= cset-list)

char-set-hash cs [bound] -> integer

Compute a hash value for the character set cs. Bound is a non-negative exact integer specifying the range of the hash function. A positive value restricts the return value to the range [0,bound).

If bound is either zero or not given, the implementation may use an implementation-specific default value, chosen to be as large as is efficiently practical. For instance, the default range might be chosen for a given implementation to map all strings into the range of integers that can be represented with a single machine word.

Invariant:

(char-set= cs1 cs2) => (= (char-set-hash cs1 b) (char-set-hash cs2 b))

A legal but nonetheless discouraged implementation:

(define (char-set-hash cs . maybe-bound) 1)

Rationale: allowing the user to specify an explicit bound simplifies user code by removing the mod operation that typically accompanies every hash computation, and also may allow the implementation of the hash function to exploit a reduced range to efficiently compute the hash value. E.g., for small bounds, the hash function may be computed in a fashion such that intermediate values never overflow into bignum integers, allowing the implementor to provide a fixnum-specific "fast path" for computing the common cases very rapidly.

Iterating over character sets

char-set-cursor cset -> cursor

char-set-ref cset cursor -> char

char-set-cursor-next cset cursor -> cursor

end-of-char-set? cursor -> boolean

Cursors are a low-level facility for iterating over the characters in a set. A cursor is a value that indexes a character in a char set. char-set-cursor produces a new cursor for a given char set. The set element indexed by the cursor is fetched with char-set-ref. A cursor index is incremented with char-set-cursor-next; in this way, code can step through every character in a char set. Stepping a cursor "past the end" of a char set produces a cursor that answers true to end-of-char-set?. It is an error to pass such a cursor to char-set-ref or to char-set-cursor-next.

A cursor value may not be used in conjunction with a different character set; if it is passed to char-set-ref or char-set-cursor-next with a character set other than the one used to create it, the results and effects are undefined.

Cursor values are not necessarily distinct from other types. They may be integers, linked lists, records, procedures or other values. This license is granted to allow cursors to be very "lightweight" values suitable for tight iteration, even in fairly simple implementations.

Note that these primitives are necessary to export an iteration facility for char sets to loop macros.

Example:

(define cs (char-set #\G #\a #\T #\e #\c #\h))

;; Collect elts of CS into a list.
(let lp ((cur (char-set-cursor cs)) (ans '()))
  (if (end-of-char-set? cur) ans
      (lp (char-set-cursor-next cs cur)
          (cons (char-set-ref cs cur) ans))))
  => (#\G #\T #\a #\c #\e #\h)

;; Equivalently, using a list unfold (from SRFI 1):
(unfold-right end-of-char-set? 
              (curry char-set-ref cs)
	      (curry char-set-cursor-next cs)
	      (char-set-cursor cs))
  => (#\G #\T #\a #\c #\e #\h)

Rationale: Note that the cursor API's four functions "fit" the functional protocol used by the unfolders provided by the list, string and char-set SRFIs (see the example above). By way of contrast, here is a simpler, two-function API that was rejected for failing this criterion. Besides char-set-cursor, it provided a single function that mapped a cursor and a character set to two values, the indexed character and the next cursor. If the cursor had exhausted the character set, then this function returned false instead of the character value, and another end-of-char-set cursor. In this way, the other three functions of the current API were combined together.

char-set-fold kons knil cs -> object

This is the fundamental iterator for character sets. Applies the function kons across the character set cs using initial state value knil. That is, if cs is the empty set, the procedure returns knil. Otherwise, some element c of cs is chosen; let cs' be the remaining, unchosen characters. The procedure returns

(char-set-fold kons (kons c knil) cs')

Examples:

;; CHAR-SET-MEMBERS
(lambda (cs) (char-set-fold cons '() cs))

;; CHAR-SET-SIZE
(lambda (cs) (char-set-fold (lambda (c i) (+ i 1)) 0 cs))

;; How many vowels in the char set?
(lambda (cs) 
  (char-set-fold (lambda (c i) (if (vowel? c) (+ i 1) i))
                 0 cs))

char-set-unfold  f p g seed [base-cs] -> char-set

char-set-unfold! f p g seed base-cs -> char-set

This is a fundamental constructor for char-sets.

• G is used to generate a series of "seed" values from the initial seed: seed, (g seed), (g² seed), (g³ seed), ...
• P tells us when to stop -- when it returns true when applied to one of these seed values.
• F maps each seed value to a character. These characters are added to the base character set base-cs to form the result; base-cs defaults to the empty set. char-set-unfold! adds the characters to base-cs in a linear-update -- it is allowed, but not required, to side-effect and use base-cs's storage to construct the result.

More precisely, the following definitions hold, ignoring the optional-argument issues:

(define (char-set-unfold p f g seed base-cs) 
  (char-set-unfold! p f g seed (char-set-copy base-cs)))

(define (char-set-unfold! p f g seed base-cs)
  (let lp ((seed seed) (cs base-cs))
        (if (p seed) cs                                 ; P says we are done.
            (lp (g seed)                                ; Loop on (G SEED).
                (char-set-adjoin! cs (f seed))))))      ; Add (F SEED) to set.

(Note that the actual implementation may be more efficient.)

Examples:

                         
(port->char-set p) = (char-set-unfold eof-object? values
                                      (lambda (x) (read-char p))
                                      (read-char p))

(list->char-set lis) = (char-set-unfold null? car cdr lis)

char-set-for-each proc cs -> unspecified

Apply procedure proc to each character in the character set cs. Note that the order in which proc is applied to the characters in the set is not specified, and may even change from one procedure application to another.

Nothing at all is specified about the value returned by this procedure; it is not even required to be consistent from call to call. It is simply required to be a value (or values) that may be passed to a command continuation, e.g. as the value of an expression appearing as a non-terminal subform of a begin expression. Note that in R5RS, this restricts the procedure to returning a single value; non-R5RS systems may not even provide this restriction.

char-set-map proc cs -> char-set

proc is a char->char procedure. Apply it to all the characters in the char-set cs, and collect the results into a new character set.

Essentially lifts proc from a char->char procedure to a char-set -> char-set procedure.

Example:

(char-set-map char-downcase cset)

Creating character sets

char-set-copy cs -> char-set

Returns a copy of the character set cs. "Copy" means that if either the input parameter or the result value of this procedure is passed to one of the linear-update procedures described below, the other character set is guaranteed not to be altered.

A system that provides pure-functional implementations of the linear-operator suite could implement this procedure as the identity function -- so copies are not guaranteed to be distinct by eq?.

char-set char₁ ... -> char-set

Return a character set containing the given characters.

list->char-set  char-list [base-cs] -> char-set

list->char-set! char-list base-cs -> char-set

Return a character set containing the characters in the list of characters char-list.

If character set base-cs is provided, the characters from char-list are added to it. list->char-set! is allowed, but not required, to side-effect and reuse the storage in base-cs; list->char-set produces a fresh character set.

string->char-set  s [base-cs] -> char-set

string->char-set! s base-cs -> char-set

Return a character set containing the characters in the string s.

If character set base-cs is provided, the characters from s are added to it. string->char-set! is allowed, but not required, to side-effect and reuse the storage in base-cs; string->char-set produces a fresh character set.

char-set-filter  pred cs [base-cs] -> char-set

char-set-filter! pred cs base-cs -> char-set

Returns a character set containing every character c in cs such that (pred c) returns true.

If character set base-cs is provided, the characters specified by pred are added to it. char-set-filter! is allowed, but not required, to side-effect and reuse the storage in base-cs; char-set-filter produces a fresh character set.

An implementation may not save away a reference to pred and invoke it after char-set-filter or char-set-filter! returns -- that is, "lazy," on-demand implementations are not allowed, as pred may have external dependencies on mutable data or have other side-effects.

Rationale: This procedure provides a means of converting a character predicate into its equivalent character set; the cs parameter allows the programmer to bound the predicate's domain. Programmers should be aware that filtering a character set such as char-set:full could be a very expensive operation in an implementation that provided an extremely large character type, such as 32-bit Unicode. An earlier draft of this library provided a simple predicate->char-set procedure, which was rejected in favor of char-set-filter for this reason.

ucs-range->char-set  lower upper [error? base-cs] -> char-set

ucs-range->char-set! lower upper error? base-cs -> char-set

Lower and upper are exact non-negative integers; lower <= upper.

Returns a character set containing every character whose ISO/IEC 10646 UCS-4 code lies in the half-open range [lower,upper).

• If the requested range includes unassigned UCS values, these are silently ignored (the current UCS specification has "holes" in the space of assigned codes).
• If the requested range includes "private" or "user space" codes, these are handled in an implementation-specific manner; however, a UCS- or Unicode-based Scheme implementation should pass them through transparently.
• If any code from the requested range specifies a valid, assigned UCS character that has no corresponding representative in the implementation's character type, then (1) an error is raised if error? is true, and (2) the code is ignored if error? is false (the default). This might happen, for example, if the implementation uses ASCII characters, and the requested range includes non-ASCII characters.

If character set base-cs is provided, the characters specified by the range are added to it. ucs-range->char-set! is allowed, but not required, to side-effect and reuse the storage in base-cs; ucs-range->char-set produces a fresh character set.

Note that ASCII codes are a subset of the Latin-1 codes, which are in turn a subset of the 16-bit Unicode codes, which are themselves a subset of the 32-bit UCS-4 codes. We commit to a specific encoding in this routine, regardless of the underlying representation of characters, so that client code using this library will be portable. I.e., a conformant Scheme implementation may use EBCDIC or SHIFT-JIS to encode characters; it must simply map the UCS characters from the given range into the native representation when possible, and report errors when not possible.

->char-set x -> char-set

Coerces x into a char-set. X may be a string, character or char-set. A string is converted to the set of its constituent characters; a character is converted to a singleton set; a char-set is returned as-is. This procedure is intended for use by other procedures that want to provide "user-friendly," wide-spectrum interfaces to their clients.

Querying character sets

char-set-size cs -> integer

Returns the number of elements in character set cs.

char-set-count pred cs -> integer

Apply pred to the chars of character set cs, and return the number of chars that caused the predicate to return true.

char-set->list cs -> character-list

This procedure returns a list of the members of character set cs. The order in which cs's characters appear in the list is not defined, and may be different from one call to another.

char-set->string cs -> string

This procedure returns a string containing the members of character set cs. The order in which cs's characters appear in the string is not defined, and may be different from one call to another.

char-set-contains? cs char -> boolean

This procedure tests char for membership in character set cs.

The MIT Scheme character-set package called this procedure char-set-member?, but the argument order isn't consistent with the name.

char-set-every pred cs -> boolean

char-set-any   pred cs -> boolean

The char-set-every procedure returns true if predicate pred returns true of every character in the character set cs. Likewise, char-set-any applies pred to every character in character set cs, and returns the first true value it finds. If no character produces a true value, it returns false. The order in which these procedures sequence through the elements of cs is not specified.

Note that if you need to determine the actual character on which a predicate returns true, use char-set-any and arrange for the predicate to return the character parameter as its true value, e.g.

(char-set-any (lambda (c) (and (char-upper-case? c) c)) 
              cs)

Character-set algebra

char-set-adjoin cs char₁ ... -> char-set

char-set-delete cs char₁ ... -> char-set

Add/delete the char_i characters to/from character set cs.

char-set-adjoin! cs char₁ ... -> char-set

char-set-delete! cs char₁ ... -> char-set

Linear-update variants. These procedures are allowed, but not required, to side-effect their first parameter.

char-set-complement cs -> char-set

char-set-union cs₁ ... -> char-set

char-set-intersection cs₁ ... -> char-set

char-set-difference cs₁ cs₂ ... -> char-set

char-set-xor cs₁ ... -> char-set

char-set-diff+intersection cs₁ cs₂ ... -> [char-set char-set]

These procedures implement set complement, union, intersection, difference, and exclusive-or for character sets. The union, intersection and xor operations are n-ary. The difference function is also n-ary, associates to the left (that is, it computes the difference between its first argument and the union of all the other arguments), and requires at least one argument.

Boundary cases:

(char-set-union) => char-set:empty
(char-set-intersection) => char-set:full
(char-set-xor) => char-set:empty
(char-set-difference cs) => cs

char-set-diff+intersection returns both the difference and the intersection of the arguments -- it partitions its first parameter. It is equivalent to

(values (char-set-difference cs1 cs2 ...)
        (char-set-intersection cs1 (char-set-union cs2 ...)))

but can be implemented more efficiently.

Programmers should be aware that char-set-complement could potentially be a very expensive operation in Scheme implementations that provide a very large character type, such as 32-bit Unicode. If this is a possibility, sets can be complimented with respect to a smaller universe using char-set-difference.

char-set-complement! cs -> char-set

char-set-union! cs₁ cs₂ ... -> char-set

char-set-intersection! cs₁ cs₂ ... -> char-set

char-set-difference! cs₁ cs₂ ... -> char-set

char-set-xor! cs₁ cs₂ ... -> char-set

char-set-diff+intersection! cs₁ cs₂ cs₃ ... -> [char-set char-set]

These are linear-update variants of the set-algebra functions. They are allowed, but not required, to side-effect their first (required) parameter.

char-set-diff+intersection! is allowed to side-effect both of its two required parameters, cs₁ and cs₂.

Standard character sets

Several character sets are predefined for convenience:

Note that there may be characters in char-set:letter that are neither upper or lower case---this might occur in implementations that use a character type richer than ASCII, such as Unicode. A "graphic character" is one that would put ink on your page. While the exact composition of these sets may vary depending upon the character type provided by the underlying Scheme system, here are the definitions for some of the sets in an ASCII implementation:

Note that the existence of the char-set:ascii set implies that the underlying character set is required to be at least as rich as ASCII (including ASCII's control characters).

Rationale: The name choices reflect a shift from the older "alphabetic/numeric" terms found in R5RS and Posix to newer, Unicode-influenced "letter/digit" lexemes.

Unicode, Latin-1 and ASCII definitions of the standard character sets

In Unicode Scheme implementations, the base character sets are compatible with Java's Unicode specifications. For ASCII or Latin-1, we simply restrict the Unicode set specifications to their first 128 or 256 codes, respectively. Scheme implementations that are not based on ASCII, Latin-1 or Unicode should attempt to preserve the sense or spirit of these definitions.

The following descriptions frequently make reference to the "Unicode character database." This is a file, available at URL

Each line contains a description of a Unicode character. The first semicolon-delimited field of the line gives the hex value of the character's code; the second field gives the name of the character, and the third field gives a two-letter category. Other fields give simple 1-1 case-mappings for the character and other information; see

for further description of the file's format. Note in particular the two-letter category specified in the the third field, which is referenced frequently in the descriptions below.

char-set:lower-case

For Unicode, we follow Java's specification: a character is lowercase if

• it is not in the range [U+2000,U+2FFF], and
• the Unicode attribute table does not give a lowercase mapping for it, and
• at least one of the following is true:
  • the Unicode attribute table gives a mapping to uppercase for the character, or
  • the name for the character in the Unicode attribute table contains the words "SMALL LETTER" or "SMALL LIGATURE".

The lower-case ASCII characters are

Latin-1 adds another 33 lower-case characters to the ASCII set:

Note that three of these have no corresponding Latin-1 upper-case character:

(The compatibility micro character uppercases to the non-Latin-1 Greek capital mu; the German sharp s character uppercases to the pair of characters "SS," and the capital y-with-diaeresis is non-Latin-1.)

(Note that the Java spec for lowercase characters given at

is inconsistent. U+00B5 MICRO SIGN fulfills the requirements for a lower-case character (as of Unicode 3.0), but is not given in the numeric list of lower-case character codes.)

(Note that the Java spec for isLowerCase() given at

gives three mutually inconsistent definitions of "lower case." The first is the definition used in this SRFI. Following text says "A character is considered to be lowercase if and only if it is specified to be lowercase by the Unicode 2.0 standard (category Ll in the Unicode specification data file)." The former spec excludes U+00AA FEMININE ORDINAL INDICATOR and U+00BA MASCULINE ORDINAL INDICATOR; the later spec includes them. Finally, the spec enumerates a list of characters in the Latin-1 subset; this list excludes U+00B5 MICRO SIGN, which is included in both of the previous specs.)

char-set:upper-case

For Unicode, we follow Java's specification: a character is uppercase if

• it is not in the range [U+2000,U+2FFF], and
• the Unicode attribute table does not give an uppercase mapping for it (this excludes titlecase characters), and
• at least one of the following is true:
  • the Unicode attribute table gives a mapping to lowercase for the character, or
  • the name for the character in the Unicode attribute table contains the words "CAPITAL LETTER" or "CAPITAL LIGATURE".

The upper-case ASCII characters are

Latin-1 adds another 30 upper-case characters to the ASCII set:

char-set:title-case

In Unicode, a character is titlecase if it has the category Lt in the character attribute database. There are very few of these characters; here is the entire 31-character list as of Unicode 3.0:

There are no ASCII or Latin-1 titlecase characters.

char-set:letter

In Unicode, a letter is any character with one of the letter categories (Lu, Ll, Lt, Lm, Lo) in the Unicode character database.

There are 52 ASCII letters

There are 117 Latin-1 letters. These are the 115 characters that are members of the Latin-1 char-set:lower-case and char-set:upper-case sets, plus

(These two letters are considered lower-case by Unicode, but not by Java or SRFI 14.)

char-set:digit

In Unicode, a character is a digit if it has the category Nd in the character attribute database. In Latin-1 and ASCII, the only such characters are 0123456789. In Unicode, there are other digit characters in other code blocks, such as Gujarati digits and Tibetan digits.

char-set:hex-digit

The only hex digits are 0123456789abcdefABCDEF.

char-set:letter+digit

The union of char-set:letter and char-set:digit.

char-set:graphic

A graphic character is one that would put ink on paper. The ASCII and Latin-1 graphic characters are the members of

char-set:printing

A printing character is one that would occupy space when printed, i.e., a graphic character or a space character. char-set:printing is the union of char-set:whitespace and char-set:graphic.

char-set:whitespace

In Unicode, a whitespace character is either

• a character with one of the space, line, or paragraph separator categories (Zs, Zl or Zp) of the Unicode character database.
• U+0009 Horizontal tabulation (\t control-I)
• U+000A Line feed (\n control-J)
• U+000B Vertical tabulation (\v control-K)
• U+000C Form feed (\f control-L)
• U+000D Carriage return (\r control-M)

There are 24 whitespace characters in Unicode 3.0:

The ASCII whitespace characters are the first six characters in the above list -- line feed, horizontal tabulation, vertical tabulation, form feed, carriage return, and space. These are also exactly the characters recognised by the Posix isspace() procedure. Latin-1 adds the no-break space.

Note: Java's isWhitespace() method is incompatible, including

and excluding

Java's excluding the no-break space means that tokenizers can simply break character streams at "whitespace" boundaries. However, the exclusion introduces exceptions in other places, e.g. char-set:printing is no longer simply the union of char-set:graphic and char-set:whitespace.

char-set:iso-control

The ISO control characters are the Unicode/Latin-1 characters in the ranges [U+0000,U+001F] and [U+007F,U+009F].

ASCII restricts this set to the characters in the range [U+0000,U+001F] plus the character U+007F.

Note that Unicode defines other control characters which do not belong to this set (hence the qualifying prefix "iso-" in the name). This restriction is compatible with the Java IsISOControl() method.

char-set:punctuation

In Unicode, a punctuation character is any character that has one of the punctuation categories in the Unicode character database (Pc, Pd, Ps, Pe, Pi, Pf, or Po.)

ASCII has 23 punctuation characters:

!"#%&'()*,-./:;?@[\]_{}

Latin-1 adds six more:

Note that the nine ASCII characters $+<=>^`|~ are not punctuation. They are "symbols."

char-set:symbol

In Unicode, a symbol is any character that has one of the symbol categories in the Unicode character database (Sm, Sc, Sk, or So). There are nine ASCII symbol characters:

$+<=>^`|~

Latin-1 adds 18 more:

char-set:blank

Blank chars are horizontal whitespace. In Unicode, a blank character is either

• a character with the space separator category (Zs) in the Unicode character database.
• U+0009 Horizontal tabulation (\t control-I)

There are eighteen blank characters in Unicode 3.0:

The ASCII blank characters are the first two characters above -- horizontal tab and space. Latin-1 adds the no-break space.

Java doesn't have the concept of "blank" characters, so there are no compatibility issues.

Reference implementation

This SRFI comes with a reference implementation. It resides at:

I have placed this source on the Net with an unencumbered, "open" copyright. Some of the code in the reference implementation bears a distant family relation to the MIT Scheme implementation, and being derived from that code, is covered by the MIT Scheme copyright (which is a generic BSD-style open-source copyright -- see the source file for details). The remainder of the code was written by myself for scsh or for this SRFI; I have placed this code under the scsh copyright, which is also a generic BSD-style open-source copyright.

The code is written for portability and should be simple to port to any Scheme. It has only the following deviations from R4RS, clearly discussed in the comments:

• an error procedure;
• the R5RS values procedure for producing multiple return values;
• a simple check-arg procedure for argument checking;
• let-optionals* and :optional macros for for parsing, checking and defaulting optional arguments from rest lists;
• The SRFI-19 define-record-type form;
• bitwise-and for the hash function;
• %latin1->char and %char->latin1.

The library is written for clarity and well-commented; the current source is about 375 lines of source code and 375 lines of comments and white space. It is also written for efficiency. Fast paths are provided for common cases.

This is not to say that the implementation can't be tuned up for a specific Scheme implementation. There are notes in comments addressing ways implementors can tune the reference implementation for performance.

In short, I've written the reference implementation to make it as painless as possible for an implementor -- or a regular programmer -- to adopt this library and get good results with it.

The code uses a rather simple-minded, inefficient representation for ASCII/Latin-1 char-sets -- a 256-character string. The character whose code is i is in the set if s[i] = ASCII 1 (soh, or ^a); not in the set if s[i] = ASCII 0 (nul). A much faster and denser representation would be 16 or 32 bytes worth of bit string. A portable implementation using bit sets awaits standards for bitwise logical-ops and byte vectors.

"Large" character types, such as Unicode, should use a sparse representation, taking care that the Latin-1 subset continues to be represented with a dense 32-byte bit set.

Acknowledgements

The design of this library benefited greatly from the feedback provided during the SRFI discussion phase. Among those contributing thoughtful commentary and suggestions, both on the mailing list and by private discussion, were Paolo Amoroso, Lars Arvestad, Alan Bawden, Jim Bender, Dan Bornstein, Per Bothner, Will Clinger, Brian Denheyer, Kent Dybvig, Sergei Egorov, Marc Feeley, Matthias Felleisen, Will Fitzgerald, Matthew Flatt, Arthur A. Gleckler, Ben Goetter, Sven Hartrumpf, Erik Hilsdale, Shiro Kawai, Richard Kelsey, Oleg Kiselyov, Bengt Kleberg, Donovan Kolbly, Bruce Korb, Shriram Krishnamurthi, Bruce Lewis, Tom Lord, Brad Lucier, Dave Mason, David Rush, Klaus Schilling, Jonathan Sobel, Mike Sperber, Mikael Staldal, Vladimir Tsyshevsky, Donald Welsh, and Mike Wilson. I am grateful to them for their assistance.

I am also grateful the authors, implementors and documentors of all the systems mentioned in the introduction. Aubrey Jaffer should be noted for his work in producing Web-accessible versions of the R5RS spec, which was a tremendous aid.

This is not to imply that these individuals necessarily endorse the final results, of course.

During this document's long development period, great patience was exhibited by Mike Sperber, who is the editor for the SRFI, and by Hillary Sullivan, who is not.

References & links

[Java]

The following URLs provide documentation on relevant Java classes.
http://java.sun.com/products/jdk/1.2/docs/api/java/lang/Character.html
http://java.sun.com/products/jdk/1.2/docs/api/java/lang/String.html
http://java.sun.com/products/jdk/1.2/docs/api/java/lang/StringBuffer.html
http://java.sun.com/products/jdk/1.2/docs/api/java/text/Collator.html
http://java.sun.com/products/jdk/1.2/docs/api/java/text/package-summary.html

[MIT-Scheme]

http://www.swiss.ai.mit.edu/projects/scheme/

[R5RS]

Revised⁵ report on the algorithmic language Scheme.
R. Kelsey, W. Clinger, J. Rees (editors).
Higher-Order and Symbolic Computation, Vol. 11, No. 1, September, 1998.
and ACM SIGPLAN Notices, Vol. 33, No. 9, October, 1998.
Available at http://www.schemers.org/Documents/Standards/.

[SRFI]

The SRFI web site.
http://srfi.schemers.org/

[SRFI-14]

SRFI-14: String libraries.
http://srfi.schemers.org/srfi-14/

This document, in HTML:

http://srfi.schemers.org/srfi-14/srfi-14.html

This document, in plain text format:

http://srfi.schemers.org/srfi-14/srfi-14.txt

Source code for the reference implementation:

http://srfi.schemers.org/srfi-14/srfi-14.scm

Scheme 48 module specification, with typings:

http://srfi.schemers.org/srfi-14/srfi-14-s48-module.scm

Regression-test suite:

http://srfi.schemers.org/srfi-14/srfi-14-tests.scm

[Unicode]

http://www.unicode.org/

[UnicodeData]

The Unicode character database.
ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.txt
ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.html

Copyright

Certain portions of this document -- the specific, marked segments of text describing the R5RS procedures -- were adapted with permission from the R5RS report.

All other text is copyright (C) Olin Shivers (1998, 1999, 2000). All Rights Reserved.

This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Scheme Request For Implementation process or editors, except as needed for the purpose of developing SRFIs in which case the procedures for copyrights defined in the SRFI process must be followed, or as required to translate it into languages other than English.

The limited permissions granted above are perpetual and will not be revoked by the authors or their successors or assigns.

This document and the information contained herein is provided on an "as is" basis and the authors and the SRFI editors disclaim all warranties, express or implied, including but not limited to any warranty that the use of the information herein will not infringe any rights or any implied warranties of merchantability or fitness for a particular purpose.