12.9.1 Readtables

The dispatch table in Delimiters and Dispatch corresponds to the default readtable. By creating a new readtable and installing it via the current-readtable parameter, the reader’s behavior can be extended.

A readtable is consulted at specific times by the reader:

• when looking for the start of a datum;

• when determining how to parse a datum that starts with #;

• when looking for a delimiter to terminate a symbol or number;

• when looking for an opener (such as (), closer (such as )), or . after the first character parsed as a sequence for a pair, list, vector, or hash table; or

• when looking for an opener after #‹n› in a vector of specified length ‹n›.

The readtable is ignored at other times. In particular, after parsing a character that is mapped to the default behavior of ;, the readtable is ignored until the comment’s terminating newline is discovered. Similarly, the readtable does not affect string parsing until a closing double-quote is found. Meanwhile, if a character is mapped to the default behavior of (, then it starts sequence that is closed by any character that is mapped to a close parenthesis ). An apparent exception is that the default parsing of | quotes a symbol until a matching character is found, but the parser is simply using the character that started the quote; it does not consult the readtable.

For many contexts, #f identifies the default readtable. In particular, #f is the initial value for the current-readtable parameter, which causes the reader to behave as described in The Reader.

The possible combinations for key, mode, and action are as follows:

• char 'terminating-macro proc — causes char to be parsed as a delimiter, and an unquoted/uncommented char in the input string triggers a call to the reader macro proc; the activity of proc is described further below. Conceptually, characters like ;, (, and ) are mapped to terminating reader macros in the default readtable.

• char 'non-terminating-macro proc — like the 'terminating-macro variant, but char is not treated as a delimiter, so it can be used in the middle of an identifier or number. Conceptually, # is mapped to a non-terminating macro in the default readtable.

• char 'dispatch-macro proc — like the 'non-terminating-macro variant, but for char only when it follows a # (or, more precisely, when the character follows one that has been mapped to the behavior of #hash in the default readtable).

• char like-char readtable — causes char to be parsed in the same way that like-char is parsed in readtable, where readtable can be #f to indicate the default readtable. Mapping a character to the same actions as | in the default reader means that the character starts quoting for symbols, and the same character terminates the quote; in contrast, mapping a character to the same action as a " means that the character starts a string, but the string is still terminated with a closing ". Finally, mapping a character to an action in the default readtable means that the character’s behavior is sensitive to parameters that affect the original character; for example, mapping a character to the same action as a curly brace { in the default readtable means that the character is disallowed when the read-curly-brace-as-paren parameter is set to #f.

• #f 'non-terminating-macro proc — replaces the macro used to parse characters with no specific mapping: i.e., characters (other than # or |) that can start a symbol or number with the default readtable.

If multiple 'dispatch-macro mappings are provided for a single char, all but the last one are ignored. Similarly, if multiple non-'dispatch-macro mappings are provided for a single char, all but the last one are ignored.

A reader macro proc must accept six arguments, and it can optionally accept two arguments. The first two arguments are always the character that triggered the reader macro and the input port for reading. When the reader macro is triggered by read-syntax (or read-syntax/recursive), the procedure is passed four additional arguments that represent a source location. When the reader macro is triggered by read (or read/recursive), the procedure is passed only two arguments if it accepts two arguments, otherwise it is passed six arguments where the last four are all #f. See Reader-Extension Procedures for information on the procedure’s results.

A reader macro normally reads characters from the given input port to produce a value to be used as the “reader macro-expansion” of the consumed characters. The reader macro might produce a special-comment value (see Special Comments) to cause the consumed character to be treated as whitespace, and it might use read/recursive or read-syntax/recursive.

• either a character (mapping is to same behavior as the character in the default readtable), 'terminating-macro, or 'non-terminating-macro; this result reports the main (i.e., non-'dispatch-macro) mapping for key. When the result is a character, then key is mapped to the same behavior as the returned character in the default readtable.

• either #f or a reader-macro procedure; the result is a procedure when the first result is 'terminating-macro or 'non-terminating-macro.

• either #f or a reader-macro procedure; the result is a procedure when the character has a 'dispatch-macro mapping in readtable to override the default dispatch behavior.

Note that reader-macro procedures for the default readtable are not directly accessible. To invoke default behaviors, use read/recursive or read-syntax/recursive with a character and the #f readtable.

Examples:

; Provides raise-read-error and raise-read-eof-error

> (require syntax/readerr)

> (define (skip-whitespace port)       ; Skips whitespace characters, sensitive to the current       ; readtable's definition of whitespace       (let ([ch (peek-char port)])         (unless (eof-object? ch)           ; Consult current readtable:           (let-values ([(like-ch/sym proc dispatch-proc)                         (readtable-mapping (current-readtable) ch)])             ; If like-ch/sym is whitespace, then ch is whitespace             (when (and (char? like-ch/sym)                        (char-whitespace? like-ch/sym))               (read-char port)               (skip-whitespace port))))))

> (define (skip-comments read-one port src)       ; Recursive read, but skip comments and detect EOF       (let loop ()         (let ([v (read-one)])           (cond            [(special-comment? v) (loop)]            [(eof-object? v)             (let-values ([(l c p) (port-next-location port)])               (raise-read-eof-error                "unexpected EOF in tuple" src l c p 1))]            [else v]))))

> (define (parse port read-one src)       ; First, check for empty tuple       (skip-whitespace port)       (if (eq? #\> (peek-char port))           null           (let ([elem (read-one)])             (if (special-comment? elem)                 ; Found a comment, so look for > again                 (parse port read-one src)                 ; Non-empty tuple:                 (cons elem                       (parse-nonempty port read-one src))))))

> (define (parse-nonempty port read-one src)       ; Need a comma or closer       (skip-whitespace port)       (case (peek-char port)         [(#\>) (read-char port)          ; Done          null]         [(#\,) (read-char port)          ; Read next element and recur          (cons (skip-comments read-one port src)                (parse-nonempty port read-one src))]         [else          ; Either a comment or an error; grab location (in case          ; of error) and read recursively to detect comments          (let-values ([(l c p) (port-next-location port)]                       [(v) (read-one)])            (cond             [(special-comment? v)              ; It was a comment, so try again              (parse-nonempty port read-one src)]             [else              ; Wasn't a comment, comma, or closer; error              ((if (eof-object? v)                   raise-read-eof-error                   raise-read-error)               "expected `,' or `>'" src l c p 1)]))]))

> (define (make-delims-table)       ; Table to use for recursive reads to disallow delimiters       ;  (except those in sub-expressions)       (letrec ([misplaced-delimiter                 (case-lambda                  [(ch port) (unexpected-delimiter ch port #f #f #f #f)]                  [(ch port src line col pos)                   (raise-read-error                    (format "misplaced `~a' in tuple" ch)                    src line col pos 1)])])         (make-readtable (current-readtable)                         #\, 'terminating-macro misplaced-delimiter                         #\> 'terminating-macro misplaced-delimiter)))

> (define (wrap l)       `(make-tuple (list ,@l)))

> (define parse-open-tuple       (case-lambda        [(ch port)         ; `read' mode         (wrap (parse port                      (lambda ()                        (read/recursive port #f                                        (make-delims-table)))                      (object-name port)))]        [(ch port src line col pos)         ; `read-syntax' mode         (datum->syntax          #f          (wrap (parse port                       (lambda ()                         (read-syntax/recursive src port #f                                                (make-delims-table)))                       src))          (let-values ([(l c p) (port-next-location port)])            (list src line col pos (and pos (- p pos)))))]))

> (define tuple-readtable       (make-readtable #f #\< 'terminating-macro parse-open-tuple))

> (parameterize ([current-readtable tuple-readtable])       (read (open-input-string "<1 , 2 , \"a\">")))

'(make-tuple (list 1 2 "a"))

> (parameterize ([current-readtable tuple-readtable])       (read (open-input-string              "< #||# 1 #||# , #||# 2 #||# , #||# \"a\" #||# >")))

'(make-tuple (list 1 2 "a"))

> (define tuple-readtable+       (make-readtable tuple-readtable                       #\* 'terminating-macro (lambda a                                                (make-special-comment #f))                       #\_ #\space #f))

> (parameterize ([current-readtable tuple-readtable+])       (read (open-input-string "< * 1 __,__  2 __,__ * \"a\" * >")))

'(make-tuple (list 1 2 "a"))