sunet/scheme/lib/url.scm

;;; HTTP 1.1 Request-URI parsing and unparsing -*- Scheme -*-

;;; Copyright (c) 1995 by Olin Shivers.
;;; For copyright information, see the file COPYING which comes with
;;; the distribution.

;;; References:
;;; RFC 2616 Hypertext Transfer Protocol -- HTTP/1.1
;;; RFC 2396 Uniform Resource Identifiers (URI): Generic Syntax
;;;
;;; RFC 2616 adopts definitions of regexps from RFC 2396.


;;; RFC 2616 is ambiguous in defining Request_URIS:
;;;
;;; section 5.1.2 states:
;;; HTTP 1.1 Request-URIS are of the form
;;; Request-URI    = "*" | absoluteURI | abs_path | authority
;;;
;;; whilst section 3.2.2 defines the 'http_URL'
;;; http_URL = "http://" host [ ":" port ] [ abs_path [ "?" query ]]


;;; Since allowing for general absoluteURIs doesn't make too much sense
;;; we implement only Request_URIs as follows:
;;; Request-URI    = ( http_URL | abs_path) ["#" fragment]
;;;
;;; where http_URL is a subset of absoluteURI

;;; [ "#" fragment ] is allowed even though 
;;; RFC 2616 disallowes the #fragment part 
;;; (while RFC 1945 for HTTP/1.0 allowed it).
;;; (This is for compatibility with buggy clients).




;;; RexExps for Request-URIs as scsh SREs
;;; stick to RFC terminology throughout

(define digit (rx numeric))

(define alpha (rx alphabetic))

(define alphanum (rx alphanumeric))

(define hex (rx hex-digit))

(define escaped (rx (: "%" ,hex ,hex)))

(define mark (rx ( "-_.!~*'()")))

(define unreserved (rx (| ,alphanum ,mark)))

(define reserved (rx ( ";/?:@&=+$,")))

(define uric (rx (| ,reserved ,unreserved ,escaped)))

(define fragment (rx (* ,uric)))

(define query (rx (* ,uric)))

(define pchar-charset (rx ( ":@&=+$,")))

(define pchar (rx (| ,unreserved ,escaped ,pchar-charset)))

(define param (rx (* ,pchar)))

(define segment (rx (:
		     (* ,pchar)
		     (* (: ";" ,param)))))

(define path-segments (rx (:
			   ,segment 
			   (* (: "/" ,segment)))))

(define abs_path (rx (: 
		      "/" 
		      ,path-segments)))


(define port (rx (* ,digit)))

(define IPv4address (rx (+ ,digit) "." (+ ,digit) "." (+ ,digit) "." (+ ,digit)))

(define toplabel (rx (:
		      (|
		       ,alpha
		       (: 
		       ,alpha 
		       (* (| ,alphanum "-"))
		       ,alphanum)))))

(define domainlabel (rx (:
			 (|
			  ,alphanum
			  (: ,alphanum 
			     (* (| ,alphanum "-"))
			     ,alphanum)))))

(define hostname (rx (:
		      (* (: ,domainlabel "."))
		      ,toplabel 
		      (? "."))))

(define host (rx (| ,hostname ,IPv4address)))

(define http_URL (rx (:
		      "http://" 
		      (submatch 
		       ,host)
		      (? 
		       (: ":" (submatch ,port)))
		      (? 
		       (: (submatch ,abs_path)
			  (? 
			   (: "?" (submatch ,query))))))))

(define http_URL_with_frag (rx (: bos ,@http_URL (? "#" ,fragment) eos)))

(define abs_path_with_frag (rx (: bos (submatch ,abs_path) (? "#" ,fragment) eos)))

(define Request-URI (rx (| ,@http_URL_with_frag ,@abs_path_with_frag)))



;;; parse a HTTP 1.1 Request_URI
;;; 
;;; return matches of regexps host, port, abs_path, query;
;;;
;;; If request-uri is a relative URI, host, port and query are #f;
;;; port and query are also #f if they are not given in an absolute URI.
;;; If there's no abs_path given, or abs_path is "/", path is the empty list;
;;; otherwise it is a list containing the path's segments.
;;;
;;; don't decode yet!

(define (parse-uri request-uri)
   (cond 

    ((regexp-search abs_path_with_frag request-uri)
     => (lambda (match)
	  (values #f #f (split-abs-path (match:substring match 1)) #f)))

    ((regexp-search http_URL_with_frag request-uri)
     =>(lambda (match)
	 (let ((host (match:substring match 1))
	       (port (match:substring match 2))
	       (path (split-abs-path (match:substring match 3)))
	       (query (match:substring match 4)))
	   (values host port path query))))

    (else
     (fatal-syntax-error "Request-URI syntactically faulty"))))
 
  
;;; split the string abs-path at slashes, return list of 'segments' (see RegExp definition above).
;;;
;;; SPLIT-PATH assumes abs-path if either #f or matches the RegExp abs_path, 
;;; no checks are done.
;;;
;;; remark: abs_path allows for strings containing several consecutive slashes;
;;; SPLIT-ABS-PATH treats them as one slash.
;;; (e.g., "/foo///bar//baz/" => ("foo" "bar" "baz")) 


(define (split-abs-path abs-path)

  (if abs-path 

      (regexp-fold-right 
       (rx (+ (~ ("/"))))
       (lambda (match i res)
	 (cons (match:substring match 0) res))
       '()
       abs-path)

      '()))


;;; record type HTTP-URL for Request_URIs
;;; 
;;; The HOST slot is a non-empty-string or #f.
;;;
;;; The PORT slot is an integer or #f.
;;;
;;; The PATH slot is the Request_URI's path split at slashes
;;; (e.g., "/foo///bar//baz/" => ("foo" "bar" "baz")) 
;;; and decoded. 
;;;
;;; The QUERY slot is a decoded non-empty-string or #f.

(define-record-type http-url :http-url
  (make-http-url host port path query)
  http-url?
  (host http-url-host)
  (port http-url-port)		
  (path http-url-path)		
  (query http-url-query))

;;; decode various parts of the Request_URI as returned by PARSE-URI;
;;; returns a HTTP-URL record.

(define (parsed-uri->http-url host port path query)

  (let 	((portnumber (and port (string->number port)))
	(decoded-path (map unescape path))
	(decoded-query (and query (unescape query))))

  (make-http-url host portnumber decoded-path decoded-query)))

;;really decode query here??

;;; parse a HTTP 1.1. Request_URI into a http-url record

(define (parse-http-url uri-string)
  (call-with-values
   (lambda () (parse-uri uri-string))
   parsed-uri->http-url))

;;; Unparse a http-url record into its corresponding Request_URI

(define (http-url->uri-string http-url)

  (let* ((host (http-url-host http-url))
	(scheme-and-host-string 
	 (if host
	     (string-append "http://" host)
	     ""))

	(port (http-url-port http-url))
	(port-string 
	 (if port
	     (string-append ":" (number->string port))
	     ""))

	(path (http-url-path http-url))
	(path-string 
	 (fold-right
	  (lambda (segment res)
	    (string-append "/" (escape-uri segment) res))
	  ""
	  path))

	(query (http-url-query http-url))
	(query-string (if query
			(string-append "?" (escape-uri query))
			"")))

    (string-append scheme-and-host-string port-string path-string query-string)))

;;TODO: separate procedures for encoding path and query



;;; decode a URI
;;; walk over string s and unescape all occurrences of RegExp 'escaped' (see above).

;;; Remark: 
;;; we assume no non-ASCII characters occur in the URI; therefore the
;;; ascii table is used for conversion of the octet the hexnumber
;;; represents to a char.

;;; Caution:
;;; a URI must be separated into its components (i.e. parsed by
;;; PARSE-URI) before the escaped characters within those components
;;; can be safely decoded. Don't use UNESCAPE on an unparsed URI.

(define (unescape s)
  (regexp-fold 
   escaped
   (lambda (start-search match res)
     (let* ((start-match (match:start match))
	    (hexchar-low (string-ref s (+ start-match 2)))
	    (hexchar-high (string-ref s (+ start-match 1)))
	    (hex-low (hexchar->int hexchar-low))
	    (hex-high (hexchar->int hexchar-high))
	    (ascii (+ (* 16 hex-high) hex-low)))
       (string-append 
	res 
	(substring s start-search start-match)
	(string (ascii->char ascii)))))
   ""
   s
   (lambda (start-search res)
    (string-append res (substring s start-search (string-length s))))))


; make use of the fact that numbers and characters are in order in the ascii table
(define (hexchar->int c)      
  (- (char->ascii c) 
     (if (char-numeric? c)
	 (char->ascii #\0)         
	 (- (if (char-upper-case? c)
		(char->ascii #\A)
		(char->ascii #\a))
	    10))))