Reasonably complete and up-to-date docs.

This commit is contained in:
sperber 2003-01-14 15:02:44 +00:00
parent 8ed2a48176
commit ff8061c4ea
1 changed files with 138 additions and 190 deletions

View File

@ -1,218 +1,166 @@
\chapter{Handle URIs}\label{cha:uri}
%
\begin{description}
\item[Used files:] uri.scm
\item[Name of the package:] uri
\end{description}
%
\chapter{Parsing and Processing URIs}\label{cha:uri}
The \ex{uri} structure contains a library for dealing with URIs.
\section{Notes on URI Syntax}
\section{Overview}
A URI (Uniform Resource Identifier) is of following syntax:
%
\begin{inset}
[scheme] : \semvar{path} [{\normalfont?\/} search] [{\normalfont\#} fragmentid]
[scheme] : \var{path} [{\normalfont?\/} search] [{\normalfont\#} fragid]
\end{inset}
Parts in brackets may be ommitted. The last part is usually referred
to as fragid in this document.
%
Parts in brackets may be ommitted.
As you see, the URI contains characters like \verb|:| to indicate its
different parts. But what, if the \semvar{scheme} contains \verb|:| as
part of its name? For this purpose, some special characters are
\emph{escaped} if they are a regular part of a name and not indicators
for the structure of a URI. Escape-sequences are of following scheme:
\verb|\%hh| where \verb|h| is a hexadecimal digit. The hexadecimal
number refers to the (US) ASCII code of the escaped character, e.g.\
\ex{\%20} is space (ASCII character 32) and \ex{\%61} is `a' (ASCII
character 97). This module provides procedures to escape and unescape
strings that are meant to be used in a URI.
The URI contains characters like \verb|:| to indicate its different
parts. Some special characters are \emph{escaped} if they are a
regular part of a name and not indicators for the structure of a URI.
Escape sequences are of following scheme: \verb|\%hh| where \verb|h|
is a hexadecimal digit. The hexadecimal number refers to the
ASCII of the escaped character, e.g.\ \ex{\%20} is space (ASCII
32) and \ex{\%61} is `a' (ASCII 97). This module
provides procedures to escape and unescape strings that are meant to
be used in a URI.
\section{Procedures}
\begin{defundesc}{parse-uri} {uri-string } {scheme path search
\defun{parse-uri} {uri-string } {scheme path search
frag-id} \label{proc:parse-uri}
Parses an \semvar{uri\=string} in the possible four fields, as
mentioned above in \emph{Overview}. These four fields are returned
as a multiple value. They are \emph{not} unescaped, as the rules for
parsing the \semvar{path} component in particular need unescaped
text, and are dependent on \semvar{scheme}. The URL parser is
responsible for doing this. If the \semvar{scheme}, \semvar{search}\
or \semvar{fragid} portions are not specified, they are \sharpf.
Otherwise, \semvar{scheme}, \semvar{search}, and \semvar{fragid} are
strings. \semvar{path} is a non-empty string list -- the path split
\begin{desc}
Parses an \var{uri\=string} into its four fields.
The fields are \emph{not} unescaped, as the rules for
parsing the \var{path} component in particular need unescaped
text, and are dependent on \var{scheme}. The URL parser is
responsible for doing this. If the \var{scheme}, \var{search}
or \var{fragid} portions are not specified, they are \sharpf.
Otherwise, \var{scheme}, \var{search}, and \var{fragid} are
strings. \var{path} is a non-empty string list----the path split
at slashes.
For those of you who are interested, here is a description of the
parsing technique. It is inwards from both ends.
\begin{itemize}
\item First we search forwards for the first reserved character
(\verb|=|, \verb|;|, \verb|/|, \verb|#|, \verb|?|, \verb|:| or
\verb|space|). If it's a colon, then that's the \semvar{scheme}
part, otherwise we have no \semvar{scheme} part. At all events we
remove it.
\item Then we search backwards from the end for the last reserved
char. If it's a sharp, then that's the \semvar{fragid} part --
remove it.
\item Then we search backwards from the end for the last reserved
char. If it's a question-mark, then that's the \semvar{search}
part -- remove it.
\item What's left is the path. We split at slashes. The empty string
becomes a list containing the empty string.
\end{itemize}
This scheme is tolerant of the various ways people build broken
URI's out there on the Net\footnote{So it is not absolutely conform
with RFC~1630}, e.g. \verb|=| is a reserved character, but used
unescaped in the search-part. It was given to me\footnote{That's
Olin Shivers.} by Dan Connolly of the W3C and slightly modified.
\end{defundesc}
\end{desc}
\begin{defundesc}{unescape-uri} {string [start] [end]} {string}
Unescapes a string. This procedure should only be used \emph{after}
the URL (!) was parsed, since unescaping may introduce characters
that blow up the parse (that's why escape sequences are used in URIs
;-). Escape sequences are of the scheme as described in ``Overview''.
\end{defundesc}
Here is a description of the parsing technique. It is inwards from
both ends:
\begin{itemize}
\item First, the code searches forwards for the first reserved
character (\verb|=|, \verb|;|, \verb|/|, \verb|#|, \verb|?|,
\verb|:| or \verb|space|). If it's a colon, then that's the
\var{scheme} part, otherwise there is no \var{scheme} part. At
all events, it is removed.
\item Then the code searches backwards from the end for the last reserved
char. If it's a sharp, then that's the \var{fragid} part---remove it.
\item Then the code searches backwards from the end for the last reserved
char. If it's a question-mark, then that's the \var{search}
part----remove it.
\item What's left is the path. The code split it at slashes. The
empty string becomes a list containing the empty string.
\end{itemize}
%
This scheme is tolerant of the various ways people build broken
URI's out there on the Net\footnote{So it does not absolutely conform
to RFC~1630.}, e.g.\ \verb|=| is a reserved character, but used
unescaped in the search-part. It was given to me\footnote{That's
Olin Shivers.} by Dan Connolly of the W3C and slightly modified.
\defun{unescape-uri}{string [start] [end]}{string}
\begin{desc}
\ex{Unescape-uri} unescapes a string. If \var{start} and/or \var{end} are
specified, they specify start and end positions within \var{string}
should be unescaped.
\end{desc}
%
This procedure should only be used \emph{after} the URI was parsed,
since unescaping may introduce characters that blow up the
parse---that's why escape sequences are used in URIs.
%\texttt{uri-escaped-chars} \hfill
%\texttt{char-set}\index{\texttt{uri-escaped-chars}}
\defvar{uri-escaped-chars}{char-set}
\begin{desc}
A set of characters that are escaped in URIs. These are the
following characters: dollar (\verb|$|), minus (\verb|-|),%fool Xemacs$
underscore (\verb|_|), at (\verb|@|), dot (\verb|.|), and-sign
(\verb|&|), exclamation mark (\verb|!|), asterisk (\verb|*|),
backslash (\verb|\|), double quote (\verb|"|), single quote
(\verb|'|), open brace (\verb|(|), close brace (\verb|)|), comma
(\verb|,|) plus (\verb|+|) and all other characters that are neither
letters nor digits (such as space and control characters).
This is a set of characters (in the sense of SRFI~14) which are
escaped in URIs. These are the
following characters: \verb|$|, \verb|-|, \verb|_|, \verb|@|, %$
\verb|.|, \verb|&|, \verb|!|, \verb|*|, \verb|\|, \verb|"|,
\verb|'|, \verb|(|, \verb|)|, \verb|,|, \verb|+|, and all other
characters that are neither letters nor digits (such as space and
control characters).
\end{desc}
\begin{defundesc}{escape-uri} {string [escaped-chars]} {string}
Escapes characters of \semvar{string} that are given with
\semvar{escaped\=chars}. \semvar{escaped\=chars} default to
\ex{uri\=escaped\=chars}. Be careful with using this procedure to
chunks of text with syntactically meaningful reserved characters
(e.g., paths with URI slashes or colons) -- they'll be escaped, and
lose their special meaning. E.g.\ it would be a mistake to apply
\ex{escape-uri} to
``\ex{//lcs.\ob{}mit.\ob{}edu:8001\ob/foo\ob/bar.html}'' because the
sla\-shes and co\-lons would be escaped. Note that \ex{esacpe-uri}
doesn't check this as it would lose his meaning.
\end{defundesc}
\begin{defundesc}{resolve-uri} {cscheme cp scheme p} {scheme path}
%FIXME{Sorry, I can't figure out what resolve-uri is inteded to do.
%Perhaps I find it out later.}
%There is a paragraph in the spec, that describes someting like
%resolve-uri does. We have to check this.
To be done.
\end{defundesc}
\begin{defundesc}{split-uri-path} {uri start end} {list}
Splits uri at slashes. Only the substring given with \semvar{start}
(inclusive) and \semvar{end} (exclusive) as indices is considered.
\semvar{start} and $\semvar{end} - 1$ have to be within the range of
\semvar{uri}. Otherwise an index-out-of-range exception will be
raised. Example: \codex{(split-uri-path "foo/bar/colon" 4 11)}
results to \codex{'("bar" "col")}
\end{defundesc}
\begin{defundesc}{uri-path-list->path} {plist} {string}
Generates a path out of an uri-path-list by inserting slashes
between the elements of \semvar{plist}. If you want to use the
resulting string for further operation, you should escape the
elements of \semvar{plist} in case the contain slashes. This doesn't
escape them for you, you must do that yourself like
\ex{(uri-path-list->path (map escape-uri pathlist))}.
\end{defundesc}
\begin{defundesc}{simplify-uri-path} {path} {list}
Removes `\ex{.}' and `\ex{..}' entries from path. The result is
a (maybe empty) list representing a path that does not contain any
`\ex{.}' or `\ex{..}'\,. The list can only be empty if the path
did not start with a slash (for the rare occasion someone wants to
simplify a relative path). The result is \sharpf{} if the path tries
to back up past root, for example by `\ex{/..}' or
`\ex{/foo\ob/..\ob/..}' or just `\ex{..}'\,. `\ex{//}' may occur
somewhere in the path referring to root but not being backed up.
Examples:
%FIXME: Can't we have a better environment for examples like these?
\begin{alltt}
(simplify-uri-path
(split-uri-path "/foo/bar/baz/.." 0 15))
\end{alltt}
results to
\codex{'("" "foo" "bar")}
\begin{alltt}
(simplify-uri-path
(split-uri-path "foo/bar/baz/../../.." 0 20))
\end{alltt}
results to
\codex{'()}
\begin{alltt}
(simplify-uri-path
(split-uri-path "/foo/../.." 0 10))
\end{alltt}
results to
\codex{\sharpf ; tried to back up root}
\begin{alltt}
(simplify-uri-path
(split-uri-path "foo/bar//" 0 9))
\end{alltt}
results to
\codex{'("") ; "//" refers to root}
\begin{alltt}
(simplify-uri-path
(split-uri-path "foo/bar/" 0 8))
\end{alltt}
results to
\codex{'("") ; last "/" also refers to root}
\begin{alltt}
(simplify-uri-path
(split-uri-path "/foo/bar//baz/../.." 0 19))
\end{alltt}
results to
\codex{\sharpf ; tries to back up root}
\end{defundesc}
\section{Unexported names}
\defvar{uri-reserved}{char-set}
\defun{escape-uri} {string [escaped-chars]} {string}
\begin{desc}
A list of reserved characters (semicolon, slash, hash, question
mark, double colon and space).
This procedure escapes characters of \var{string} that are in
\var{escaped\=chars}. \var{Escaped\=chars} defaults to
\ex{uri\=escaped\=chars}.
\end{desc}
%
Be careful with using this procedure to chunks of text with
syntactically meaningful reserved characters (e.g., paths with URI
slashes or colons)---they'll be escaped, and lose their special
meaning. E.g.\ it would be a mistake to apply \ex{escape-uri} to
\begin{verbatim}
//lcs.mit.edu:8001/foo/bar.html}
\end{verbatim}
%
because the sla\-shes and co\-lons would be escaped.
\defun{split-uri}{uri start end} {list}
\begin{desc}
This procedure splits \var{uri} at slashes. Only the substring given
with \var{start} (inclusive) and \var{end} (exclusive) as indices is
considered. \var{start} and $\var{end} - 1$ have to be within the
range of \var{uri}. Otherwise an \ex{index-out-of-range} exception
will be raised.
Example: \codex{(split-uri "foo/bar/colon" 4 11)} returns
\codex{("bar" "col")}
\end{desc}
\begin{defundesc}{hex-digit?} {character} {boolean}
Returns \sharpt{} if character is a hexadecimal digit (i.e., one of 1--9,
a--f, A--F), \sharpf{} otherwise.
\end{defundesc}
\defun{uri-path->uri}{plist}{string}
\begin{desc}
This procedure generates a path out of a URI path list by inserting
slashes between the elements of \var{plist}.
\end{desc}
%
If you want to use the resulting string for further operation, you
should escape the elements of \var{plist} in case they contain
slashes, like so:
%
\begin{verbatim}
(uri-path->uri (map escape-uri pathlist))
\end{verbatim}
\defun{simplify-uri-path}{path}{list}
\begin{desc}
This procedure simplifies a URI path. It removes \verb|"."| and
\verb|"/.."| entries from path, and removes parts before a root.
The result is a list, or \sharpf{} if the path tries to back up past
root.
\end{desc}
%
According to RFC~2396, relative paths are considered not to start with
\verb|/|. They are appended to a base URL path and then simplified.
So before you start to simplify a URL try to find out if it is a
relative path (i.e. it does not start with a \verb|/|).
\begin{defundesc}{hexchar->int} {character} {number}
Translates the given character to an integer, e.g. \ex{(hexchar->int
\#a)} results to 10.
\end{defundesc}
Examples:
%
\begin{alltt}
(simplify-uri-path (split-uri "/foo/bar/baz/.." 0 15))
\(\Rightarrow\) ("" "foo" "bar")
\begin{defundesc}{int->hexchar} {integer} {character}
Translates the given integer from range 1--15 into an hexadecimal
character (uses uppercase letters), e.g. \ex{(int->hexchar 14)}
results to `E'.
\end{defundesc}
(simplify-uri-path (split-uri "foo/bar/baz/../../.." 0 20))
\(\Rightarrow\) ()
\begin{defundesc}{rev-append} {list-a list-b} {list}
Performs a \ex{(append (reverse list-a) list-b)}. The comment says it
should be defined in a list package but I am wondering how often
this will be used.
\end{defundesc}
(simplify-uri-path (split-uri "/foo/../.." 0 10))
\(\Rightarrow\) #f
(simplify-uri-path (split-uri "foo/bar//" 0 9))
\(\Rightarrow\) ("")
(simplify-uri-path (split-uri "foo/bar/" 0 8))
\(\Rightarrow\) ("")
(simplify-uri-path (split-uri "/foo/bar//baz/../.." 0 19))
\(\Rightarrow\) #f
\end{alltt}
%EOF
%%% Local Variables:
%%% mode: latex