URI - Uniform Resource Identifiers (absolute and relative)

 $u1 = URI->new("http://www.perl.com");
 $u2 = URI->new("foo", "http");
 $u3 = $u2->abs($u1);
 $u4 = $u3->clone;
 $u5 = URI->new("HTTP://WWW.perl.com:80")->canonical;

 $str = $u->as_string;
 $str = "$u";

 $scheme = $u->scheme;
 $opaque = $u->opaque;
 $path   = $u->path;
 $frag   = $u->fragment;

 $u->scheme("ftp");
 $u->host("ftp.perl.com");
 $u->path("cpan/");

This module implements the CWURI class. Objects of this class represent Uniform Resource Identifier references as specified in RFC 2396 (and updated by RFC 2732).

A Uniform Resource Identifier is a compact string of characters that identifies an abstract or physical resource. A Uniform Resource Identifier can be further classified as either a Uniform Resource Locator (URL) or a Uniform Resource Name (URN). The distinction between URL and URN does not matter to the CWURI class interface. A URI-reference is a URI that may have additional information attached in the form of a fragment identifier.

An absolute URI reference consists of three parts: a scheme, a scheme-specific part and a fragment identifier. A subset of URI references share a common syntax for hierarchical namespaces. For these, the scheme-specific part is further broken down into authority, path and query components. These URIs can also take the form of relative URI references, where the scheme (and usually also the authority) component is missing, but implied by the context of the URI reference. The three forms of URI reference syntax are summarized as follows:

  <scheme>:<scheme-specific-part>#<fragment>
  <scheme>://<authority><path>?<query>#<fragment>
  <path>?<query>#<fragment>

The components into which a URI reference can be divided depend on the scheme. The CWURI class provides methods to get and set the individual components. The methods available for a specific CWURI object depend on the scheme.

The following methods construct new CWURI objects: Constructs a new URI object. The string representation of a URI is given as argument, together with an optional scheme specification. Common URI wrappers like "" and <>, as well as leading and trailing white space, are automatically removed from the CW$str argument before it is processed further. The constructor determines the scheme, maps this to an appropriate URI subclass, constructs a new object of that class and returns it. The CW$scheme argument is only used when CW$str is a relative URI. It can be either a simple string that denotes the scheme, a string containing an absolute URI reference, or an absolute CWURI object. If no CW$scheme is specified for a relative URI CW$str, then CW$str is simply treated as a generic URI (no scheme-specific methods available). The set of characters available for building URI references is restricted (see URI::Escape). Characters outside this set are automatically escaped by the URI constructor. Constructs a new absolute URI object. The CW$str argument can denote a relative or absolute URI. If relative, then it is absolutized using CW$base_uri as base. The CW$base_uri must be an absolute URI. Constructs a new file URI from a file name. See URI::file. Constructs a new absolute file URI from a file name. See URI::file.

$uri = URI::file->cwd

Returns the current working directory as a file URI. See URI::file.

$uri->clone

Returns a copy of the CW$uri.

The methods described in this section are available for all CWURI objects.

Methods that give access to components of a URI always return the old value of the component. The value returned is CWundef if the component was not present. There is generally a difference between a component that is empty (represented as CW"") and a component that is missing (represented as CWundef). If an accessor method is given an argument, it updates the corresponding component in addition to returning the old value of the component. Passing an undefined argument removes the component (if possible). The description of each accessor method indicates whether the component is passed as an escaped or an unescaped string. A component that can be further divided into sub-parts are usually passed escaped, as unescaping might change its semantics.

The common methods available for all URI are:

$uri->scheme

Sets and returns the scheme part of the CW$uri. If the CW$uri is relative, then CW$uri->scheme returns CWundef. If called with an argument, it updates the scheme of CW$uri, possibly changing the class of CW$uri, and returns the old scheme value. The method croaks if the new scheme name is illegal; a scheme name must begin with a letter and must consist of only US-ASCII letters, numbers, and a few special marks: ., +, -. This restriction effectively means that the scheme must be passed unescaped. Passing an undefined argument to the scheme method makes the URI relative (if possible). Letter case does not matter for scheme names. The string returned by CW$uri->scheme is always lowercase. If you want the scheme just as it was written in the URI in its original case, you can use the CW$uri->_scheme method instead.

$uri->opaque

Sets and returns the scheme-specific part of the CW$uri (everything between the scheme and the fragment) as an escaped string.

$uri->path

Sets and returns the same value as CW$uri->opaque unless the URI supports the generic syntax for hierarchical namespaces. In that case the generic method is overridden to set and return the part of the URI between the host name and the fragment.

$uri->fragment

Returns the fragment identifier of a URI reference as an escaped string.

$uri->as_string

Returns a URI object to a plain string. URI objects are also converted to plain strings automatically by overloading. This means that CW$uri objects can be used as plain strings in most Perl constructs.

$uri->canonical

Returns a normalized version of the URI. The rules for normalization are scheme-dependent. They usually involve lowercasing the scheme and Internet host name components, removing the explicit port specification if it matches the default port, uppercasing all escape sequences, and unescaping octets that can be better represented as plain characters. For efficiency reasons, if the CW$uri is already in normalized form, then a reference to it is returned instead of a copy. Tests whether two URI references are equal. URI references that normalize to the same string are considered equal. The method can also be used as a plain function which can also test two string arguments. If you need to test whether two CWURI object references denote the same object, use the '==' operator. Returns an absolute URI reference. If CW$uri is already absolute, then a reference to it is simply returned. If the CW$uri is relative, then a new absolute URI is constructed by combining the CW$uri and the CW$base_uri, and returned. Returns a relative URI reference if it is possible to make one that denotes the same resource relative to CW$base_uri. If not, then CW$uri is simply returned.

The following methods are available to schemes that use the common/generic syntax for hierarchical namespaces. The descriptions of schemes below indicate which these are. Unknown schemes are assumed to support the generic syntax, and therefore the following methods:

$uri->authority

Sets and returns the escaped authority component of the CW$uri.

$uri->path

Sets and returns the escaped path component of the CW$uri (the part between the host name and the query or fragment). The path can never be undefined, but it can be the empty string.

$uri->path_query

Sets and returns the escaped path and query components as a single entity. The path and the query are separated by a ? character, but the query can itself contain ?.

$uri->path_segments

Sets and returns the path. In a scalar context, it returns the same value as CW$uri->path. In a list context, it returns the unescaped path segments that make up the path. Path segments that have parameters are returned as an anonymous array. The first element is the unescaped path segment proper; subsequent elements are escaped parameter strings. Such an anonymous array uses overloading so it can be treated as a string too, but this string does not include the parameters. Note that absolute paths have the empty string as their first path_segment, i.e. the path CW/foo/bar have 3 path_segments; ", foo and bar".

$uri->query

Sets and returns the escaped query component of the CW$uri.

$uri->query_form

$uri->query_form( \@key_value_pairs )

$uri->query_form( \%hash )

Sets and returns query components that use the application/x-www-form-urlencoded format. Key/value pairs are separated by &, and the key is separated from the value by a = character. The form can be set either by passing separate key/value pairs, or via an array or hash reference. Passing an empty array or an empty hash removes the query component, whereas passing no arguments at all leaves the component unchanged. The order of keys is undefined if a hash reference is passed. The old value is always returned as a list of separate key/value pairs. Assigning this list to a hash is unwise as the keys returned might repeat. The values passed when setting the form can be plain strings or references to arrays of strings. Passing an array of values has the same effect as passing the key repeatedly with one value at a time. All the following statements have the same effect:

    $uri->query_form(foo => 1, foo => 2);
    $uri->query_form(foo => [1, 2]);
    $uri->query_form([ foo => 1, foo => 2 ]);
    $uri->query_form([ foo => [1, 2] ]);
    $uri->query_form({ foo => [1, 2] });

The CWURI::QueryParam module can be loaded to add further methods to manipulate the form of a URI. See URI::QueryParam for details.

$uri->query_keywords

$uri->query_keywords( \@keywords )

Sets and returns query components that use the keywords separated by + format. The keywords can be set either by passing separate keywords directly or by passing a reference to an array of keywords. Passing an empty array removes the query component, whereas passing no arguments at all leaves the component unchanged. The old value is always returned as a list of separate words.

For schemes where the authority component denotes an Internet host, the following methods are available in addition to the generic methods.

$uri->userinfo

Sets and returns the escaped userinfo part of the authority component. For some schemes this is a user name and a password separated by a colon. This practice is not recommended. Embedding passwords in clear text (such as URI) has proven to be a security risk in almost every case where it has been used.

$uri->host

Sets and returns the unescaped hostname. If the CW$new_host string ends with a colon and a number, then this number also sets the port.

$uri->port

Sets and returns the port. The port is a simple integer that should be greater than 0. If a port is not specified explicitly in the URI, then the URI scheme's default port is returned. If you don't want the default port substituted, then you can use the CW$uri->_port method instead.

$uri->host_port

Sets and returns the host and port as a single unit. The returned value includes a port, even if it matches the default port. The host part and the port part are separated by a colon: :.

$uri->default_port

Returns the default port of the URI scheme to which CW$uri belongs. For http this is the number 80, for ftp this is the number 21, etc. The default port for a scheme can not be changed.

Scheme-specific support is provided for the following URI schemes. For CWURI objects that do not belong to one of these, you can only use the common and generic methods.

data:

The data URI scheme is specified in RFC 2397. It allows inclusion of small data items as immediate data, as if it had been included externally. CWURI objects belonging to the data scheme support the common methods and two new methods to access their scheme-specific components: CW$uri->media_type and CW$uri->data. See URI::data for details.

file:

An old specification of the file URI scheme is found in RFC 1738. A new RFC 2396 based specification in not available yet, but file URI references are in common use. CWURI objects belonging to the file scheme support the common and generic methods. In addition, they provide two methods for mapping file URIs back to local file names; CW$uri->file and CW$uri->dir. See URI::file for details.

ftp:

An old specification of the ftp URI scheme is found in RFC 1738. A new RFC 2396 based specification in not available yet, but ftp URI references are in common use. CWURI objects belonging to the ftp scheme support the common, generic and server methods. In addition, they provide two methods for accessing the userinfo sub-components: CW$uri->user and CW$uri->password.

gopher:

The gopher URI scheme is specified in <draft-murali-url-gopher-1996-12-04> and will hopefully be available as a RFC 2396 based specification. CWURI objects belonging to the gopher scheme support the common, generic and server methods. In addition, they support some methods for accessing gopher-specific path components: CW$uri->gopher_type, CW$uri->selector, CW$uri->search, CW$uri->string.

http:

The http URI scheme is specified in RFC 2616. The scheme is used to reference resources hosted by HTTP servers. CWURI objects belonging to the http scheme support the common, generic and server methods.

https:

The https URI scheme is a Netscape invention which is commonly implemented. The scheme is used to reference HTTP servers through SSL connections. Its syntax is the same as http, but the default port is different.

ldap:

The ldap URI scheme is specified in RFC 2255. LDAP is the Lightweight Directory Access Protocol. An ldap URI describes an LDAP search operation to perform to retrieve information from an LDAP directory. CWURI objects belonging to the ldap scheme support the common, generic and server methods as well as ldap-specific methods: CW$uri->dn, CW$uri->attributes, CW$uri->scope, CW$uri->filter, CW$uri->extensions. See URI::ldap for details.

ldapi:

Like the ldap URI scheme, but uses a UNIX domain socket. The server methods are not supported, and the local socket path is available as CW$uri->un_path. The ldapi scheme is used by the OpenLDAP package. There is no real specification for it, but it is mentioned in various OpenLDAP manual pages.

ldaps:

Like the ldap URI scheme, but uses an SSL connection. This scheme is deprecated, as the preferred way is to use the start_tls mechanism.

mailto:

The mailto URI scheme is specified in RFC 2368. The scheme was originally used to designate the Internet mailing address of an individual or service. It has (in RFC 2368) been extended to allow setting of other mail header fields and the message body. CWURI objects belonging to the mailto scheme support the common methods and the generic query methods. In addition, they support the following mailto-specific methods: CW$uri->to, CW$uri->headers.

mms:

The mms URL specification can be found at <http://sdp.ppona.com/> CWURI objects belonging to the mms scheme support the common, generic, and server methods, with the exception of userinfo and query-related sub-components.

news:

The news, nntp and snews URI schemes are specified in <draft-gilman-news-url-01> and will hopefully be available as an RFC 2396 based specification soon. CWURI objects belonging to the news scheme support the common, generic and server methods. In addition, they provide some methods to access the path: CW$uri->group and CW$uri->message.

nntp:

See news scheme.

pop:

The pop URI scheme is specified in RFC 2384. The scheme is used to reference a POP3 mailbox. CWURI objects belonging to the pop scheme support the common, generic and server methods. In addition, they provide two methods to access the userinfo components: CW$uri->user and CW$uri->auth

rlogin:

An old specification of the rlogin URI scheme is found in RFC 1738. CWURI objects belonging to the rlogin scheme support the common, generic and server methods.

rtsp:

The rtsp URL specification can be found in section 3.2 of RFC 2326. CWURI objects belonging to the rtsp scheme support the common, generic, and server methods, with the exception of userinfo and query-related sub-components.

rtspu:

The rtspu URI scheme is used to talk to RTSP servers over UDP instead of TCP. The syntax is the same as rtsp.

rsync:

Information about rsync is available from http://rsync.samba.org. CWURI objects belonging to the rsync scheme support the common, generic and server methods. In addition, they provide methods to access the userinfo sub-components: CW$uri->user and CW$uri->password.

sip:

The sip URI specification is described in sections 19.1 and 25 of RFC 3261. CWURI objects belonging to the sip scheme support the common, generic, and server methods with the exception of path related sub-components. In addition, they provide two methods to get and set sip parameters: CW$uri->params_form and CW$uri->params.

sips:

See sip scheme. Its syntax is the same as sip, but the default port is different.

snews:

See news scheme. Its syntax is the same as news, but the default port is different.

telnet:

An old specification of the telnet URI scheme is found in RFC 1738. CWURI objects belonging to the telnet scheme support the common, generic and server methods.

tn3270:

These URIs are used like telnet URIs but for connections to IBM mainframes. CWURI objects belonging to the tn3270 scheme support the common, generic and server methods.

ssh:

Information about ssh is available at http://www.openssh.com/. CWURI objects belonging to the ssh scheme support the common, generic and server methods. In addition, they provide methods to access the userinfo sub-components: CW$uri->user and CW$uri->password.

urn:

The syntax of Uniform Resource Names is specified in RFC 2141. CWURI objects belonging to the urn scheme provide the common methods, and also the methods CW$uri->nid and CW$uri->nss, which return the Namespace Identifier and the Namespace-Specific String respectively. The Namespace Identifier basically works like the Scheme identifier of URIs, and further divides the URN namespace. Namespace Identifier assignments are maintained at <http://www.iana.org/assignments/urn-namespaces>. Letter case is not significant for the Namespace Identifier. It is always returned in lower case by the CW$uri->nid method. The CW$uri->_nid method can be used if you want it in its original case.

urn:isbn:

The CWurn:isbn: namespace contains International Standard Book Numbers (ISBNs) and is described in RFC 3187. A CWURI object belonging to this namespace has the following extra methods (if the Business::ISBN module is available): CW$uri->isbn, CW$uri->isbn_publisher_code, CW$uri->isbn_country_code, CW$uri->isbn_as_ean.

urn:oid:

The CWurn:oid: namespace contains Object Identifiers (OIDs) and is described in RFC 3061. An object identifier consists of sequences of digits separated by dots. A CWURI object belonging to this namespace has an additional method called CW$uri->oid that can be used to get/set the oid value. In a list context, oid numbers are returned as separate elements.

The following configuration variables influence how the class and its methods behave:

$URI::ABS_ALLOW_RELATIVE_SCHEME

Some older parsers used to allow the scheme name to be present in the relative URL if it was the same as the base URL scheme. RFC 2396 says that this should be avoided, but you can enable this old behaviour by setting the CW$URI::ABS_ALLOW_RELATIVE_SCHEME variable to a TRUE value. The difference is demonstrated by the following examples:

  URI->new("http:foo")->abs("http://host/a/b")
      ==>  "http:foo"

  local $URI::ABS_ALLOW_RELATIVE_SCHEME = 1;
  URI->new("http:foo")->abs("http://host/a/b")
      ==>  "http:/host/a/foo"

$URI::ABS_REMOTE_LEADING_DOTS

You can also have the abs() method ignore excess .. segments in the relative URI by setting CW$URI::ABS_REMOTE_LEADING_DOTS to a TRUE value. The difference is demonstrated by the following examples:

  URI->new("../../../foo")->abs("http://host/a/b")
      ==> "http://host/../../foo"

  local $URI::ABS_REMOTE_LEADING_DOTS = 1;
  URI->new("../../../foo")->abs("http://host/a/b")
      ==> "http://host/foo"

Using regexp variables like CW$1 directly as arguments to the URI methods does not work too well with current perl implementations. I would argue that this is actually a bug in perl. The workaround is to quote them. Example:

   /(...)/ || die;
   $u->query("$1");

As an alternative to this module, the following (official) regular expression can be used to decode a URI:

  my($scheme, $authority, $path, $query, $fragment) =
  $uri =~ m|(?:([^:/?#]+):)?(?://([^/?#]*))?([^?#]*)(?:\?([^#]*))?(?:#(.*))?|;

The CWURI::Split module provides the function uri_split() as a readable alternative.

URI::file, URI::WithBase, URI::QueryParam, URI::Escape, URI::Split, URI::Heuristic

RFC 2396: Uniform Resource Identifiers (URI): Generic Syntax, Berners-Lee, Fielding, Masinter, August 1998.

http://www.iana.org/assignments/uri-schemes

http://www.iana.org/assignments/urn-namespaces

http://www.w3.org/Addressing/

Copyright 1995-2003 Gisle Aas.

Copyright 1995 Martijn Koster.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

This module is based on the CWURI::URL module, which in turn was (distantly) based on the CWwwwurl.pl code in the libwww-perl for perl4 developed by Roy Fielding, as part of the Arcadia project at the University of California, Irvine, with contributions from Brooks Cutter.

CWURI::URL was developed by Gisle Aas, Tim Bunce, Roy Fielding and Martijn Koster with input from other people on the libwww-perl mailing list.

CWURI and related subclasses was developed by Gisle Aas.