5 use warnings::illegalproto ();
7 use Web::Dispatch::Wrapper ();
9 our $VERSION = '0.020';
12 my ($class, $app_package) = @_;
13 $app_package ||= caller;
14 $class->_export_into($app_package);
15 eval "package $app_package; use Web::Dispatch::Wrapper; use Moo; 1"
16 or die "Failed to setup app package: $@";
18 warnings::illegalproto->unimport;
22 my ($class, $app_package) = @_;
25 *{"${app_package}::PSGI_ENV"} = sub () { -1 };
26 require Web::Simple::Application;
27 unshift(@{"${app_package}::ISA"}, 'Web::Simple::Application');
29 (my $name = $app_package) =~ s/::/\//g;
30 $INC{"${name}.pm"} = 'Set by "use Web::Simple;" invocation';
37 Web::Simple - A quick and easy way to build simple web applications
47 sub dispatch_request {
49 [ 200, [ 'Content-type', 'text/plain' ], [ 'Hello world!' ] ]
52 [ 405, [ 'Content-type', 'text/plain' ], [ 'Method not allowed' ] ]
56 HelloWorld->run_if_script;
58 If you save this file into your cgi-bin as C<hello-world.cgi> and then visit:
60 http://my.server.name/cgi-bin/hello-world.cgi/
62 you'll get the "Hello world!" string output to your browser. At the same time
63 this file will also act as a class module, so you can save it as HelloWorld.pm
64 and use it as-is in test scripts or other deployment mechanisms.
66 Note that you should retain the ->run_if_script even if your app is a
67 module, since this additionally makes it valid as a .psgi file, which can
68 be extremely useful during development.
70 For more complex examples and non-CGI deployment, see
71 L<Web::Simple::Deployment>. To get help with L<Web::Simple>, please connect to
72 the irc.perl.org IRC network and join #web-simple.
76 The philosophy of L<Web::Simple> is to keep to an absolute bare minimum for
77 everything. It is not designed to be used for large scale applications;
78 the L<Catalyst> web framework already works very nicely for that and is
79 a far more mature, well supported piece of software.
81 However, if you have an application that only does a couple of things, and
82 want to not have to think about complexities of deployment, then L<Web::Simple>
83 might be just the thing for you.
85 The only public interface the L<Web::Simple> module itself provides is an
88 use Web::Simple 'NameOfApplication';
90 This sets up your package (in this case "NameOfApplication" is your package)
91 so that it inherits from L<Web::Simple::Application> and imports L<strictures>,
92 as well as installs a C<PSGI_ENV> constant for convenience, as well as some
95 Importing L<strictures> will automatically make your code use the C<strict> and
96 C<warnings> pragma, so you can skip the usual:
99 use warnings FATAL => 'aa';
101 provided you 'use Web::Simple' at the top of the file. Note that we turn
102 on *fatal* warnings so if you have any warnings at any point from the file
103 that you did 'use Web::Simple' in, then your application will die. This is,
104 so far, considered a feature.
106 When we inherit from L<Web::Simple::Application> we also use L<Moo>, which is
107 the the equivalent of:
110 package NameOfApplication;
112 extends 'Web::Simple::Application';
115 So you can use L<Moo> features in your application, such as creating attributes
116 using the C<has> subroutine, etc. Please see the documentation for L<Moo> for
119 It also exports the following subroutines for use in dispatchers:
121 response_filter { ... };
123 redispatch_to '/somewhere';
127 $INC{"NameOfApplication.pm"} = 'Set by "use Web::Simple;" invocation';
129 so that perl will not attempt to load the application again even if
131 require NameOfApplication;
133 is encountered in other code.
135 =head1 DISPATCH STRATEGY
137 L<Web::Simple> despite being straightforward to use, has a powerful system
138 for matching all sorts of incoming URLs to one or more subroutines. These
139 subroutines can be simple actions to take for a given URL, or something
140 more complicated, including entire L<Plack> applications, L<Plack::Middleware>
141 and nested subdispatchers.
145 sub dispatch_request {
146 # matches: GET /user/1.htm?show_details=1
148 sub (GET + /user/* + ?show_details~ + .htm|.html|.xhtml) {
149 my ($self, $user_id, $show_details) = @_;
152 # matches: POST /user?username=frew
153 # POST /user?username=mst&first_name=matt&last_name=trout
154 sub (POST + /user + ?username=&*) {
155 my ($self, $username, $misc_params) = @_;
158 # matches: DELETE /user/1/friend/2
159 sub (DELETE + /user/*/friend/*) {
160 my ($self, $user_id, $friend_id) = @_;
163 # matches: PUT /user/1?first_name=Matt&last_name=Trout
164 sub (PUT + /user/* + ?first_name~&last_name~) {
165 my ($self, $user_id, $first_name, $last_name) = @_;
170 # matches: PUT /user/1/role/1
171 sub (PUT + /role/*) {
175 # matches: DELETE /user/1/role/1
176 sub (DELETE + /role/*) {
183 =head2 The dispatch cycle
185 At the beginning of a request, your app's dispatch_request method is called
186 with the PSGI $env as an argument. You can handle the request entirely in
187 here and return a PSGI response arrayref if you want:
189 sub dispatch_request {
190 my ($self, $env) = @_;
191 [ 404, [ 'Content-type' => 'text/plain' ], [ 'Amnesia == fail' ] ]
194 However, generally, instead of that, you return a set of dispatch subs:
196 sub dispatch_request {
198 sub (/) { redispatch_to '/index.html' },
199 sub (/user/*) { $self->show_user($_[1]) },
203 Well, a sub is a valid PSGI response too (for ultimate streaming and async
204 cleverness). If you want to return a PSGI sub you have to wrap it into an
207 sub dispatch_request {
210 # This is pure PSGI here, so read perldoc PSGI
214 If you return a subroutine with a prototype, the prototype is treated
215 as a match specification - and if the test is passed, the body of the
216 sub is called as a method and passed any matched arguments (see below for more details).
218 You can also return a plain subroutine which will be called with just C<$env>
219 - remember that in this case if you need C<$self> you B<must> close over it.
221 If you return a normal object, L<Web::Simple> will simply return it upwards on
222 the assumption that a response_filter (or some arbitrary L<Plack::Middleware>)
223 somewhere will convert it to something useful. This allows:
225 sub dispatch_request {
227 sub (.html) { response_filter { $self->render_zoom($_[0]) } },
228 sub (/user/*) { $self->users->get($_[1]) },
231 to render a user object to HTML, if there is an incoming URL such as:
233 http://myweb.org/user/111.html
235 This works because as we descend down the dispachers, we first match
236 C<sub (.html)>, which adds a C<response_filter> (basically a specialized routine
237 that follows the L<Plack::Middleware> specification), and then later we also
238 match C<sub (/user/*)> which gets a user and returns that as the response.
239 This user object 'bubbles up' through all the wrapping middleware until it hits
240 the C<response_filter> we defined, after which the return is converted to a
243 However, two types of object are treated specially - a C<Plack::Component> object
244 will have its C<to_app> method called and be used as a dispatcher:
246 sub dispatch_request {
248 sub (/static/...) { Plack::App::File->new(...) },
252 A L<Plack::Middleware> object will be used as a filter for the rest of the
253 dispatch being returned into:
255 ## responds to /admin/track_usage AND /admin/delete_accounts
257 sub dispatch_request {
260 Plack::Middleware::Session->new(%opts);
262 sub (/admin/track_usage) {
263 ## something that needs a session
265 sub (/admin/delete_accounts) {
266 ## something else that needs a session
270 Note that this is for the dispatch being B<returned> to, so if you want to
271 provide it inline you need to do:
273 ## ALSO responds to /admin/track_usage AND /admin/delete_accounts
275 sub dispatch_request {
279 Plack::Middleware::Session->new(%opts);
282 ## something that needs a session
284 sub (/delete_accounts) {
285 ## something else that needs a session
290 And that's it - but remember that all this happens recursively - it's
291 dispatchers all the way down. A URL incoming pattern will run all matching
292 dispatchers and then hit all added filters or L<Plack::Middleware>.
294 =head2 Web::Simple match specifications
296 =head3 Method matches
300 A match specification beginning with a capital letter matches HTTP requests
301 with that request method.
307 A match specification beginning with a / is a path match. In the simplest
308 case it matches a specific path. To match a path with a wildcard part, you
312 $self->handle_user($_[1])
314 This will match /user/<anything> where <anything> does not include a literal
315 / character. The matched part becomes part of the match arguments. You can
316 also match more than one part:
319 my ($self, $user_1, $user_2) = @_;
321 sub (/domain/*/user/*) {
322 my ($self, $domain, $user) = @_;
324 and so on. To match an arbitrary number of parts, use C<**>:
327 my ($self, $match) = @_;
329 This will result in a single element for the entire match. Note that you can do
331 sub (/page/**/edit) {
333 to match an arbitrary number of parts up to but not including some final
336 Note: Since Web::Simple handles a concept of file extensions, C<*> and C<**>
337 matchers will not by default match things after a final dot, and this
338 can be modified by using C<*.*> and C<**.*> in the final position, e.g.:
340 /one/* matches /one/two.three and captures "two"
341 /one/*.* matches /one/two.three and captures "two.three"
342 /** matches /one/two.three and captures "one/two"
343 /**.* matches /one/two.three and captures "one/two.three"
349 Will match C</foo/> on the beginning of the path B<and> strip it. This is
350 designed to be used to construct nested dispatch structures, but can also prove
351 useful for having e.g. an optional language specification at the start of a
354 Note that the '...' is a "maybe something here, maybe not" so the above
355 specification will match like this:
358 /foo/ # match and strip path to '/'
359 /foo/bar/baz # match and strip path to '/bar/baz'
365 Will match on C</foo/bar/baz>, but also include C</foo>. Otherwise it
366 operates the same way as C</foo/...>.
368 /foo # match and strip path to ''
369 /foo/ # match and strip path to '/'
370 /foo/bar/baz # match and strip path to '/bar/baz'
372 Please note the difference between C<sub(/foo/...)> and C<sub(/foo...)>. In
373 the first case, this is expecting to find something after C</foo> (and fails to
374 match if nothing is found), while in the second case we can match both C</foo>
375 and C</foo/more/to/come>. The following are roughly the same:
377 sub (/foo) { 'I match /foo' },
379 sub (/bar) { 'I match /foo/bar' },
380 sub (/*) { 'I match /foo/{id}' },
386 sub (~) { 'I match /foo' },
387 sub (/bar) { 'I match /foo/bar' },
388 sub (/*) { 'I match /foo/{id}' },
391 You may prefer the latter example should you wish to take advantage of
392 subdispatchers to scope common activities. For example:
395 my $user_rs = $schema->resultset('User');
396 sub (~) { $user_rs },
397 sub (/*) { $user_rs->find($_[1]) },
400 You should note the special case path match C<sub (~)> which is only meaningful
401 when it is contained in this type of path match. It matches to an empty path.
403 =head4 Naming your patch matches
405 Any C<*>, C<**>, C<*.*>, or C<**.*> match can be followed with C<:name> to make it into a named
408 sub (/*:one/*:two/*:three/*:four) {
409 "I match /1/2/3/4 capturing { one => 1, two => 2, three => 3, four => 4 }"
412 sub (/**.*:allofit) {
413 "I match anything capturing { allofit => \$whole_path }"
416 In the specific case of a simple single-* match, the * may be omitted, to
419 sub (/:one/:two/:three/:four) {
420 "I match /1/2/3/4 capturing { one => 1, two => 2, three => 3, four => 4 }"
423 =head4 C</foo> and C</foo/> are different specs
425 As you may have noticed with the difference between C<sub(/foo/...)> and
426 C<sub(/foo...)>, trailing slashes in path specs are significant. This is
427 intentional and necessary to retain the ability to use relative links on
428 websites. Let's demonstrate on this link:
430 <a href="bar">bar</a>
432 If the user loads the url C</foo/> and clicks on this link, they will be
433 sent to C</foo/bar>. However when they are on the url C</foo> and click this
434 link, then they will be sent to C</bar>.
436 This makes it necessary to be explicit about the trailing slash.
438 =head3 Extension matches
442 will match .html from the path (assuming the subroutine itself returns
443 something, of course). This is normally used for rendering - e.g.:
446 response_filter { $self->render_html($_[1]) }
453 will match any extension and supplies the extension as a match argument.
455 =head3 Query and body parameter matches
457 Query and body parameters can be match via
459 sub (?<param spec>) { # match URI query
460 sub (%<param spec>) { # match body params
462 The body spec will match if the request content is either
463 application/x-www-form-urlencoded or multipart/form-data - the latter
464 of which is required for uploads - see below.
466 The param spec is elements of one of the following forms:
468 param~ # optional parameter
469 param= # required parameter
470 @param~ # optional multiple parameter
471 @param= # required multiple parameter
472 :param~ # optional parameter in hashref
473 :param= # required parameter in hashref
474 :@param~ # optional multiple in hashref
475 :@param= # required multiple in hashref
476 * # include all other parameters in hashref
477 @* # include all other parameters as multiple in hashref
479 separated by the C<&> character. The arguments added to the request are
480 one per non-C<:>/C<*> parameter (scalar for normal, arrayref for multiple),
481 plus if any C<:>/C<*> specs exist a hashref containing those values.
483 Please note that if you specify a multiple type parameter match, you are
484 ensured of getting an arrayref for the value, EVEN if the current incoming
485 request has only one value. However if a parameter is specified as single
486 and multiple values are found, the last one will be used.
488 For example to match a C<page> parameter with an optional C<order_by> parameter one
491 sub (?page=&order_by~) {
492 my ($self, $page, $order_by) = @_;
493 return unless $page =~ /^\d+$/;
496 $_[1]->search_rs({}, $p);
500 to implement paging and ordering against a L<DBIx::Class::ResultSet> object.
502 Another Example: To get all parameters as a hashref of arrayrefs, write:
505 my ($self, $params) = @_;
508 To get two parameters as a hashref, write:
510 sub(?:user~&:domain~) {
511 my ($self, $params) = @_; # params contains only 'user' and 'domain' keys
513 You can also mix these, so:
515 sub (?foo=&@bar~&:coffee=&@*) {
516 my ($self, $foo, $bar, $params);
518 where $bar is an arrayref (possibly an empty one), and $params contains
519 arrayref values for all parameters B<not> mentioned and a scalar value for
520 the 'coffee' parameter.
522 Note, in the case where you combine arrayref, single parameter and named
523 hashref style, the arrayref and single parameters will appear in C<@_> in the
524 order you defined them in the protoype, but all hashrefs will merge into a
525 single C<$params>, as in the example above.
527 =head3 Upload matches
529 sub (*foo=) { # param specifier can be anything valid for query or body
531 The upload match system functions exactly like a query/body match, except
532 that the values returned (if any) are C<Web::Dispatch::Upload> objects.
534 Note that this match type will succeed in two circumstances where you might
535 not expect it to - first, when the field exists but is not an upload field
536 and second, when the field exists but the form is not an upload form (i.e.
537 content type "application/x-www-form-urlencoded" rather than
538 "multipart/form-data"). In either of these cases, what you'll get back is
539 a C<Web::Dispatch::NotAnUpload> object, which will C<die> with an error
540 pointing out the problem if you try and use it. To be sure you have a real
543 $upload->is_upload # returns 1 on a valid upload, 0 on a non-upload field
545 and to get the reason why such an object is not an upload, call
547 $upload->reason # returns a reason or '' on a valid upload.
549 Other than these two methods, the upload object provides the same interface
550 as L<Plack::Request::Upload> with the addition of a stringify to the temporary
551 filename to make copying it somewhere else easier to handle.
553 =head3 Combining matches
555 Matches may be combined with the + character - e.g.
557 sub (GET + /user/*) {
559 to create an AND match. They may also be combined withe the | character - e.g.
563 to create an OR match. Matches can be nested with () - e.g.
565 sub ((GET|POST) + /user/*) {
567 and negated with ! - e.g.
569 sub (!/user/foo + /user/*) {
571 ! binds to the immediate rightmost match specification, so if you want
572 to negate a combination you will need to use
574 sub ( !(POST|PUT|DELETE) ) {
576 and | binds tighter than +, so
578 sub ((GET|POST) + /user/*) {
582 sub (GET|POST + /user/*) {
586 sub ((GET + /admin/...) | (POST + /admin/...)) {
590 sub (GET + /admin/... | POST + /admin/...) {
592 are not - the latter is equivalent to
594 sub (GET + (/admin/...|POST) + /admin/...) {
596 which will never match!
600 Note that for legibility you are permitted to use whitespace:
602 sub (GET + /user/*) {
604 but it will be ignored. This is because the perl parser strips whitespace
605 from subroutine prototypes, so this is equivalent to
609 =head3 Accessing parameters via C<%_>
611 If your dispatch specification causes your dispatch subroutine to receive
612 a hash reference as its first argument, the contained named parameters
613 will be accessible via C<%_>.
615 This can be used to access your path matches, if they are named:
617 sub (GET + /foo/:path_part) {
619 ['Content-type' => 'text/plain'],
620 ["We are in $_{path_part}"],
624 Or, if your first argument would be a hash reference containing named
627 sub (GET + /foo + ?:some_param=) {
629 ['Content-type' => 'text/plain'],
630 ["We received $_{some_param} as parameter"],
634 Of course this also works when all you are doing is slurping the whole set
635 of parameters by their name:
637 sub (GET + /foo + ?*) {
639 ['Content-type' => 'text/plain'],
640 [exists($_{foo}) ? "Received a foo: $_{foo}" : "No foo!"],
644 Note that only the first hash reference will be available via C<%_>. If
645 you receive additional hash references, you will need to access them as
648 =head3 Accessing the PSGI env hash
650 In some cases you may wish to get the raw PSGI env hash - to do this,
651 you can either use a plain sub:
658 or use the C<PSGI_ENV> constant exported to retrieve it from C<@_>:
660 sub (GET + /foo + ?some_param=) {
662 my $env = $_[PSGI_ENV];
665 but note that if you're trying to add a middleware, you should simply use
666 Web::Simple's direct support for doing so.
668 =head1 EXPORTED SUBROUTINES
670 =head2 response_filter
673 # Hide errors from the user because we hates them, preciousss
674 if (ref($_[0]) eq 'ARRAY' && $_[0]->[0] == 500) {
675 $_[0] = [ 200, @{$_[0]}[1..$#{$_[0]}] ];
680 The response_filter subroutine is designed for use inside dispatch subroutines.
682 It creates and returns a special dispatcher that always matches, and calls
683 the block passed to it as a filter on the result of running the rest of the
684 current dispatch chain.
686 Thus the filter above runs further dispatch as normal, but if the result of
687 dispatch is a 500 (Internal Server Error) response, changes this to a 200 (OK)
688 response without altering the headers or body.
692 redispatch_to '/other/url';
694 The redispatch_to subroutine is designed for use inside dispatch subroutines.
696 It creates and returns a special dispatcher that always matches, and instead
697 of continuing dispatch re-delegates it to the start of the dispatch process,
698 but with the path of the request altered to the supplied URL.
700 Thus if you receive a POST to C</some/url> and return a redispatch to
701 C</other/url>, the dispatch behaviour will be exactly as if the same POST
702 request had been made to C</other/url> instead.
704 Note, this is not the same as returning an HTTP 3xx redirect as a response;
705 rather it is a much more efficient internal process.
707 =head1 CHANGES BETWEEN RELEASES
709 =head2 Changes between 0.004 and 0.005
713 =item * dispatch {} replaced by declaring a dispatch_request method
715 dispatch {} has gone away - instead, you write:
717 sub dispatch_request {
719 sub (GET /foo/) { ... },
723 Note that this method is still B<returning> the dispatch code - just like
726 Also note that you need the C<< my $self = shift >> since the magic $self
729 =item * the magic $self variable went away.
731 Just add C<< my $self = shift; >> while writing your C<< sub dispatch_request { >>
732 like a normal perl method.
734 =item * subdispatch deleted - all dispatchers can now subdispatch
736 In earlier releases you needed to write:
738 subdispatch sub (/foo/...) {
741 sub (GET /bar/) { ... },
746 As of 0.005, you can instead write simply:
751 sub (GET /bar/) { ... },
758 =head2 Changes since Antiquated Perl
762 =item * filter_response renamed to response_filter
764 This is a pure rename; a global search and replace should fix it.
766 =item * dispatch [] changed to dispatch {}
770 dispatch [ sub(...) { ... }, ... ];
774 dispatch { sub(...) { ... }, ... };
780 =head1 DEVELOPMENT HISTORY
782 Web::Simple was originally written to form part of my Antiquated Perl talk for
783 Italian Perl Workshop 2009, but in writing the bloggery example I realised
784 that having a bare minimum system for writing web applications that doesn't
785 drive me insane was rather nice and decided to spend my attempt at nanowrimo
786 for 2009 improving and documenting it to the point where others could use it.
788 The Antiquated Perl talk can be found at L<http://www.shadowcat.co.uk/archive/conference-video/> and the slides are reproduced in this distribution under
789 L<Web::Simple::AntiquatedPerl>.
791 =head1 COMMUNITY AND SUPPORT
795 irc.perl.org #web-simple
797 =head2 No mailing list yet
799 Because mst's non-work email is a bombsite so he'd never read it anyway.
801 =head2 Git repository
803 Gitweb is on http://git.shadowcat.co.uk/ and the clone URL is:
805 git clone git://git.shadowcat.co.uk/catagits/Web-Simple.git
809 Matt S. Trout (mst) <mst@shadowcat.co.uk>
813 Devin Austin (dhoss) <dhoss@cpan.org>
815 Arthur Axel 'fREW' Schmidt <frioux@gmail.com>
817 gregor herrmann (gregoa) <gregoa@debian.org>
819 John Napiorkowski (jnap) <jjn1056@yahoo.com>
821 Josh McMichael <jmcmicha@linus222.gsc.wustl.edu>
823 Justin Hunter (arcanez) <justin.d.hunter@gmail.com>
825 Kjetil Kjernsmo <kjetil@kjernsmo.net>
827 markie <markie@nulletch64.dreamhost.com>
829 Christian Walde (Mithaldu) <walde.christian@googlemail.com>
831 nperez <nperez@cpan.org>
833 Robin Edwards <robin.ge@gmail.com>
835 Andrew Rodland (hobbs) <andrew@cleverdomain.org>
837 Robert Sedlacek (phaylon) <r.sedlacek@shadowcat.co.uk>
841 Copyright (c) 2011 the Web::Simple L</AUTHOR> and L</CONTRIBUTORS>
846 This library is free software and may be distributed under the same terms