[Catalyst] Catalyst, REST and Firefox

Jonas Alves jonas.alves at gmail.com
Thu Jan 17 22:35:16 GMT 2008


On Jan 17, 2008 6:03 PM, Christopher H. Laco <claco at chrislaco.com> wrote:
>
> Jonas Alves wrote:
> > On Jan 17, 2008 2:32 PM, Christopher H. Laco <claco at chrislaco.com> wrote:
> >> I've touched on this before, and posted about it on UP:
> >> http://use.perl.org/~jk2addict/journal/35411
> >>
> >> In a nutshell, Firefox 2.x Accept header totaly screws with the REST
> >> controller when you use it as a base for View negotiations. If you have
> >> a default type of text/html pointed to a View::TT, REST will see
> >> text/xml from Firefox and try and send that instead, based on this header:
> >>
> >> text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5
> >>
> >> What does everyone think about a config option/toggle to tell REST to
> >> ignore the Accept header, allowing it to fall back to the default
> >> content-type in config when no Cntent-Type header or content-type params
> >> are specified?
> >>
> >> -=Chris
> >>
> >
> > I have a subclass of Action::Serialize that does this:
> >
> > my $default       = $controller->config->{serialize}{default};
> > my $pcontent_type = $c->req->preferred_content_type;
> > my $content_types = $c->req->accepted_content_types_qvalue;
> > my $ordered_content_types = $c->req->accepted_content_types;
> > my $max_qvalue = $content_types->{$pcontent_type};
> > my %max_qvalue_content_types = map {
> >         $content_types->{$_} eq $max_qvalue ? ($_ => $default) : ()
> > } keys %$content_types;
> > my $content_type = $max_qvalue_content_types{$default}
> >                                  || $pcontent_type
> >                                  || $c->req->content_type;
> >
> > And in a subclass of Request::REST mixed with Plugin::Flavour:
> >
> > sub preferred_content_type {
> >     my $self = shift;
> >     if ($self->flavour) {
> >         my $type = $self->{ext_map}{$self->flavour}
> >                          || $self->_ext_to_type($self->flavour);
> >         return $type;
> >     }
> >     $self->accepted_content_types->[0];
> > }
> >
> > sub accepted_content_types_qvalue {
> >     my $self = shift;
> >     return $self->{content_types_qvalue} if $self->{content_types_qvalue};
> >
> >     my %types;
> >     if ($self->method eq "GET" && $self->param('content-type')) {
> >         $types{ $self->param('content-type') } = 2;
> >     }
> >
> >     # This is taken from chansen's Apache2::UploadProgress.
> >     if ( $self->header('Accept') ) {
> >         $self->accept_only(1) unless keys %types;
> >
> >         my $accept_header = $self->header('Accept');
> >         my $counter       = 0;
> >
> >         foreach my $pair ( split_header_words($accept_header) ) {
> >             my ( $type, $qvalue ) = @{$pair}[ 0, 3 ];
> >             next if $types{$type};
> >             $qvalue = 1 unless defined $qvalue;
> >             $types{$type} = sprintf '%.3f', $qvalue;
> >         }
> >     }
> >     return $self->{content_types_qvalue} = \%types;
> > }
> >
> >
> > That way all works fine. If you add an extension to the uri (like
> > .json or .xml), it serves that content-type. Else it tries to find the
> > greater qvalue on the Accept Header and tries to serve that. If it has
> > more than one content-type with the same max qvalue it tries to find
> > the default content-type in that list and serve it. If it isn't in the
> > max qvalue list than it serves the first content-type on that list.
> > I think that is a sane approach.
> >
>
> Well volunteered! I can't speak for the flavour stuff, but I'd think the
> q value logic would be well served in the REST package. :-)
>
> I'd personally like the flavour stuff in there as well. In my case, I'm
> half way there. I have a type= that takes a friendly name (atom, rss,
> json) ... along wieh adding some of those content types to MIME::Types
> since it's missing a few of them for type->extension mapping.
>

Well, I don't mind to help in that too. I think that the flavour stuff
is very useful. Especially if you need to use the REST interface from
some lib/framework/api where isn't easy to set the Accept header. And
I think that is much more clean than the query-string type parameter.

-- 
Jonas



More information about the Catalyst mailing list