Re: Request#accept_media_types

[mynyml <martin.aumont@gmail.com>]

@Joshua: Nice catch about the quality values. The specs do specify
they can contain up to three decimal digits. I updated the code to
reflect that.

Now about Apache's behaviour of assigning 0.01 to */* and 0.02 to type/
*: I didn't know about this so I did some reading. Seems that Apache
does that as a fix for browsers that don't send any quality values at
all, as it assumes they don't really know what they're doing.

  http://httpd.apache.org/docs/1.3/content-negotiation.html#better

Turns out though there is another part of the HTTP spec (that I had
previously missed), which is that more specific types override less
specific ones. This automatically takes care of apache's behaviour. I
updated the code to include this too.

@Ryan: The code was becoming too repetitive so I split the media type
handling into it's own class. I put it in Rack::Mime::MimeType. I
guess that could provide a starting point. It could eventually
subclass something more generic.

@Yehuda: TBH I don't think I know enough performance optimization
tricks to add significant improvements, but I added the basics by
caching the parsed output.


Latest patch is @
  http://github.com/mynyml/rack/commit/2f0a743fe3a55b07cf0fc75ba06cb3180ddef7be

Questions, suggestions, patches?
Anyone wants to tackle performance? =)


On Jun 15, 3:58 pm, Ryan Tomayko <r...@tomayko.com> wrote:
> I haven't reviewed the patch yet but I'd love to have a set of simple
> utility classes/methods/mixins for parsing basic HTTP header value
> types. Accept, Accept-Language, Accept-Encoding, Cache-Control, etc.
> all follow a few simple syntax grammars. We could then use the
> utilities from Rack::Request, or directly from framework/app code if
> more control is needed.
>
> Ryan
>
> On Sun, Jun 14, 2009 at 8:39 AM, Yehuda Katz<wyc...@gmail.com> wrote:
> > Take a look at what we do in Merb. We've optimized our Accept header parsing
> > quite a bit (for performance; it can be shockingly slow), and handle quality
> > factors (both absolute and relative).
>
> > -- Yehuda
>
> > On Sun, Jun 14, 2009 at 9:50 AM, Joshua Hull <joshbu...@gmail.com> wrote:
>
> >> Just a couple of comments on this:
>
> >> types = types.select {|type| media_type_quality(type).between?(0.1, 1) }
>
> >> this should probably check that its greater than 0 and less than or
> >> equal to 1 instead
>
> >> types = types.select {|type| quality = media_type_quality(type);
> >> quality > 0 && quality <= 1 }
>
> >> so that values of 0.01 still get through.
>
> >> As well, the default value of all types is 1.0, but we probably want
> >> to do something similar to what apache does, and assign */* a value of
> >> 0.01 and type/* a value of 0.02.
>
> >> Cheers
> >> Joshua
>
> >> On Fri, Jun 12, 2009 at 5:48 PM, mynyml<martin.aum...@gmail.com> wrote:
>
> >> > Patch is at:
> >> >http://github.com/mynyml/rack/commit/c9953dbc3f834fdcbcd1ebc8b71b64ea...
> >> > -----
> >> > Patch to add parsing of the HTTP_ACCEPT header's media types. Orders
> >> > types by "quality" (preference level) (following HTTP 1.1 specs).
>
> >> > The main advantage of being able to parse the Accept header is to
> >> > centralize the lookup for the request's "desired" type(s). Usually, a
> >> > client asks for a specific media type to be sent back by appending an
> >> > extension to the URL; not only does this require parsing by every
> >> > middleware/app that needs to know about this media type, but it also
> >> > ignores the Accept hierarchy, which can be usefull for cascading down
> >> > possible handlings of the resquest (think of a respond_to mechanism,
> >> > for instance).
>
> >> > I've encountered many use cases already:
>
> >> > o
> >> >http://github.com/rack/rack-contrib/blob/3f42d3afe7323d322567d77cd404...
> >> > ohttp://github.com/mynyml/rack-abstract-format
> >> > ohttp://github.com/mynyml/rack-supported-media-types
> >> > ohttp://github.com/mynyml/rack-js4xhr
> >> > ohttp://github.com/mynyml/rack-respond_to
>
> >> > which led me to write the Accept header parsing code:
>
> >> > ohttp://github.com/mynyml/rack-accept-media-types
>
> >> > In my rack clone, the request_accept branch is a straight port of the
> >> > rack-accept-media-types gem, and the request_accept-compact branch
> >> > minimizes the code into two methods (instead of two classes). The
> >> > latter reflects the rest of the Request class much better so it's
> >> > probably a more appropriate patch.
>
> >> > Feedback/questions/suggestions ?
>
> >> > Thanks
>
> > --
> > Yehuda Katz
> > Developer | Engine Yard
> > (ph) 718.877.1325
>
>