Mime-type on serializers

[Luigi Bai <lp...@focalpoint.com>]

Please see below for a proposal, following the use-case:

On Sat, 18 Sep 2004, Jorg Heymans wrote:

>
>
> Luigi Bai wrote:
>> Hi,
>> 
>> I've got a XML database with images of various sorts of mime-types (jpeg, 
>> gif, png, etc) stored in text() nodes as Base64. I have a trivial 
>> Base64Serializer that reconstitutes them on the way out. However, I think 
>> I'm stuck with a mime-type problem. It seems I have to have a separately 
>> instantiated Serializer for each possible mime-type:
>
> correct
>
>> 
>> <map:select type="mimeSelector">
>>  <map:when test="image/png">
>>   <map:serialize type="base64" mime-type="image/png" />
>>  </map:when>
>> ...etc...
>> 
>> I find this verbose, and annoying to maintain if someone saves an image 
>> with a new mime-type.
>
> i agree, others might not.
>
>> 
>> The Serializer interface has a getMimeType() method which allows a 
>> Serializer to declare its mime-type, but only at setup() time (correct me 
>> if I'm wrong?). I need to declare the mime-type later; the first 
> correct. getMimeType() is called during setup() which is called during 
> pipeline setup before any data is processed.
>
>
> IIRC it works this ways so that cocoon can start streaming SAX events to the 
> client as they roll in, and doesn't have to wait until the serializer is 
> fully finished processing.
>
>

I would like to propose a change to how mime-types in a pipeline work, so 
a Serializer can set the mime-type after setup() (i.e., if the mime-type 
depends on some SAX events).

I would like Serializer to have a method
boolean willSetMimeType()

to let the pipeline builder know that despite returning null in 
getMimeType(), the Serializer *will* want to set the mime type later in 
the stream. This will tell AbstractProcessingPipeline.setupPipeline() not 
to bail if it can't determine a mimeType.

The problem is that I can't get a handle on *how* the Serializer would 
later set the mime-type; it doesn't seem to have access to the Environment 
in which that information is stored.

The Pipeline would also probably need to be told not to start streaming to 
the client right away, but to probably do that when the Serializer's 
finished. Maybe it can just deduce that from the Serializer returning 
"true" from willSetMimeType().

This would throw a wrench in the caching policy of creating the Serializer 
to see what the mime-type of the cached object is; you'd have to run the 
Serializer each time to determine the mime-type. UNLESS, of course, the 
mime-type is cached with the object. Which seems like a reasonable thing 
to do.

Any comments?