[C1] proposed change to formatter interface

[Robin Green <gr...@hotmail.com>]

As a product of discussions with Uli on cocoon-users on outputting non-XML 
content directly, such as .DOC or .PDF files, I think it would be an 
improvement to tweak the Formatter interface to take an OutputStream instead 
of a Writer (and then tweak all the implementations, of course). This would 
serve two purposes:

1. Efficiency and reducing likelihood of OutOfMemoryErrors - Strings can be 
converted into bytes in smaller chunks - rather than in one go as it is now, 
which requires effectively two buffers (String and byte[]) equal to the size 
of the entire output. (Actually there are 3 buffers! One can easily be 
removed, which I'll do - it was just a very quick hack.) When generating 
large output (e.g. PDF) this is a big waste and might require unnecessary 
cache freeing-up. Plus, if a serializer has a faster way of converting chars 
to bytes under a given encoding than String.getBytes (), this is the only to 
take advantage of it!

2. I'd like to do a LiteralFormatter to do what Uli is trying to do but in a 
way more suited to C1. (Unlike if you use XSP, this takes full advantage of 
C1's response header support - heh, not much yet, but I plan to add more - 
and caching, so there _are_ possible advantages). Anyway, it's 
counterintuitive to output binary data through a Writer and store it in a 
String. It would actually work, providing you choose an appropriate 
encoding, but it just makes more sense to use an OutputStream.

(As a knock-on effect .../framework/Page.java would store cacheable content 
as a byte array rather than a String.)



_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.