> -----Oorspronkelijk bericht-----
> Van: Michiel Meeuwissen [mailto:[EMAIL PROTECTED]
> Verzonden: dinsdag 16 maart 2004 11:31
> Aan: [EMAIL PROTECTED]
> Onderwerp: Re: image magic (moved from users mail group)
> 
> 
> Ernst Bunders <[EMAIL PROTECTED]> wrote:
> > I move this thread to dev becouse i think there are some 
> developer related
> > issues i would like to point out.
> > Last weekend i looked into the current imagemagic and JAI 
> implementation and
> > allso in the imagemagick java api that Michiel mentioned as 
> an option for a
> > third converter implementation. I find that the current 
> situation is a bit
> > awkward, and maybe will need to be changed in the future, 
> with massive
> > backwards-compatability issues:
> > 
> > 1 there is a small number of commands that are being abstracted from
> > implementation. This set is allso the currently implemented 
> in ConvertJAI.
> > Apart from this set all other commands are just passed to 
> imagemagick on the
> > fly. Errors can not be cougt and if the imagemagick people 
> would decide
> > (unlikely, but beyond hour control) to change these commands all
> > non-abstracted imagetemplates in mmabse templates would be broken.
> 
> I'm not sure whether this was a good idea. It seemed like a 
> more or less
> sensible, because it's hopeless the wrap every single one of 
> the massive number
> of options that imagemagick provides in an 'abstraction' of 
> our own. And
> then also hope that we could implement then excactly the same as in
> imagemagick somewhere else.

I think an abstract set would have to be a limited set. 

> 
> > 2 Due to an apperant change in java (1.4.1) specifications 
> it is no longer
> > possible to pass characters outside the ascii set to 
> imagemagick. This makes
> > all text functions of imagemagick pretty much useless. The 
> current java api
> > supports a limited set of commmands and i think text is not 
> supported (yet).
> 
> Yes, this really sucks. I invested litteraly hours, if not 
> days, to get text
> in images working (it was not possible at all before I did 
> that), only to
> find out that all effort is rendered useless in this ****** java 1.4
> 
> > 3 If you would want to make a more complete JAI     
> implementation you would
> > practically hav to rebuild imagemagick in java in order to 
> ensure backwards
> > competability.
> 
> > What dous this lead to? I think it is unfortunate that the 
> abstraction of
> > commands in the image templates has been dropped. It is not 
> mmbase-like
> > seems and it seems that we are stuck with one 
> implementation (imagemagick),
> > and that implementation dous not function well anymore 
> becouse of the
> > different behavour of Runtime.exec (in java 1.4.1). 
> 
> The other alternative was not to provide the options of 
> imagemagick, in
> which case you can just as wel not use it. Furthermore, some of these
> 'abstractions' IIRC gave a kind of feel of compatibility which was not
> really there, because they only provide approximately the same result.
> 
> > It is perhaps not an immediate problem, but still it is 
> messy, and perhaps
> > needs some kind of attention.
> 
> I'd go for ImageMagic only, with 'jai' only as a fall-back. 
> This provides a
> lot of features, which we don't have to support.
> 
> It's a real petty that Runtime.exec is useless nowadays in 
> java, so we have
> to look for an alternative. I suppose the java-interface to 
> imagemagick
> would supply the same features.
> 
> Consicely, I think supporting both complete abstraction of the
> convert-interface (independent from the used 'templates'), 
> and a reasonalbly
> rich set of features is impossible.
> 
> The current situation of a limited set of features which are 
> 'abstract' and
> everything else simple translates directly to imagesmagick is 
> I think simple
> and maintainable, though it should perhaps be documented more 
> explicitily
> that using image-magick-features makes your site dependent on it.
> 

It seems to me that a partly abstracted set is useless for it has no
function and is only confusing for users. Either you abstract or you don't.
Allso, the java api to imagemagick is not finished yet. It is being
developed by one guy hwo is actually inviting people to make requests for
implementation. So this problem could solve itself in time.
So it might be a good idear to start on an implementation of this api, with
featurs becomming available in time. This way we can allso produce
meaningfull error messages. Downside is we would have to develope a parser
for the templates. 

Ernst

> 
> Michiel
> 
> 
> -- 
> Michiel Meeuwissen 
> Mediapark C101 Hilversum  
> +31 (0)35 6772979
> nl_NL eo_XX en_US
> mihxil'
>  [] ()
> 

Reply via email to