[jrawio-dev] Re: Re: Re: Re: Re: API CHANGES: *.raw packages renamed to *.craw

[Igor Malinin <igor@alveolo.org>]

Fabrizio Giudici wrote:
> And, yes, I'd like to know how many people are using jrawio. For the 
> mailing list, we're about 15 - I suppose there could be more people 
> using it? Also, if it doesn't violate any reserved information, I'd be 
> happy to know how people are using it.

I just started to experiment with it. I had the idea of opensource 
'platform' for raw tools year ago, then forgot it, then not so far ago 
my colleague shared with me some similar ideas, so I started 
investigation again. I had not so much free time to contribute 
something, except the patch for my own recently bought 500D.

I did some very basic RAW file reading, experimented with debayering, 
some color management and basic processing like exposure compensation. 
Tried to compare performance with C/C++ code. Unfortunately I can only 
state that C code is MUCH fasted than Java, and it is not 20-30%... It 
is 2-3 times difference, depending on hardware! :(

Then was fighting with Swing - it took me several evenings to show large 
picture on JScrollPane with acceptable scrolling performance, with SWT - 
just first try worked! But finally got comparable performance with 
Swing. Found out that color space conversion is not so good performing 
in Java, actually it is terribly slow to do it on the fly, and 
implementation is not trying to optimize it even for most common image 
data representations.

Then tried to search alternatives to NetBeans RCP, I wanted some 
Swing/OSGi based platform, but it appeared that there are none :( Don't 
want to do anything based on NB - probably it is only my old experience 
with a this platform, but in my memory it sits quite firmly as the worst 
thing to base your product on - experience from integrating RefactorIt 
to everything in the world.


As with jrawio project itself I had many issues too and as I told 
already I would like to share this so probably my problems could 
encourage to make the project more friendly, especially for contributions.

First, it is not very good to the project to change 'homes' so 
frequently. Yes, now I understand that official home is still at 
jrawio.tidalwave.it, but it was not so obvious when I sow only something 
like "site is moving" @tidalwave and project created @java.net. Then I 
did checkout from SVN @java.net, experimented with it, did a patch for 
my 500D, then wanted to contribute, but discovered... guess what - SVN 
repository is empty! After some search (non obvious again) I discovered 
that sources is on kenai now, and repository is Mercurial. Uhh.. Start 
learning new thing.. Happily there is EclipseMercurial, and more happily 
it even working somehow! And happily it became usable just few days 
ago!!! It still not perfect, but at least I can do something from IDE 
now. So, created a patch, submitted it, only to discover than it failed 
to work well after pulling applied patch. :( So, hardly something can be 
reverted now, but IMHO it was too early to experiment with Hg. Tried Hg 
year ago too, experience is much better now, but tools are still far 
from famous CVS/SVN. Hopefully it will change very soon - the speed of 
MercurialEclipse development is amazing, but it is mostly from single 
person unfortunately.

Second my problem is Maven. Probably I most unlucky person with this 
tool, I like the idea behind of Maven very much, but how it is 
implemented - just terrible. Hardly you like to revert it now (I would 
prefer just Ant/Ivy), but improvement of experience could be greatly 
improved, I hope so. One problem - the build takes enormous amount of 
time, and sometimes failing. I hoped that this large time will be only 
on first run (to pre-load dependencies), but this happens every time. 4 
hours (and even more) for build such a small chunk of code is not 
normal! Another problem is the documentation, if there was some small 
description on your site on how to use it, a list of most useful Maven 
goals, to build generated java sources only for example without building 
whole project + running unit tests.
And - do we really need this 'generation' step at all? It just code, and 
in Java it would be more flexible than XML. It is 1:1 mapping anyway, or 
am I missing something? If it is only HTML summary generation - don't 
feel a need for it at all.

And third. Your choice for code formatting is [sorry] - terrible! Yep, I 
would accept it, consistent rules for the project is a must, and good or 
bad, they are always better than not having it, but my IDE (Eclipse) 
refuses to format your way, independent of my tries to configure it. It 
is only one thing that causing problem - but adding spaces by hand after 
every insert of any brace - it's madness! Why not take rules that Sun 
recommends for Java code? They are quite common it the community and 
when they are 'not as I would like it', I was quickly used to them an 
they are not so bad! Usually the only exception is added to them - using 
either tabs or spaces for indentation, not both mixed as Sun suggests 
(this is really terrible suggestion)...

So, it is all for this time, sorry if sometimes I used not correct 
wording. A little of emotions should not fear you, I hope it is the 
feedback you liked to get :) Anyway you are doing a great work!