Gdal Grass

in these page i'll try to make annotation about the gdal grass code. please add your note about it, suggestion, documentation to read or any thing youn think can help me to learn more about it. i haven't a c++ base, i changed the time line to give more time to thise part of the project.

As start point i had a discussion on Irc with Frank W. about the problems i had tring to load grass raster in ossim :

[2009-04-10 16:46:08] <epifanio> hi FrankW 
[2009-04-10 16:46:25] <epifanio> i've just a little question to you :-/
[2009-04-10 16:46:40] <epifanio> maybe OT on qgis, so i used provate message
[2009-04-10 16:46:51] <FrankW> sure, fire away
[2009-04-10 16:47:08] <epifanio> i'm tring to learn more about the ossim_gdal_plugin
[2009-04-10 16:47:16] <epifanio> are you familiar with it ?
[2009-04-10 16:47:38] <epifanio> the problem you expose me the last time was related to the color table
[2009-04-10 16:47:40] <FrankW> I had done some work on it at one time.
[2009-04-10 16:47:55] <epifanio> the gdal plugin probably only recognizes 8 bit pseudo color data
[2009-04-10 16:48:12] <epifanio> but if gdal can read it try translating to 3 band tiff then planet could read it
[2009-04-10 16:48:59] <FrankW> Potentially the GDAL pct2rgb.py script could be used to convert to an RGB file.
[2009-04-10 16:49:13] <FrankW> But I'm not even positive that supports more than 8bit images! 
[2009-04-10 16:49:32] <FrankW> Ultimately though I imagine bulk translation would not be the desired approach.
[2009-04-10 16:50:07] <epifanio> yes it is ... beacose it implies a data duplication
[2009-04-10 16:50:12] <epifanio> i think ...
[2009-04-10 16:50:29] <FrankW> You *want* data duplication?
[2009-04-10 16:50:35] <epifanio> no
[2009-04-10 16:50:47] <epifanio> i prefer to avoid data duplication
[2009-04-10 16:51:19] <FrankW> ah, that was what I expected.
[2009-04-10 16:51:34] <epifanio> beacouse my targhet is to read grass data directly
[2009-04-10 16:51:44] <FrankW> Understood.
[2009-04-10 16:51:59] <FrankW> The comment "yes it is" in reply to my suggesting bulk translation was not desirable confused me.
[2009-04-10 16:52:30] <epifanio> ahhh ... my poor english sorry :-(
[2009-04-10 16:52:37] <FrankW> It may be that the ossimGdalTileSource::loadIndexTo3BandTile method could be updated to support non-8bit data. 
[2009-04-10 16:53:08] <FrankW> I'm looking at OSSIM code from mid-2006 I happen to have sitting around, but I agree that this code is 8bit oriented, but it seems like it could be correctedly reasonably easily.
[2009-04-10 16:53:18] <epifanio> do you have "under hand" the ossim_plugin_source code ?
[2009-04-10 16:53:23] <epifanio> ok 
[2009-04-10 16:53:57] <FrankW> What happens when you load your non-8bit image by the way?  A crash? An exception?  A runtime error?  Or just the wrong visual results?
[2009-04-10 16:54:10] <epifanio> no, it load the data
[2009-04-10 16:54:21] <epifanio> recognize the right projection too
[2009-04-10 16:54:31] <epifanio> but it iis sympli no "rendered"
[2009-04-10 16:54:58] <epifanio> i can have a black image or a "band series" 
[2009-04-10 16:55:37] <epifanio> like the image is composed by different coloured strip
[2009-04-10 16:55:56] <epifanio> i can produce a scrrenshot
[2009-04-10 16:58:30] <epifanio> these is a screen in imagelinker for the elevation.10m from the spearfish dataset :http://www.geofemengineering.it/data/elevation.10m.png
[2009-04-10 17:00:24] <epifanio> while the slope : http://www.geofemengineering.it/data/slope.png  is rendered
[2009-04-10 17:01:28] <epifanio> FrankW: these the link :  http://www.geofemengineering.it/data/elevation.10m.png
[2009-04-10 17:01:32] <FrankW> I was able to look at the slope.png
[2009-04-10 17:01:48] <epifanio> can you see it ?
[2009-04-10 17:01:54] <epifanio> the elevation.10m
[2009-04-10 17:02:11] <FrankW> Based on the screen shot of the elevation (I see it now), I believe a 16 or 32bit buffer is just being directly intepreted as 8bit. 
[2009-04-10 17:02:44] <FrankW> This reassures me that it may be easy to modify loadIndexTo3BandTile() to support non-eight bit imagery.
[2009-04-10 17:03:48] <epifanio> ohh :-)  i haven't knowledge about cpp/c code,  but if you pooint me on the right source ... i'l try to read and learn a bit from the source file
[2009-04-10 17:04:18] <epifanio> so i need to point my attention on :  ossimGdalTileSource.cpp .h
[2009-04-10 17:04:26] <epifanio> ?
[2009-04-10 17:04:30] <FrankW> yes
[2009-04-10 17:05:12] <FrankW> http://trac.osgeo.org/ossim/browser/trunk/ossim_plugins/gdal/ossimGdalTileSource.cpp#L1516
[2009-04-10 17:05:35] <FrankW> The core problem is that the result of getBuf() is cast to ossim_uint8* without regard to the true type of the buffer.
[2009-04-10 17:06:03] <FrankW> I believe the code needs to be modified to handle three cases - 8bit, 16bit and 32bit integers and to error out gracefully if it isn't one of those cases.
[2009-04-10 17:06:45] <FrankW> Down at line 1567 where the s[sample] value it fetched from the input buffer this would need to be altered to fetch it based on the type of the buffer.
[2009-04-10 17:07:23] <FrankW> The GDALGetColorEntryAsRGB() already supports non-eight bit palette entries so that aspect should not be a problem.
[2009-04-10 17:08:11] <epifanio> huhh i'm reading .. and follow you :-)
[2009-04-10 17:09:44] <epifanio> each word is for me a great hand, thanks
[2009-04-10 17:09:49] <epifanio>  right now i haven't the nedeed knowledge to undstoand ... 
[2009-04-10 17:10:01] <epifanio> but i can document all, reading more .. and try to send a relative mail/irc  to the ossimpanet dev-team 
[2009-04-10 17:10:28] <FrankW> sounds good - those guys should have no problem helping with the code changes.  I'm glad to answer further questions too. 
[2009-04-10 17:11:16] <epifanio> maybe that's what i can do actually,  yes nhv gpots  gived me a great help and like you ar so kind 
[2009-04-10 17:11:49] <FrankW> Just now however I need to drive my daughter somewhere ... later!
[2009-04-10 17:11:59] <epifanio> i'll try to documnet all , done screenshot and the relative gdalinfo information 
[2009-04-10 17:12:11] <epifanio> thanks !!! see you at the next time :-)