iLiad DJVU for iLiad

[scotty1024]

I pulled down the sources for DJVU and looked at it a bit.

Looks doable in a short time. Anyone want to get your "15 minutes of glory" and make it happen?

[arivero]

Originally Posted by scotty1024

I pulled down the sources for DJVU and looked at it a bit.

Looks doable in a short time. Anyone want to get your "15 minutes of glory" and make it happen?

The problem could be the Xwindows application, djview. All the others will compile straigh with a decent toolchain. It is already one year since I did it, and I have not set up a toolchain yet. So if you want to try...

[scotty1024]

Hmmm isn't there a DJVU viewer in Java?

[arivero]

Originally Posted by scotty1024

Hmmm isn't there a DJVU viewer in Java?

Are you able to install a java engine ?

[Mike Kostousov]

It is not difficult to write djvu viewer. I already tried, but I don't understand how to work with iliad's X server.
I can post here simple code which opens djvu and extracts a page to a pixels buffer.

[arivero]

Hmm here could be a mixed solution, yep: instead of usinf djview (which, btw, sucks), we can have java first to extract the pages and then to show them, so we have the update. Mike, can you upload your code? I have already a old example in bash, here in http://dftuz.unizar.es/~rivero/lbdjvu.tgz

[Mike Kostousov]

#include "config.h"
#include "DjVuGlobal.h"
#include "ZPCodec.h"
#include "GException.h"
#include "GSmartPointer.h"
#include "GURL.h"
#include "GRect.h"
#include "GPixmap.h"
#include "GBitmap.h"
#include "DjVuImage.h"
#include "DjVuDocument.h"
#include "GOS.h"
#include "ByteStream.h"
#include "DjVuMessage.h"
#include "DjVuText.h"
#include "ddjvuapi.h"


GP<DjVuDocument> doc;



int openFile(const char *fileName)
{
	djvu_programname("ddjvu");
	ddjvu_context_t * context = ddjvu_context_create("djview");	
	ddjvu_cache_set_size(context,10000000);
	printf("cache_size %ld\n",ddjvu_cache_get_size(context));
	if (doc) doc->stop_init();
	GURL gurl=GURL::Filename::UTF8(GNativeString(fileName));
	doc = DjVuDocument::create_wait(gurl);
	/*	doc = DjVuDocument::create_noinit();
		if (!doc) return 1;
		GURL gurl=GURL::Filename::UTF8(GNativeString(fileName.latin1()));
		doc->start_init(gurl);
	 */	ddjvu_context_release(context);
	return 1;
}

uint * getPageNum(int pageNum,int width,int height)
{
	printf("step 1");

	GP<DjVuImage> dimg;
	dimg = doc->get_page(pageNum);
	if (!dimg || ! dimg->wait_for_complete_decode()) { printf("error no dimg\n"); }
	DjVuInfo *info = dimg->get_info();
	int colorp = dimg->is_legal_photo();
	int blackp = dimg->is_legal_bilevel();
	int compoundp = dimg->is_legal_compound();
	if (!compoundp && !colorp && !blackp) 
	{ 
		DjVuPrintErrorUTF8("%s","Warning: This is not a well formed DjVu image\n");
		if (!info)  printf("error\n");
	}

	GP<GPixmap> pm;
	GP<GBitmap> bm;
	GRect renderrect(0,0,width/2,height/2);
	GRect fullrect(0,0,width/2,height/2);
	uint * buffer = new uint[width*height];


	pm = dimg->get_pixmap(renderrect, fullrect);

	if (!pm) bm = dimg->get_bitmap(renderrect, fullrect);
	unsigned int buff_row = width;
	unsigned int* buff_ptr = new uint[width*height*10];  
	unsigned int px,py;
	unsigned int w,h;
	printf("here");
	if (bm) {
		printf("BM!!!\n");

		int grays = bm->get_grays();
		unsigned char* row_ptr = (*bm)[bm->rows()-1];
		w=bm->columns();
		h=bm->rows();
		for (unsigned int i=0;i<w*h;i++) 
		{
			py=i/w;
			px=i%w;
			if (px==0)
			{
				row_ptr = (*bm)[bm->rows()-py-1];
				buff_ptr = buffer + buff_row*py;
			}
			int v = 255 - ((row_ptr[px] * 255) / (grays-1));
			buff_ptr[px] = (255<<24) | (v<<16) | (v<<8) | v;
		}

	} else
		if (pm) {
			printf("PM!!!\n");
			w=pm->columns();
			h=pm->rows();

			GPixel* row_ptr;
			for (unsigned int i=0;i<w*h;i++) {
				py=i/w;
				px=i%w;
				if (px==0) 
				{
					row_ptr=(*pm)[pm->rows()-1-py];
					buff_ptr = buffer + buff_row*py;
				}
				int r = row_ptr[px].r;
				int g = row_ptr[px].g;
				int b = row_ptr[px].b;
				buff_ptr[px] = (255<<24) | (r<<16) | (g<<8) | b;
			}
		}
	return buffer;

}

With compilation of djvulibre can be some problems. So, gcc parameters must be tuned and config.h also must be tuned

[scotty1024]

Well I know how to invalidate the e-ink display now.

So I ask: which djvu viewer is the best to start from?

You all let me know and I'll go modify it to work smoothly on the iLiad and post it here.

[arivero]

Originally Posted by scotty1024

So I ask: which djvu viewer is the best to start from?
.

The answer is: none.

You only have one compiling decently, djview. But on other hand it comes with a nice packaging including command line utilities, so for the time you have compiled it, you have also got the non-gui tools, particularly ddjvu

And then you find that djview depends of detecting the move of the mouse before click (no a great task for a wacom tablet, but unimplemented). And than you must rewire the keyboard to use iLiad keyboard. So perhaps it is more sensible to use you java trickies or Antartica tcl/tk (or iLiad tcl-tk?) instead of patching djview heavily.

Moreover, the ddjvu in the tar.gz above is already compiled for an ARM architecture with older dinamic libraries, so I would hope it to work in the iLiad with the more recent ones.

Given a page in $i, I use the following two command lines:

ddjvu ebook307.djvu -page=$i -size=600x800 -format=pgm -aspect=no | pnmdepth 255 > /tmp/page0.pgm
tail +4 /tmp/page0.pgm > /tmp/page1.raw

It is possible to get segments of a zoomed page:
ddjvu ebook307.djvu -page=$i -format=pgm -size=800x1000 -aspect=no -segment=800x600+0+400 > /tmp/page0a.pgm

It is possible to rotate using pnmtools (also in the package):
pnmdepth 255 /tmp/page0a.pgm | pamflip -transpose| pamflip -topbottom | tail +4 > /tmp/page1a.raw

These "raw" images were still 800x600x8bits, We had a custom utility to reduce them to 2 bits (pgm4biARM), but here in the iLiad it is not needed, the Xserver will take care off, you only need some java tool to show pgm files (so not | tail+4 neither).

There are some other tricks you can see in the bash script "lbdjbrowse". You have pnmpad, to put white space around a bad cropped page. pnmcrop to cut a preprocessed page, and pnmscalefixed to zoom or fit a page without cropping (fixed is because it uses not float point).


Let me note also xwdtopnm. Not related to djvu, but a nice tool to view screendumps of an Xserver. Also pbmtext is a stupid utility having java running and all that, but if you need to show an ASCII message in a graphical area, it is a nice shortcut to do it from bash.

Regretly I forgot to include pgm2jpg (?) or similar converters so you will still need them from debian netpbm