WHAT IS THIS
============

devanagari-from-pdf is a postprocessor to extract proper Sansk.rt (Sanskrit)
text from devanaagari PDF files WHEN those files have been properly created
with SHREE* fonts (see sample/ for an example of such file). Works as a
command-line filter.

Status
======

It is work in progress. But even now it properly converts over 99% of
syllables, even if text requires some verification and editing afterwards.

Usage
=====

0. Apply the following patch to poppler-utils' pdftotext so it would not
   convert U+00A0 (non-breaking space) to ASCII space in output. Unfortunately,
   U+00A0 is actually used in the SHREE fonts' encoding and the only workaround
   I could find so far is applying this patch and recompiling pdftotext.

diff --git a/poppler/TextOutputDev.cc b/poppler/TextOutputDev.cc
index 67a6246d..a60d9191 100644
--- a/poppler/TextOutputDev.cc
+++ b/poppler/TextOutputDev.cc
@@ -2656,7 +2656,7 @@ void TextPage::addChar(const GfxState *state, double x, double y, double dx, dou
	 }

	 // break words at space character
-    if (uLen == 1 && UnicodeIsWhitespace(u[0])) {
+    if (uLen == 1 && u[0] == ' ') { //UnicodeIsWhitespace(u[0])) {
		 charPos += nBytes;
		 endWord();
		 return;

1. Run pdftotext -layout filename.pdf
   This will generate filename.txt which is not really human-readable yet.

2. Run
   decode-shree-devanagari.py filename.txt > decoded.txt

decoded.txt is a Velthius-encoded text file. There are many ways to re-encode
Velthius to unicode devanaagari, itrans or other encodings online, this program
does not try to cover that.

TO DO (future plans)
====================

1. Complete support of all characters and their combinations as used in the PDF
   file in sample/.

2. Recode this python script as a proper part of poppler-utils' pdftotext.
   Bison and flex might be useful. Perhaps suggest an upstream patch when ready.

3. Properly handle mix of characters from non-devanaagari alphabets, e.g.:
	- 'x' in 1 x 100 on page 25 of vi1000 - govindAcArya [san].pdf
	- some kanna.da word on page 26 of the same PDF.
	This seems to be best doable from within pdftotext.