#!/usr/bin/env bash
set -euo pipefail
BASE=/home/spazio/cosmos_books/talk-to-the-entities-beginning-manual
PDF="$BASE/source/TTTE_Beginning_Manual_AUG_2019_RUSSIAN.pdf"
IMG="$BASE/ocr_images"
PAGES="$BASE/text/pages"
FULL="$BASE/text/book_full.txt"
TESS=/home/spazio/.local/opt/tesseract-local/usr/bin/tesseract
export LD_LIBRARY_PATH=/home/spazio/.local/opt/tesseract-local/usr/lib/x86_64-linux-gnu
export TESSDATA_PREFIX=/home/spazio/.local/opt/tesseract-local/usr/share/tesseract-ocr/5/tessdata
mkdir -p "$IMG" "$PAGES"
: > "$FULL"
pdftoppm -jpeg -r 150 "$PDF" "$IMG/page"
for image in "$IMG"/page-*.jpg; do
  stem=$(basename "$image" .jpg)
  num=${stem#page-}
  "$TESS" "$image" "$PAGES/$stem" -l rus --psm 3 >/dev/null 2>&1
  printf '\n\n--- PAGE %s ---\n\n' "$num" >> "$FULL"
  cat "$PAGES/$stem.txt" >> "$FULL"
  printf 'OCR %s\n' "$num"
done
rm -rf "$IMG"
printf 'DONE\n'
