#!/bin/sh
# OCR a screen region: slurp -> grim -> tesseract -> clipboard, with a notification showing what was read.
# usage: ocr [lang]   (tesseract language, default eng; e.g. `ocr eng+hin`)
# slurp takes aurora's matugen colours when they're there.

lang="${1:-eng}"
c="${XDG_CACHE_HOME:-$HOME/.cache}/matugen/aurora.json"

if command -v jq >/dev/null && [ -r "$c" ]; then
    bg=$(jq -r '.surfaceLowest // empty' "$c")
    fg=$(jq -r '.primary // empty' "$c")
fi
if [ -n "$bg" ] && [ -n "$fg" ]; then
    g=$(slurp -b "${bg}88" -c "${fg}ff" -s "#00000000" -w 1) || exit 0
else
    g=$(slurp) || exit 0
fi

text=$(grim -g "$g" - | tesseract stdin stdout -l "$lang" 2>/dev/null | sed -e 's/[[:space:]]*$//' -e '/./,$!d')
if [ -z "$text" ]; then
    notify-send -a ocr "OCR" "no text found"
    exit 1
fi
printf '%s' "$text" | wl-copy
notify-send -a ocr "OCR · copied" "$(printf '%s' "$text" | head -c 200)"