#!/usr/bin/env python3
import argparse, json, os, subprocess, time
from pathlib import Path

def run(*args):
    return subprocess.check_output(args, text=True).strip()

ap = argparse.ArgumentParser()
ap.add_argument('action', choices=['snapshot', 'click', 'type', 'key'])
ap.add_argument('values', nargs='*')
ap.add_argument('--execute', action='store_true')
a = ap.parse_args()
wid = os.environ['KVM_WINDOW_ID']
if not wid.isdecimal():
    raise SystemExit('KVM_WINDOW_ID must be a numeric X11 window ID')
if os.environ.get('XDG_SESSION_TYPE') != 'x11':
    raise SystemExit('This example requires an X11 desktop session')
if a.action != 'snapshot' and not a.execute:
    print(json.dumps({'action': a.action, 'values': a.values, 'dry_run': True}))
    raise SystemExit(0)
run('xdotool', 'windowactivate', '--sync', wid)
time.sleep(0.5)
if a.action == 'click':
    if len(a.values) != 2: raise SystemExit('click requires X Y')
    x, y = map(int, a.values)
    geo = dict(line.split('=', 1) for line in
               run('xdotool', 'getwindowgeometry', '--shell', wid).splitlines())
    if not (0 <= x < int(geo['WIDTH']) and 0 <= y < int(geo['HEIGHT'])):
        raise SystemExit('Click is outside the viewer window')
    run('xdotool', 'mousemove', '--window', wid, str(x), str(y))
    run('xdotool', 'click', '1')
elif a.action == 'type':
    if len(a.values) != 1 or len(a.values[0]) > 200:
        raise SystemExit('type requires one string of at most 200 characters')
    run('xdotool', 'type', '--clearmodifiers', '--delay', '70', '--', a.values[0])
elif a.action == 'key':
    if len(a.values) != 1 or a.values[0] not in {'Return', 'Tab', 'Escape'}:
        raise SystemExit('Allowed keys: Return, Tab, Escape')
    run('xdotool', 'key', '--clearmodifiers', a.values[0])
elif a.values:
    raise SystemExit('snapshot takes no values')
time.sleep(1)
Path('shots').mkdir(exist_ok=True)
shot = Path('shots') / f'{time.time_ns()}.png'
run('scrot', '--focused', str(shot))
print(json.dumps({'screenshot': str(shot), 'action': a.action}))