feat(actions): implement ActionExecutor with xdotool and pyautogui backends

This commit is contained in:
pyr0ball 2026-04-26 21:20:06 -07:00
parent 0dcc25164d
commit 51a22cbe47
2 changed files with 144 additions and 0 deletions

View file

@ -0,0 +1,96 @@
"""
ActionExecutor translate action name strings into OS input events.
Backends:
xdotool Linux X11; requires xdotool package installed
pyautogui cross-platform fallback
auto detects xdotool at init, falls back to pyautogui
"""
from __future__ import annotations
import logging
import os
import shutil
import subprocess
from typing import Literal
# Suppress pyautogui DISPLAY errors on headless systems
if not os.environ.get("DISPLAY"):
os.environ["DISPLAY"] = ":0"
import pyautogui
logger = logging.getLogger(__name__)
Backend = Literal["xdotool", "pyautogui", "auto"]
_XDOTOOL_CLICK: dict[str, list[str]] = {
"left_click": ["xdotool", "click", "1"],
"right_click": ["xdotool", "click", "3"],
"double_click": ["xdotool", "click", "--repeat", "2", "1"],
"middle_click": ["xdotool", "click", "2"],
"scroll_up": ["xdotool", "click", "4"],
"scroll_down": ["xdotool", "click", "5"],
}
_XDOTOOL_KEY: dict[str, str] = {
"key_enter": "Return",
"key_escape": "Escape",
"key_space": "space",
"key_tab": "Tab",
}
class ActionExecutor:
def __init__(self, backend: Backend = "auto") -> None:
if backend == "auto":
self._backend: Backend = (
"xdotool" if shutil.which("xdotool") else "pyautogui"
)
else:
self._backend = backend
def execute(self, action: str) -> None:
"""Execute a named action. Silently ignores unknown action names."""
try:
if self._backend == "xdotool":
self._xdotool(action)
else:
self._pyautogui(action)
except Exception as exc:
logger.warning("ActionExecutor error for action=%r: %s", action, exc)
def _xdotool(self, action: str) -> None:
if action in _XDOTOOL_CLICK:
subprocess.run(_XDOTOOL_CLICK[action], capture_output=True)
elif action in _XDOTOOL_KEY:
subprocess.run(
["xdotool", "key", _XDOTOOL_KEY[action]], capture_output=True
)
elif action.startswith("drag_start"):
subprocess.run(["xdotool", "mousedown", "1"], capture_output=True)
elif action.startswith("drag_end"):
subprocess.run(["xdotool", "mouseup", "1"], capture_output=True)
else:
logger.warning("xdotool: unknown action %r", action)
def _pyautogui(self, action: str) -> None:
if action == "left_click":
pyautogui.click()
elif action == "right_click":
pyautogui.rightClick()
elif action == "double_click":
pyautogui.doubleClick()
elif action == "scroll_up":
pyautogui.scroll(3)
elif action == "scroll_down":
pyautogui.scroll(-3)
elif action == "key_enter":
pyautogui.press("enter")
elif action == "key_escape":
pyautogui.press("escape")
elif action == "key_space":
pyautogui.press("space")
else:
logger.warning("pyautogui: unknown action %r", action)

View file

@ -0,0 +1,48 @@
from unittest.mock import patch, MagicMock
from merlin.actions.executor import ActionExecutor
@patch("merlin.actions.executor.subprocess")
def test_left_click_calls_xdotool(mock_sub):
ex = ActionExecutor(backend="xdotool")
ex.execute("left_click")
mock_sub.run.assert_called_once_with(["xdotool", "click", "1"], capture_output=True)
@patch("merlin.actions.executor.subprocess")
def test_right_click_calls_xdotool(mock_sub):
ex = ActionExecutor(backend="xdotool")
ex.execute("right_click")
mock_sub.run.assert_called_once_with(["xdotool", "click", "3"], capture_output=True)
@patch("merlin.actions.executor.subprocess")
def test_double_click_calls_xdotool(mock_sub):
ex = ActionExecutor(backend="xdotool")
ex.execute("double_click")
mock_sub.run.assert_called_once_with(
["xdotool", "click", "--repeat", "2", "1"], capture_output=True
)
@patch("merlin.actions.executor.subprocess")
def test_key_enter(mock_sub):
ex = ActionExecutor(backend="xdotool")
ex.execute("key_enter")
mock_sub.run.assert_called_once_with(
["xdotool", "key", "Return"], capture_output=True
)
@patch("merlin.actions.executor.subprocess")
def test_unknown_action_does_not_raise(mock_sub):
ex = ActionExecutor(backend="xdotool")
ex.execute("nonexistent_action")
mock_sub.run.assert_not_called()
@patch("merlin.actions.executor.pyautogui")
def test_pyautogui_fallback_left_click(mock_pag):
ex = ActionExecutor(backend="pyautogui")
ex.execute("left_click")
mock_pag.click.assert_called_once()