#!/usr/bin/env python3 # -*- coding: utf-8 -*- """Real Win32 input + DPI handling — the actual hands. This is the piece that was a `NotImplementedError` stub in the old code. Here it is a working SendInput implementation: absolute mouse move+click, unicode typing, named-key presses, and a mouse wheel scroll. Plus DPI awareness so a screenshot pixel and a click coordinate are the SAME point (the classic legacy-RPA bug). Everything is imported lazily and only ever runs on Windows. On any non-Windows host `is_windows()` is False and the action functions raise a clear error, so the module imports cleanly for inspection on Linux. Coordinate space: we drive the PRIMARY monitor and normalize against SM_CXSCREEN/SM_CYSCREEN, so click coords line up 1:1 with a primary-monitor ImageGrab screenshot (after set_dpi_aware()). Run the target on the primary display at 100% scale for the cleanest mapping; set_dpi_aware() makes 125%/150% report true pixels too, but 100% removes all doubt. """ from __future__ import annotations import sys import time IS_WINDOWS = sys.platform.startswith("win") def is_windows(): return IS_WINDOWS def _require_windows(): if not IS_WINDOWS: raise RuntimeError( "win_input action called on non-Windows host (sys.platform=%r). " "The implementation is real; it only runs on Windows." % sys.platform) # -------------------------------------------------------------------------- # DPI awareness # -------------------------------------------------------------------------- def set_dpi_aware(): """Make this process DPI-aware so screen capture and coordinates match. Tries Per-Monitor-V2 (Win10), then PROCESS_PER_MONITOR_DPI_AWARE (Win8.1), then the legacy SetProcessDPIAware. Safe to call once at startup; returns the method that succeeded, or 'none'. """ if not IS_WINDOWS: return "skipped-non-windows" import ctypes # Win10 1703+: SetProcessDpiAwarenessContext(-4) = PER_MONITOR_AWARE_V2 try: ctypes.windll.user32.SetProcessDpiAwarenessContext( ctypes.c_void_p(-4)) return "per-monitor-v2" except Exception: # noqa: BLE001 pass # Win8.1+: shcore.SetProcessDpiAwareness(2) = PROCESS_PER_MONITOR_DPI_AWARE try: ctypes.windll.shcore.SetProcessDpiAwareness(2) return "per-monitor" except Exception: # noqa: BLE001 pass try: ctypes.windll.user32.SetProcessDPIAware() return "system" except Exception: # noqa: BLE001 return "none" def screen_size(): """Primary-monitor size in physical pixels (after set_dpi_aware).""" _require_windows() import ctypes u = ctypes.windll.user32 return int(u.GetSystemMetrics(0)), int(u.GetSystemMetrics(1)) # SM_CXSCREEN/CY # -------------------------------------------------------------------------- # SendInput structures # -------------------------------------------------------------------------- def _structs(): import ctypes from ctypes import wintypes ULONG_PTR = ctypes.POINTER(ctypes.c_ulong) class MOUSEINPUT(ctypes.Structure): _fields_ = [("dx", wintypes.LONG), ("dy", wintypes.LONG), ("mouseData", wintypes.DWORD), ("dwFlags", wintypes.DWORD), ("time", wintypes.DWORD), ("dwExtraInfo", ULONG_PTR)] class KEYBDINPUT(ctypes.Structure): _fields_ = [("wVk", wintypes.WORD), ("wScan", wintypes.WORD), ("dwFlags", wintypes.DWORD), ("time", wintypes.DWORD), ("dwExtraInfo", ULONG_PTR)] class _INPUTunion(ctypes.Union): _fields_ = [("mi", MOUSEINPUT), ("ki", KEYBDINPUT)] class INPUT(ctypes.Structure): _fields_ = [("type", wintypes.DWORD), ("u", _INPUTunion)] return ctypes, INPUT, MOUSEINPUT, KEYBDINPUT # event flags _INPUT_MOUSE = 0 _INPUT_KEYBOARD = 1 _M_MOVE = 0x0001 _M_ABSOLUTE = 0x8000 _M_LEFTDOWN = 0x0002 _M_LEFTUP = 0x0004 _M_WHEEL = 0x0800 _K_UNICODE = 0x0004 _K_KEYUP = 0x0002 _K_EXTENDED = 0x0001 # named virtual-key codes (extended where relevant) _VK = { "enter": 0x0D, "return": 0x0D, "tab": 0x09, "escape": 0x1B, "esc": 0x1B, "space": 0x20, "backspace": 0x08, "delete": 0x2E, "del": 0x2E, "home": 0x24, "end": 0x23, "pageup": 0x21, "pagedown": 0x22, "up": 0x26, "down": 0x28, "left": 0x25, "right": 0x27, "f1": 0x70, "f2": 0x71, "f3": 0x72, "f4": 0x73, "f5": 0x74, "f6": 0x75, "f7": 0x76, "f8": 0x77, "f9": 0x78, "f10": 0x79, "f11": 0x7A, "f12": 0x7B, } _VK_EXTENDED = {0x2E, 0x24, 0x23, 0x21, 0x22, 0x26, 0x28, 0x25, 0x27} def _send(inputs): ctypes, INPUT, _MI, _KI = _structs() n = len(inputs) arr = (INPUT * n)(*inputs) sent = ctypes.windll.user32.SendInput(n, arr, ctypes.sizeof(INPUT)) if sent != n: raise RuntimeError("SendInput sent %d/%d events (input blocked? UAC-" "elevated target needs elevated python)" % (sent, n)) return sent def _abs_xy(x, y): sw, sh = screen_size() # normalize to 0..65535 over the primary screen nx = int(round(x * 65535.0 / max(1, sw - 1))) ny = int(round(y * 65535.0 / max(1, sh - 1))) return max(0, min(65535, nx)), max(0, min(65535, ny)) # -------------------------------------------------------------------------- # Public actions # -------------------------------------------------------------------------- def move_click(x, y, settle=0.12): """Absolute move to (x,y) then a left click.""" _require_windows() ctypes, INPUT, MOUSEINPUT, _KI = _structs() nx, ny = _abs_xy(x, y) def mouse(flags): mi = MOUSEINPUT(nx, ny, 0, flags | _M_ABSOLUTE, 0, None) inp = INPUT(); inp.type = _INPUT_MOUSE; inp.u.mi = mi return inp _send([mouse(_M_MOVE)]) time.sleep(0.03) _send([mouse(_M_LEFTDOWN), mouse(_M_LEFTUP)]) time.sleep(settle) def type_unicode(text): """Type a unicode string into the focused control via KEYEVENTF_UNICODE.""" _require_windows() ctypes, INPUT, _MI, KEYBDINPUT = _structs() inputs = [] for ch in str(text): code = ord(ch) down = KEYBDINPUT(0, code, _K_UNICODE, 0, None) up = KEYBDINPUT(0, code, _K_UNICODE | _K_KEYUP, 0, None) for ki in (down, up): inp = INPUT(); inp.type = _INPUT_KEYBOARD; inp.u.ki = ki inputs.append(inp) if inputs: _send(inputs) time.sleep(0.05) def press_key(name): """Press a named key (enter/tab/escape/arrows/f-keys/...).""" _require_windows() ctypes, INPUT, _MI, KEYBDINPUT = _structs() vk = _VK.get(str(name).lower()) if vk is None: # single printable char -> route through unicode typing if len(str(name)) == 1: return type_unicode(name) raise RuntimeError("unknown key name %r" % name) flags_down = _K_EXTENDED if vk in _VK_EXTENDED else 0 down = KEYBDINPUT(vk, 0, flags_down, 0, None) up = KEYBDINPUT(vk, 0, flags_down | _K_KEYUP, 0, None) seq = [] for ki in (down, up): inp = INPUT(); inp.type = _INPUT_KEYBOARD; inp.u.ki = ki seq.append(inp) _send(seq) time.sleep(0.05) def scroll(amount_notches, x=None, y=None): """Mouse wheel scroll. Positive = up, negative = down. Optionally move to (x,y) first so the scroll lands on the intended pane.""" _require_windows() ctypes, INPUT, MOUSEINPUT, _KI = _structs() if x is not None and y is not None: nx, ny = _abs_xy(x, y) mi = MOUSEINPUT(nx, ny, 0, _M_MOVE | _M_ABSOLUTE, 0, None) inp = INPUT(); inp.type = _INPUT_MOUSE; inp.u.mi = mi _send([inp]); time.sleep(0.03) delta = int(amount_notches) * 120 # WHEEL_DELTA mi = MOUSEINPUT(0, 0, delta & 0xFFFFFFFF, _M_WHEEL, 0, None) inp = INPUT(); inp.type = _INPUT_MOUSE; inp.u.mi = mi _send([inp]) time.sleep(0.1)