问题
I'm trying to create a looping python function which performs a task and prompts the user for a response and if the user does not respond in the given time the sequence will repeat.
This is loosely based off this question: How to set time limit on raw_input
The task is represented by some_function()
. The timeout is a variable in seconds. I have two problems with the following code:
The raw_input prompt does not timeout after the specified time of 4 seconds regardless of whether the user prompts or not.
When raw_input of 'q' is entered (without '' because I know anything typed is automatically entered as a string) the function does not exit the loop.
`
import thread
import threading
from time import sleep
def raw_input_with_timeout():
prompt = "Hello is it me you're looking for?"
timeout = 4
astring = None
some_function()
timer = threading.Timer(timeout, thread.interrupt_main)
try:
timer.start()
astring = raw_input(prompt)
except KeyboardInterrupt:
pass
timer.cancel()
if astring.lower() != 'q':
raw_input_with_timeout()
else:
print "goodbye"
`
回答1:
Warning: This is intended to work in *nix and OSX as requested but definitely will not work in Windows.
I've used this modification of an ActiveState recipe as a basis for the code below. It's an easy-to-use object that can read input with a timeout. It uses polling to collect characters one at a time and emulate the behavior of raw_input()
/ input()
.
Input with Timeout
Note: apparently the _getch_nix()
method below doesn't work for OP but it does for me on OSX 10.9.5. You might have luck calling _getch_osx()
instead although it seems to work in 32-bit python only since Carbon doesn't fully support 64-bit.
import sys
import time
class TimeoutInput(object):
def __init__(self, poll_period=0.05):
import sys, tty, termios # apparently timing of import is important if using an IDE
self.poll_period = poll_period
def _getch_nix(self):
import sys, tty, termios
from select import select
fd = sys.stdin.fileno()
old_settings = termios.tcgetattr(fd)
try:
tty.setraw(sys.stdin.fileno())
[i, o, e] = select([sys.stdin.fileno()], [], [], self.poll_period)
if i:
ch = sys.stdin.read(1)
else:
ch = ''
finally:
termios.tcsetattr(fd, termios.TCSADRAIN, old_settings)
return ch
def _getch_osx(self):
# from same discussion on the original ActiveState recipe:
# http://code.activestate.com/recipes/134892-getch-like-unbuffered-character-reading-from-stdin/#c2
import Carbon
if Carbon.Evt.EventAvail(0x0008)[0] == 0: # 0x0008 is the keyDownMask
return ''
else:
# The event contains the following info:
# (what,msg,when,where,mod)=Carbon.Evt.GetNextEvent(0x0008)[1]
#
# The message (msg) contains the ASCII char which is
# extracted with the 0x000000FF charCodeMask; this
# number is converted to an ASCII character with chr() and
# returned
(what,msg,when,where,mod)=Carbon.Evt.GetNextEvent(0x0008)[1]
return chr(msg & 0x000000FF)
def input(self, prompt=None, timeout=None,
extend_timeout_with_input=True, require_enter_to_confirm=True):
"""timeout: float seconds or None (blocking)"""
prompt = prompt or ''
sys.stdout.write(prompt) # this avoids a couple of problems with printing
sys.stdout.flush() # make sure prompt appears before we start waiting for input
input_chars = []
start_time = time.time()
received_enter = False
while (time.time() - start_time) < timeout:
# keep polling for characters
c = self._getch_osx() # self.poll_period determines spin speed
if c in ('\n', '\r'):
received_enter = True
break
elif c:
input_chars.append(c)
sys.stdout.write(c)
sys.stdout.flush()
if extend_timeout_with_input:
start_time = time.time()
sys.stdout.write('\n') # just for consistency with other "prints"
sys.stdout.flush()
captured_string = ''.join(input_chars)
if require_enter_to_confirm:
return_string = captured_string if received_enter else ''
else:
return_string = captured_string
return return_string
Test it
# this should work like raw_input() except it will time out
ti = TimeoutInput(poll_period=0.05)
s = ti.input(prompt='wait for timeout:', timeout=5.0,
extend_timeout_with_input=False, require_enter_to_confirm=False)
print(s)
Repeated Input
This implements your original intention as I understand it. I don't see any value to making recursive calls - I think what you want is just to get input repeatedly? Please correct me if that is wrong.
ti = TimeoutInput()
prompt = "Hello is it me you're looking for?"
timeout = 4.0
while True:
# some_function()
s = ti.input(prompt, timeout)
if s.lower() == 'q':
print "goodbye"
break
回答2:
You can set an alarm before input and then bind the alarm to a custom handler.
after the given period alarms goes off, handler raises an exception, and your custom input
function may handle the rest.
a quick example:
import signal
class InputTimedOut(Exception):
pass
def inputTimeOutHandler(signum, frame):
"called when read times out"
print 'interrupted!'
raise InputTimedOut
signal.signal(signal.SIGALRM, inputTimeOutHandler)
def input_with_timeout(timeout=0):
foo = ""
try:
print 'You have {0} seconds to type in your stuff...'.format(timeout)
signal.alarm(timeout)
foo = raw_input()
signal.alarm(0) #disable alarm
except InputTimedOut:
pass
return foo
s = input_with_timeout(timeout=3)
print 'You typed', s
Credit where it is due: Keyboard input with timeout in Python
回答3:
I do not think that there is a way to show a prompt that will expire after time passes without displaying a different message from another thread.
You can add the following line before the call to raw_input:
thread.start_new_thread(interrupt_user,())
You can define the interrupt_user
function as follows:
sleep(5)
print "\nTime up"
In the raw_input_with_time
function, do not call sleep. Instead, save the time from before the call to raw_input, and determine if the elapsed time after the call is more than 5 seconds. Also, if the user entered 'q' then it should not call itself so the loop will stop.
回答4:
Another way of doing it is to place the IO blocking in the new thread (as opposed to your proposed scheme where you have it in your main thread). The caveat for this is that there is not a clean way of killing a thread in python, so this does not play nice with repeating calls (N threads will hang around until main ends, and I think raw_input does not play nice...).
So, be warned, this works once, far from perfect solution
import threading
import Queue
def threaded_raw_input(ret_queue):
print("In thread")
prompt = "Hello is it me you're looking for?"
astring = raw_input(prompt)
ret_queue.put(astring)
if __name__ == '__main__':
print("Main")
ret_queue = Queue.Queue()
th = threading.Thread(target=threaded_raw_input, args=(ret_queue,))
th.daemon = True
th.start()
try:
astring = ret_queue.get(timeout=4)
except Queue.Empty:
print("\nToo late")
else:
print("Your input {}".format(astring))
回答5:
This is just prof of concept. Asking user for input data.
import time, os
import curses
def main(win):
win.nodelay(True)
x=0
output=""
while 1:
win.clear()
win.addstr(str("Prompt:"))
win.addstr(str(output))
x+=1
try:
key = win.getkey()
if key == os.linesep:
return output
output += str(key)
x = 0
except:
pass
if x>=50: # 5s
return output
time.sleep(0.1)
curses.wrapper(main)
回答6:
What if instead of calling some_function
when the input times out, you turn that into a background thread that keeps going with an interval of the timeout? The work will keep going while the main thread is permanently blocked on waiting for input. You may decide to react differently to that input based on what the worker is doing (working or sleeping) - you might just completely ignore it. AFAIK, there is no noticeable difference between not taking input or taking input but ignoring it. This idea leverages that.
Note: All I intend to do is to demonstrate another way of thinking about the problem that may or may not be appropriate in your particular case. I do think it is very flexible though.
Proof of concept:
from __future__ import print_function
from threading import Event, Thread
from time import sleep
def some_function():
print("Running some function")
sleep(1)
def raw_input_with_timeout():
cancel_event = Event()
wip_event = Event() # Only needed to know if working or waiting
def worker():
timeout = 4
try:
while not cancel_event.is_set():
wip_event.set()
some_function()
print("Repeating unless 'q' is entered within %d secs!" % timeout)
wip_event.clear()
cancel_event.wait(timeout)
finally:
wip_event.clear()
worker_thread = Thread(target=worker)
worker_thread.start()
try:
while not cancel_event.is_set():
try:
if raw_input() == 'q' and not wip_event.is_set():
cancel_event.set()
except KeyboardInterrupt:
pass
finally:
cancel_event.set()
worker_thread.join()
print("Goodbye")
It doesn't rely on anything platform specific; it's just simple Python code. Only after trying some alternative implementations taking input from within a thread, I realized how much of an advantage leaving user input to the main thread is.
I didn't pay too much attention to making it safe and clean, but for sure it can be done while keeping the overall structure. The biggest flaw I can see with it is that earlier input is never going away. It causes confusion when the worker outputs, obscuring the earlier input. If you press q
but not Enter
in time, pressing q
and Enter
next time results in the input of qq
even when those q
s are not next to each other on the screen. Generally this is how command line applications work, so I'm not sure if it's worth fixing. You might consider accepting input consisting only of q
s as cancelation as well. Another option would be to read from stdin
directly, not using raw_input
.
Some idea to make the structure of the code nicer would be to make the main thread even dumber and have it pass all input to the worker (using a queue) to decide what to do with.
来源:https://stackoverflow.com/questions/32193435/user-input-with-a-timeout-in-a-loop