Python CLI Unbuffered 19 Sep 2014
Suppose you just inherited a python script that does some complex/expensive task and prints logging information to STDOUT. I say inherited because if you are writing your own script to do anything, you should really be using the logging module instead of just printing stuff to the terminal.
Now printing messages to directly to the terminal has a few disadvantages, but that’s beyond the scope of this article, and the bottom line is that your have a python script and you want to redirect its output to a file for later processing or just because you don’t want to risk losing part of the information because it’s larger than your terminal buffer. One thing you can do is use tee for that:
python script.py | tee -a script.log
While that works fine, you may run into the following inconvenience: if
your script does not output a lot of text, it’ll get buffered and you’ll
only see it in chunks. Python does that for you so that it won’t be
really slow if you’re writing a LOT of text to STDOUT
. :D
You can see the buffering result by using the following code in your
script.py
and redirecting the output to script.log
as showed above.
import time
for i in range(10):
print i
time.sleep(5)
You’ll notice that you have to wait the full 50 seconds to see anything on the screen, which is not ideal for such a short amount of text. So, if you know you’re not printing a lot of text and you need immediate feedback, you can bypass that:
python -u script.py | tee -a script.log
That makes python treat STDOUT
, STDERR
, and STDIN
completely
unbuffered. This will significantly slow down your execution if you’re
printing a lot of text, but will come in really handy if the amount of
text is not too high and you need to see what the script is doing right
away.