Ipython Notebook doesn't seem to print results in real time, but seems to buffer in a certain way and then bulk output the prints. How can I make ipython print my results as soon as the print command is processed?
Example code:
import time
def printer():
for i in range(100):
time.sleep(5)
print i
Supposing that the above code is in a file that is imported. How could I make it that when I call the printer function it prints a number every 5 seconds and not all the numbers at the very end?
Please note that I cannot edit the function printer()
because I get it from some external module. I want the to change the configs of ipython notebook somehow so that it doesn't use a buffer. Therefore, I also do not wish to use sys.stdout.flush(), I want to do it in real-time according to the question, I don't want any buffer to start with.
I also tried loading ipython notebook with the command:
ipython notebook --cache-size=0
but that also doesn't seem to work.
Since Python 3.3, print()
has an additional flush argument that can be used to force flushing:
for i in range(10):
print(i, flush=True)
time.sleep(1)
This is merely one of the answers to the question suggested by Carsten incorporating the __getattr__
delegation suggested by diedthreetimes in a comment:
import sys
oldsysstdout = sys.stdout
class flushfile():
def __init__(self, f):
self.f = f
def __getattr__(self,name):
return object.__getattribute__(self.f, name)
def write(self, x):
self.f.write(x)
self.f.flush()
def flush(self):
self.f.flush()
sys.stdout = flushfile(sys.stdout)
In the original answer, the __getattr__
method is not implemented. Without that, it fails. Other variants in answers to that question also fail in a notebook.
In a notebook, sys.stdout
is an instance of IPython.kernel.zmq.iostream.OutStream
and has a number of methods and attributes not present in the usual sys.stdout
. Delegating __getattr__
allows a flushfile
to masquerade as a ...zmq.iostream.OutStream
duck.
This works in a python 2.7 notebook run with ipython 3.1.0
Try this:
from IPython.display import display, clear_output
display("Hello World") # print string
display(df) # print object such as dataframe
clear_output(wait=True) # use this if need to clear the output before display, good for dynamic updates