A while ago I switched from Enthought's old EPD to their newer Canopy system. For the most part it's nice, but one aspect has been particularly vexing.
Whenever I run a python script, either from within the Canopy iPython environment or from the command line, none of my print statements actually get printed right away when that part of the script is hit. Instead, multiple prints seem to get executed all at once at a later time.
As an example...
import numpy as np
print "About to start long computation..."
a = np.random.randn(1e8)
print "Computation finished."
doesn't print the first statement until after a is finished being generated, when both statements are printed simultaneously. (You can tell when the calculation is occurring by watching the CPU monitor.)
Does anyone know what's going on here? If relevant, I'm running Canopy 1.0.0.1160, with Python 2.7.3 64bit on a Windows 7 machine.
No, this is not a change between EPD and Canopy. While I suppose there might be some python distributions which default to buffering off, EPD was not one of them -- the performance hit could have been too severe (as kindall's comment mentions.) Better to let the programmer decide when it's important for the user to see console output immediately (typically for status updates).
BTW, IPython in the Canopy GUI is simply IPython QtConsole. If you are depending on console I/O, you may also need to be aware of this longstanding issue with QtConsole:
I don't think there's a reasonable workaround for Canopy IPython, other than do it "properly", i.e. with flush.