Use publish / subscribe architecture.

[tbnorth]

This is really just a placeholder for some research.

For the Leo Editor Pane development I've tested a publish / subscribe architecture, or signal / connect architecture. It includes an experimental emission of a signal from v.setBodyString(). I tested the performance impact with this code:

nd = p.v.insertAsLastChild()
import time
start = time.time()
for i in range(100000):
    new = nd.insertAsLastChild()
    new.b = 'test'
g.es(time.time()-start)
p.deleteAllChildren()

Twenty trials with and without the signal emission, mean 1.3 sec without, 1.4 seconds with, 8% slowdown. But I think it's important to emphasize that this is 8% on a very fast (76,000 nodes per second) operation. So it's not 8% on the load time for a file, it's 8% on a very short subset of the load time for a file.

The tests had really high variance (ranges of 0.99-2.91 and 1.07-2.38 respectively), which seems strange. Observed on two different machines - something to do with dynamic CPU clockspeed adjustment maybe?

[tbnorth]

For comparison, I think LeoPyRef.leo has only 7800 nodes, so a minuscule impact typically.

[edreamleo]

On Sun, Apr 2, 2017 at 3:46 PM, Terry Brown ***@***.***> wrote: This is really just a placeholder for some research.

​An excellent use of a git issue, imo.​

Twenty trials with and without the signal emission, mean 1.3 sec without, 1.4 seconds with, 8% slowdown. But I think it's important to emphasize that this is 8% on a very fast (76,000 nodes per second) operation. So it's not 8% on the load time for a file, it's 8% on a very short subset of the load time for a file.

​Seems fine to me. For more accurate results, consider using the timeit module <https://docs.python.org/2/library/timeit.html>. Somewhere there is a humorous quote that using timeit allows one to investigate code that will have negligible performance impact so that one can move on :-) Edward

[edreamleo]

While trying to find the quote, I ran across this stack overflow answer. I thought the fourth answer (starting with "I'll let you in on a secret: the best way to use timeit is on the command line.") was the most interesting/surprising.

[edreamleo]

This will not happen unless Terry does it.