#py3

Trac 開發版 1.5.1 對 Python 3 的說明

在 Trac 開發版 1.5.1 的 Change Log 裡可以看到說明：

This will be the only release in the 1.5.x release line that supports Python 2.7. Future releases will support Python 3.5+.

先前在 mailing list 上有看到 Python 3 的計畫，不過好像是第一次在 changelog 裡面看到說明了…

I have asked this question quite a few times in the past. Which would be better? List comprehension, or a map function to convert every item in a list or tuple to another?

Upon Googling, you will get very different answers from different people.

Let's try it with the timeit module.

".".join([str(x) for x in range(20)]) ".".join(map(str, range(20))

As you can see, both lines use range(20) to create a generator of 20 integer values. Then, either the map function, or the list comprehensor loops through the list converting each value to a string and returning the new value.

These are the results I found:

spike@Jacob-Laptop MINGW64 ~ $ python -mtimeit "'.'.join(map(str, range(20)))" 100000 loops, best of 3: 4.27 usec per loop spike@Jacob-Laptop MINGW64 ~ $ python -mtimeit "'.'.join([str(x) for x in range(20)])" 100000 loops, best of 3: 5.64 usec per loop

So which one of these is the best method?

Turns out, using map() was slightly faster. But, after trying this, I realised there was another option. Using a tuple directly. This should save the interpreter a little memory footprint, but in larger cases this would scale.

".".join(tuple(str(x) for x in range(20)))

spike@Jacob-Laptop MINGW64 ~ $ python -mtimeit "'.'.join(tuple(str(x) for x in range(20)))" 100000 loops, best of 3: 6.21 usec per loop

Wew lenny! (╯°□°) ╯︵ ┻━┻

That is a lot slower than I would have expected. Looks like map() is still king, in terms of speed.

All tests results were taken from timeit using best of three 100,000 loop tests.