This is a line I read from a text file:
[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]
I used readline() to read it in as a string. Now what is the fastest way to convert it back to an array?
Thank you!
4 Answers
I'm not sure that this is the fastest, but it's definitely the safest/easiest:
import ast
lst = ast.literal_eval(s)
regular eval would work too:
lst = eval(s)
Some basic timings from my machine:
>>> s = '[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]'
>>> def f1():
... eval(s)
...
>>> def f2():
... ast.literal_eval(s)
...
>>> timeit.timeit('f1()', 'from __main__ import f1')
31.415852785110474
>>> timeit.timeit('f2()', 'from __main__ import f2')
46.25958704948425
So, according to my computer, eval is about 50% faster than ast.literal_eval. However, eval is terribly unsafe and should never be used on any string unless you trust it completely. Unless this is a real demonstratable bottleneck and you trust the input 100%, I would consider the little bit of extra time worth it in exchange for being able to sleep soundly at night.
7 Comments
Tim Peters
And by far the safest.
eval(s) may be the fastest.
mgilson
@JonathonReinhart -- Yeah, seriously. I'm not sure I've ever seen it go up that fast.
Jonathon Reinhart
@mgilson That's what she said.
mgilson
@TimPeters -- You're correct.
timeit shows that eval is about 50% faster on my machine.
Loïc Faure-Lacroix
This use of eval is disturbing.
|
Since we care about speed, in this particular case I might use json.loads:
>>> import ast, json
>>> s = "[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]"
>>> %timeit ast.literal_eval(s)
10000 loops, best of 3: 61.6 µs per loop
>>> %timeit eval(s)
10000 loops, best of 3: 45.7 µs per loop
>>> %timeit json.loads(s)
100000 loops, best of 3: 6.61 µs per loop
>>> json.loads(s)
[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]
(Note that this works here because this line is sufficiently JSON-like. It can't be used everywhere that ast.literal_eval can because not all Python literal syntax is valid JSON.)
3 Comments
Tim Peters
Cool! I wouldn't have guessed that - impressive :-)
mgilson
Sure DSM... I post a good answer and then you just have to go and trump it. Kudos to you (and since Kudos aren't an official unit of StackOverflow currency, +1 along with it).
Loïc Faure-Lacroix
I prefer the solution with json as using eval could end terribly wrong.
If that's the string, go here http://docs.python.org/2/library/functions.html#eval
>>> s = "[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]"
>>> eval(s)
[54, 95, 45, -97, -51, 84, 0, 32, -55, 14, 50, 54, 68, -3, 57, 88, -1]
2 Comments
Giupo
@ᴋᴇʏsᴇʀ , I accept the suggestion, but I disagree on explaining
eval. It's there, in the official python docs I posted, and IMHO it's more instructive to go there and read instead of repeating a piece of information here (and it's more "pythonic" too.)
keyser
I'm not saying it's a bad link :p (it most definitely isn't), I'm just saying you should have at least once sentence of meta-text :)
I'm using python 3.6
Here's my result of an actual string separated by spaces
stringValue = "123 456 789 012 345 678"
String to List
intValue = list(map(int, stringValue.split(' ')))
result
time taken 7.586999345221557e-06
[76, 96, 127, 255, 136, 164]
time taken trying to print it 6.697199933114462e-05
String to List to numpy
intValue = np.array(list(map(int, stringValue.split(' '))))
result
time taken 2.631999996083323e-05
[ 76 96 127 255 136 164]
time taken trying to print it 0.002241893000245909
String to numpy array
intValue = np.fromstring(stringValue,dtype=int,sep=' ')
time taken 1.3830000170855783e-05
[ 76 96 127 255 136 164]
time taken trying to print it 0.0003395890007595881
list, not anarray. There are subtle differences...