python3 -c 'import random, time, itertools; any(time.sleep(0.01) or print(random.choice("\u2571\u2572"), end="", flush=True) for x in itertools.repeat(None))'
Quite a few bytes can be golfed out of that still:
python3 -c 'while I := __import__: I("time").sleep(0.01); print(I("random").choice("\u2571\u2572"), end="", flush=1)'
My reasoning including most failures:
Whitespace is mostly trivial and not worth mentioning, except that the space between "or" and "print" can be eliminated by moving the `time.sleep` to before a string literal. Alternatively, using `!=` works almost anywhere (though not with some alternative ideas), or `;` with the `while` version.
There are several shorter ways to get an infinite loop:
itertools.repeat(None) # for reference
itertools.repeat(0)
itertools.count()
iter(lambda: 0, 1) # also removes the import, so 10 chars shorter
iter(''.upper, 1) # same length when spaces removed, shorter if not removed
iter(int.mro, 1) # spooky, but one char shorter still
range(9**99) # much longer than the age of the universe
range(9**9) # 44 days
# but it turns out we can skip it entirely
The loop itself:
any(expr for x in it) # theoretically prevents the memory leak
[expr for x in it] # wastes 800 bytes per sec, 69 MB/day, 1 GB per 14 days
while 1: expr # breaks one-liner unless longer `__import__("")` is used, but worth it since it eliminates the entire iterable, even before doing I=__import__
while I := __import__: # 2 chars shorter than doing the assignment inside the loop
I looked into alternatives to calling sleep at all, but computers are too fast (and variable in speed) nowadays. `os.sync` looked promising but is only slow the first time. Trying to pass its return value as an argument also failed.
`flush=1` is shorter than `flush=True`. Otherwise ... I tried `sys.stdout` but hardly anything was even close:
any(time.sleep(0.01) or print(random.choice("\u2571\u2572"), end="", flush=True) for it)
any(time.sleep(0.01) or print(random.choice("\u2571\u2572"), end="", flush=1) for it)
sys.stdout.writelines(time.sleep(0.1) or random.choice("\u2571\u2572")+"\0"*8191 for it) # can avoid the flush by filling the buffer manually, but requires several more chars to import sys as well
Random:
random.choice("\u2571\u2572")
# UTF-8 b"\xe2\x95\xb1" doesn't seem useful
chr(9585+random.randrange(2)) # sigh
chr(9585+(random.random()<.5)) # nope
chr(9585+(os.getrandom(1)[0]&1)) # + has tigher precedence than &
# that last would save 1 byte with normal `import` (since the identifier is repeated), but loses with `__import__`
# is there a way to eliminate parens? Is >> or | or - useful? What about *splat?
https://github.com/the-real-tokai/macuahuitl/blob/master/tem...
(how about this fancy version with SVG output? :D No longer a single line though.)
Quite a few bytes can be golfed out of that still:
My reasoning including most failures:Whitespace is mostly trivial and not worth mentioning, except that the space between "or" and "print" can be eliminated by moving the `time.sleep` to before a string literal. Alternatively, using `!=` works almost anywhere (though not with some alternative ideas), or `;` with the `while` version.
There are several shorter ways to get an infinite loop:
The loop itself: I looked into alternatives to calling sleep at all, but computers are too fast (and variable in speed) nowadays. `os.sync` looked promising but is only slow the first time. Trying to pass its return value as an argument also failed.`flush=1` is shorter than `flush=True`. Otherwise ... I tried `sys.stdout` but hardly anything was even close:
Random:I used to do this on my VIC-20 a million years ago. Just looking at this line brought back the visual image like it was yesterday.