1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
|
#!/usr/bin/env python3
"""A decorator to help with dead simple parallelization."""
from enum import Enum
import functools
import typing
class Method(Enum):
THREAD = 1
PROCESS = 2
REMOTE = 3
def parallelize(
_funct: typing.Optional[typing.Callable] = None,
*,
method: Method = Method.THREAD
) -> typing.Callable:
"""Usage:
@parallelize # defaults to thread-mode
def my_function(a, b, c) -> int:
...do some slow / expensive work, e.g., an http request
@parallelize(method=Method.PROCESS)
def my_other_function(d, e, f) -> str:
...do more really expensice work, e.g., a network read
@parallelize(method=Method.REMOTE)
def my_other_other_function(g, h) -> int:
...this work will be distributed to a remote machine pool
This decorator will invoke the wrapped function on:
Method.THREAD (default): a background thread
Method.PROCESS: a background process
Method.REMOTE: a process on a remote host
The wrapped function returns immediately with a value that is
wrapped in a SmartFuture. This value will block if it is either
read directly (via a call to result._resolve) or indirectly (by
using the result in an expression, printing it, hashing it,
passing it a function argument, etc...). See comments on the
SmartFuture class for details.
Note: you may stack @parallelized methods and it will "work".
That said, having multiple layers of Method.PROCESS or
Method.REMOTE may prove to be problematic because each process in
the stack will use its own independent pool which may overload
your machine with processes or your network with remote processes
beyond the control mechanisms built into one instance of the pool.
Be careful.
"""
def wrapper(funct: typing.Callable):
@functools.wraps(funct)
def inner_wrapper(*args, **kwargs):
import executors
import smart_future
# Look for as of yet unresolved arguments in _funct's
# argument list and resolve them now.
newargs = []
for arg in args:
newargs.append(smart_future.SmartFuture.resolve(arg))
newkwargs = {}
for kw in kwargs:
newkwargs[kw] = smart_future.SmartFuture.resolve(
kwargs[kw]
)
executor = None
if method == Method.PROCESS:
executor = executors.DefaultExecutors().process_pool()
elif method == Method.THREAD:
executor = executors.DefaultExecutors().thread_pool()
elif method == Method.REMOTE:
executor = executors.DefaultExecutors().remote_pool()
assert executor is not None
future = executor.submit(funct, *newargs, **newkwargs)
# Wrap the future that's returned in a SmartFuture object
# so that callers do not need to call .result(), they can
# just use is as normal.
return smart_future.SmartFuture(future)
return inner_wrapper
if _funct is None:
return wrapper
else:
return wrapper(_funct)
|
