brendon zhang added the comment:
did you remember to run `ulimit -S -s unlimited`? If you don't increase the
soft stack limit, your OS will kill the program.
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
reduce the upper bound for the recursion depth until it does not segfault
eg
hi = 12
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
No, the example you provided does not trigger the same bug. It is explicitly to
do with calling recursively from within the generator expression.
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
hmm, I can't edit my previous posts. I meant to say
python3.6 consumerbug.py
python3.7 consumerbug.py
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
Okay, I attached a complete minimal example below.
In your shell, run
ulimit -S -s unlimited
Then try executing with various python versions 3.6 and 3.7
python3.6 benchbug.py
python3.7 benchbug.py
You will notice that python 3.7 has a significant
Change by brendon zhang :
--
nosy: +rhettinger, vstinner
___
Python tracker
<https://bugs.python.org/issue40225>
___
___
Python-bugs-list mailing list
Unsub
brendon zhang added the comment:
update 2:
This affects ALL functions which exhaust a generator expression. If that
generator expression makes a recursive call, then the cost of evaluating it is
O(depth), when it should be only O(1).
You can see demonstrate that this doesn't just affect
Change by brendon zhang :
--
title: generator exhaustion for builtin functions in nested call is O(depth) ->
recursive call within generator expression is O(depth)
___
Python tracker
<https://bugs.python.org/issu
Change by brendon zhang :
--
title: generator exhaustion for builtin functions in recursion is O(depth) ->
generator exhaustion for builtin functions in nested call is O(depth)
___
Python tracker
<https://bugs.python.org/issu
Change by brendon zhang :
--
title: max() performance regression (quadratic time) -> generator exhaustion
for builtin functions in recursion is O(depth)
___
Python tracker
<https://bugs.python.org/issu
brendon zhang added the comment:
this affects ALL builtin functions (eg all(), any(), sum(), sorted(), etc...)
that accept generator as input and exhaust it.
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
update:
it is specifically caused by passing in a generator expression to max(), where
the generator invokes recursive function.
I added another file to demonstrate this
--
Added file: https://bugs.python.org/file49044/maxbug2.py
brendon zhang added the comment:
You can get the replicate this issue even when removing lru_cache(None) and
calling max with iterable of size 1.
eg.
best = max(solve(j) for j in [i-1])
--
___
Python tracker
<https://bugs.python.org/issue40
brendon zhang added the comment:
Something about calling max() in deeply nested recursion context appears to
make the overall complexity O(n^2) instead of O(n)
--
___
Python tracker
<https://bugs.python.org/issue40
New submission from brendon zhang :
There is a performance regression of the max (and also min) function
implementation starting in python 3.7.
I provide code and associated benchmarks in the file attachment.
--
components: Library (Lib)
files: maxbug.py
messages: 365978
nosy
brendon zhang added the comment:
np!
--
___
Python tracker
<https://bugs.python.org/issue40114>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by brendon zhang :
--
resolution: -> rejected
stage: patch review -> resolved
status: open -> closed
___
Python tracker
<https://bugs.python.or
brendon zhang added the comment:
"lru_cache(maxsize=128, typed=False)(user_function)"
oh ok! I overlooked that you could do this. There isn't really a need then for
supporting user_function w/ the maxsize argument in single call.
"The code in the PR is valiant attempt but is
brendon zhang added the comment:
update:
ignore concern #1. I used __text_signature__ attribute to reset the inspect
signature to something useful
https://github.com/python/cpython/pull/19226/commits/eea367f64d2a83d0987a3f7a155ac015306e960b
brendon zhang added the comment:
Two concerns is
1. the method signature (inspect.signature(lru_cache)) is not (maxsize=128,
typed=False) anymore, and so it is poor documentation.
2. not good reuse of TypeErrors related to incorrect number of inputs / type of
inputs
New submission from brendon zhang :
Existing implementation of lru_cache user_function option supports typed_code
argument, but does not support maxsize argument, because when we pass in the
user function as a positional argument, it gets assigned to the maxsize
parameter, and therefore
Change by brendon zhang :
--
keywords: +patch
pull_requests: +18588
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/19226
___
Python tracker
<https://bugs.python.org/issu
Change by brendon zhang :
--
components: Library (Lib)
nosy: brendon-zh...@hotmail.com, rhettinger
priority: normal
severity: normal
status: open
title: support maxsize argument for lru_cache's user_function option
type: behavior
versions: Python 3.8
23 matches
Mail list logo
