functools.lru_cache user specified cachedict support

[wdv4758h]

BPO	26082
Nosy	@rhettinger, @MojoVampire
Files	functools.lru_cache-user-specified-cachedict.patch: add user specified cachedict support

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/rhettinger'
closed_at = <Date 2017-04-25.14:36:07.821>
created_at = <Date 2016-01-11.19:23:57.160>
labels = ['type-feature', 'library']
title = 'functools.lru_cache user specified cachedict support'
updated_at = <Date 2017-04-25.14:36:07.820>
user = 'https://bugs.python.org/wdv4758h'

bugs.python.org fields:

activity = <Date 2017-04-25.14:36:07.820>
actor = 'rhettinger'
assignee = 'rhettinger'
closed = True
closed_date = <Date 2017-04-25.14:36:07.821>
closer = 'rhettinger'
components = ['Library (Lib)']
creation = <Date 2016-01-11.19:23:57.160>
creator = 'wdv4758h'
dependencies = []
files = ['41584']
hgrepos = []
issue_num = 26082
keywords = ['patch']
message_count = 3.0
messages = ['258001', '258049', '292261']
nosy_count = 3.0
nosy_names = ['rhettinger', 'josh.r', 'wdv4758h']
pr_nums = []
priority = 'normal'
resolution = 'rejected'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue26082'
versions = ['Python 3.6']

[wdv4758h]

Currently, lru_cache will automatically construct a Python dictionary in the function as cachedict. IMHO, it will be much more flexible to let users specified their cachedict, so they can use any kind of dict-like calss as their cachedict. Thus, users can use any dictionary implementation and save result in any form they want.

for example :

use OrderedDict

.. code-block:: python

    from functools import lru_cache
    from collections import OrderedDict

    @lru_cache(maxsize=None, cache=OrderedDict())
    def func(*args, **kwargs):
        pass

save by pickle

.. code-block:: python

    import os
    import pickle
    from functools import lru_cache

    filename = "cache.pickle"
    cache = {}

    def load_cache():
        global cache
        if os.path.isfile(filename):
            with open(filename, "rb") as f:
                cache = pickle.load(f)

    def store_cache():
        with open(filename, "wb") as f:
            pickle.dump(cache, f)

    load_cache()

    @lru_cache(maxsize=None, cache=cache)
    def func(*args, **kwargs):
        pass

[MojoVampire]

Given that lru_cache uses the cache dict in very specific ways, supporting arbitrary mapping types would be extremely hard. Among other things:

1. The C code uses the concrete dict APIs (including private APIs) that would not work on arbitrary mappings that don't directly inherit from dict (and often wouldn't work properly even if they do inherit from dict). Changing to use the abstract mapping APIs would slow all use cases for an extremely uncommon use case.

2. The C code is operating under the assumption that specific operations cannot release the GIL (e.g. dict insertion and deletion is done after precomputing the hash of the key, so it's impossible for Python byte code to be executed), so it can safely ignore thread safety issues. If a non-dict mapping was provided, implemented in Python rather than C, these assumptions could easily be violated.

3. This is basically a superset of the request from bpo-23030, which rhettinger has rejected (you can read the rationale there)

[rhettinger]

Marking as closed/rejected for the reasons lists by Josh. The functools.lru_cache() decorator is somewhat tightly focused and is trying to do one thing well. Another reason is that at some point, we want to be able to change the internals (perhaps using the new compact/ordereddict) and that would be precluded by this feature request.