Repository navigation
add a cell construtor, and expose the cell type in Lib/types.py #80092
Description
| BPO | 35911 |
|---|---|
| Nosy | @pitrou, @serhiy-storchaka, @1st1, @pierreglaser |
| PRs | |
| Files |
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 = None
closed_at = <Date 2019-02-07.19:37:16.139>
created_at = <Date 2019-02-06.10:55:47.309>
labels = ['type-feature', '3.8']
title = 'add a cell construtor, and expose the cell type in Lib/types.py'
updated_at = <Date 2019-02-07.19:37:16.138>
user = 'https://github.com/pierreglaser'bugs.python.org fields:
activity = <Date 2019-02-07.19:37:16.138>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2019-02-07.19:37:16.139>
closer = 'pitrou'
components = []
creation = <Date 2019-02-06.10:55:47.309>
creator = 'pierreglaser'
dependencies = []
files = ['48105', '48106']
hgrepos = []
issue_num = 35911
keywords = ['patch']
message_count = 3.0
messages = ['334924', '334931', '335037']
nosy_count = 4.0
nosy_names = ['pitrou', 'serhiy.storchaka', 'yselivanov', 'pierreglaser']
pr_nums = ['11771', '11771']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue35911'
versions = ['Python 3.8']
Activity
pierreglaser commentedon Feb 6, 2019
cell objects are containers for the free variables of functions defined in a
local scope. They are located in a function's __closure__ attribute (when
it is not None). A cell is a very simple object, with a single (optional, e.g
the cell can be empty) attribute: cell_contents.
The C/Python API provides a constructor to create such objects (PyCell_New).
However no cell.__new__ method is exposed to the pure python user. Workarounds
exist, but are hacky, and involve the creation of intermediate, unused
functions.
Why would cell-creation be useful? because creating cells happens in pickle
extensions modules designed to save user-defined functions and classes
(https://github.com/cloudpipe/cloudpickle) (*). These moudules are dependencies of
many widely-used data science frameworks (pyspark, ray, dask). Exposing a cell
constructor will simplify theses extensions code base, and alleviate their
maintenance cost.
I propose to add and expose a simple cell constructor, that accepts 0 (empty
cells) or 1 arguments. I also propose to expose the cell type in Lib/types.py
(as types.CellType)
(*): see related issues: https://bugs.python.org/issue35900
pitrou commentedon Feb 6, 2019
Hi Pierre,
You'll want to submit your patch as a GitHub pull request.
Also your PR will have to contain:
You may want to read more in the developer's guide:
https://devguide.python.org/
pitrou commentedon Feb 7, 2019
New changeset df8d2cd by Antoine Pitrou (Pierre Glaser) in branch 'master':
bpo-35911: add cell constructor (GH-11771)
df8d2cd