PyFloat_FromString deprecated form

[jimjjewett]

BPO	1650903
Nosy	@gvanrossum, @birkenfeld

^{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 2008-01-06.22:29:46.829>
created_at = <Date 2007-02-02.19:41:29.000>
labels = []
title = 'PyFloat_FromString deprecated form'
updated_at = <Date 2008-01-06.22:29:46.829>
user = 'https://bugs.python.org/jimjjewett'

bugs.python.org fields:

activity = <Date 2008-01-06.22:29:46.829>
actor = 'admin'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['None']
creation = <Date 2007-02-02.19:41:29.000>
creator = 'jimjjewett'
dependencies = []
files = []
hgrepos = []
issue_num = 1650903
keywords = []
message_count = 6.0
messages = ['31165', '31166', '31167', '31168', '31169', '31170']
nosy_count = 3.0
nosy_names = ['gvanrossum', 'georg.brandl', 'jimjjewett']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue1650903'
versions = ['Python 3.0']

[jimjjewett]

PyFloat_FromString takes two arguments, and the only purpose of the second is to avoid changing the API.

Extracts from Tim's 2000 comment:

Since we can't change the interface of a public API function, pend is still supported but now *officially* useless: if pend is not NULL, *pend is set to NULL.

[birkenfeld]

What do you propose to do?

[birkenfeld]

(cont'd)
C API functions shouldn't even change signature in Py 3000. Therefore, a new name would be needed...

[jimjjewett]

I didn't realize that a decision had been made about the C API stability.

I would indeed propose changing it, but probably not if it the rest of the API (even for strings and unicode?) is supposed to stay stable.

[gvanrossum]

I think I should relax that statement a bit. IMO it's OK to change the signature (for Py3k) since the compiler will report an error if you call it with the old signature. What's *not* OK is to keep the signature (or to keep it compatible anyway) but change the *behavior*, as that would cause bugs that are only caught (if at all) at runtime, and will hence be hard to debug. Examples of bad changes would be changing the reference count behavior of an API, changing the type of object returned (if a specific type was previously documented) or making it possible to return NULL where this was previous not possible.

[birkenfeld]

Okay, I fixed this case in rev. 54433.