Deprecate time.clock()

[vstinner]

BPO	14309
Nosy	@malemburg, @pitrou, @vstinner, @giampaolo

^{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 2012-04-29.00:53:40.332>
created_at = <Date 2012-03-15.00:52:01.905>
labels = ['library']
title = 'Deprecate time.clock()'
updated_at = <Date 2012-04-29.00:54:22.864>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2012-04-29.00:54:22.864>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2012-04-29.00:53:40.332>
closer = 'python-dev'
components = ['Library (Lib)']
creation = <Date 2012-03-15.00:52:01.905>
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 14309
keywords = []
message_count = 10.0
messages = ['155831', '155866', '155873', '156063', '156335', '156341', '157519', '157520', '159555', '159556']
nosy_count = 5.0
nosy_names = ['lemburg', 'pitrou', 'vstinner', 'giampaolo.rodola', 'python-dev']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue14309'
versions = ['Python 3.3']

[vstinner]

Python 3.3 has 3 functions to get time:

• time.clock()
• time.steady()
• time.time()

Antoine Pitrou suggested to deprecated time.clock() in msg120149 (issue bpo-10278).

"The problem is time.clock(), since it does two wildly different things
depending on the OS. I would suggest to deprecate time.clock() at the same time as we add time.wallclock(). For the Unix-specific definition of time.clock(), there is already os.times() (which gives even richer information)."

(time.wallclock was the old name of time.steady)

[malemburg]

STINNER Victor wrote:

New submission from STINNER Victor <victor.stinner@gmail.com>:

Python 3.3 has 3 functions to get time:

• time.clock()
• time.steady()
• time.time()

Antoine Pitrou suggested to deprecated time.clock() in msg120149 (issue bpo-10278).

"The problem is time.clock(), since it does two wildly different things
depending on the OS. I would suggest to deprecate time.clock() at the same time as we add time.wallclock(). For the Unix-specific definition of time.clock(), there is already os.times() (which gives even richer information)."

(time.wallclock was the old name of time.steady)

Strong -1 on this idea.

time.clock() has been in use for ages in many many scripts. We don't
want to carelessly break all those.

[vstinner]

time.clock() has been in use for ages in many many scripts.
We don't want to carelessly break all those.

I don't want to remove the function, just mark it as deprecated to
avoid confusion. It will only be removed from the next major Python.

[malemburg]

STINNER Victor wrote:

STINNER Victor <victor.stinner@gmail.com> added the comment:

> time.clock() has been in use for ages in many many scripts.
> We don't want to carelessly break all those.

I don't want to remove the function, just mark it as deprecated to
avoid confusion. It will only be removed from the next major Python.

Why ? There's no other single function providing the same functionality,
so it's not even a candidate for deprecation.

Similar functionality is available via several different functions,
but that's true for a lot functions in th stdlib.

[vstinner]

There's no other single function providing the same functionality

time.clock() is not portable: it is a different clock depending on the OS. To write portable code, you have to use the right function:

• time.time()
• time.steady()
• os.times(), resource.getrusage()

On Windows, time.clock() should be replaced by time.steady(). On UNIX, time.clock() can be replaced with "usage=os.times(); usage[0]+usage[1]" for example.

Which kind of use case is not covered by these functions?

[malemburg]

STINNER Victor wrote:

STINNER Victor <victor.stinner@gmail.com> added the comment:

> There's no other single function providing the same functionality

time.clock() is not portable: it is a different clock depending on the OS. To write portable code, you have to use the right function:

• time.time()
• time.steady()
• os.times(), resource.getrusage()

time.clock() does exactly what the docs say: you get access to
a CPU timer. It's normal that CPU timers work differently on
different OSes.

On Windows, time.clock() should be replaced by time.steady().

What for ? time.clock() uses the same timer as time.steady() on Windows,
AFAICT, so all you change is the name of the function.

On UNIX, time.clock() can be replaced with "usage=os.times(); usage[0]+usage[1]" for example.

And what's the advantage of that over using time.clock() directly ?

[vstinner]

I misunderstood the time.clock() function. It counts the CPU time while the process is active, whereas a monotonic clock counts elapsed time even during a sleep. time.clock() and time.monotonic() are different clocks for different purposes.

I wrote the PEP-418 which contains a list of all available OS clocks. It lists monotonic clocks, but also "process time" and "thread time" clocks. And it has a "Deferred API: time.perf_counter()" section.

Something can be done to provide portable functions to get the user and system times. See for example Tools/pybench/systimes.py written by Marc-Andre Lemburg.

Python 3.3 gives access to clock_gettime(CLOCK_PROCESS_CPUTIME_ID) and clock_gettime(CLOCK_THREAD_CPUTIME_ID).

[vstinner]

Something can be done to provide portable functions to get
the user and system times.

Lib/profile.py tests also various functions to choose the best one in Profile constructor.

[python-dev]

New changeset 314c3faea2fb by Victor Stinner in branch 'default':
Close bpo-14309: Deprecate time.clock()
http://hg.python.org/cpython/rev/314c3faea2fb

[vstinner]

The PEP-418 has been accepted: read it to understand why time.clock() is now deprecated.