Hi,

I could not find in documentation: is there a way of temporarily shutting off all ticktock calls? This is helpful for switching between debug and production modes.

Thanks!

Hello.

Thanks for your effort on the library, works really nice.

I would like to polish the output a bit - i.e. suppress the line numbers from the tock event, since In some cases they are redundant. For a :

@ticktick("aaa")
def a():
   pass

I want: [aaaa] 1s200 ms and now I have [aaaa:3-5] 1s200 ms

And for a:

a_clock = tick(name="train batch")
a_clock.tock()

I want: [train batch] 100ms and now I get [train batch:11] 100ms

Rationale: In the annotation case - this is obvious that a method with decorator will just get a single tock() for every tick - and so there is no need to differentiate the times, and hence it would be best if the output shown would just contain the name passed to @ticktock()

In the manual case - I would say, that this should be user-controlled. I am not sure what the API would be best, but maybe something like tick(single_tock=True) or tock(only=True) that would made the tock() not record any name for the tock and therefore the ouptut might just contain the name passed to tick ?

Hi,

I noticed there is no carriage return in the ticktock message, so I wrote the following, but I was wondering if there could be a less verbose way of doing this:

from ticktock import ticktock
from ticktock.timer import ClockCollection, set_collection
from ticktock import renderers

set_collection(
    collection=ClockCollection(renderer=renderers.StandardRenderer(format=renderers.StandardRenderer()._format + '\n')))


@ticktock
def test():
    for aa in range(100_000):
        pass


test()
print('done')

Thanks!

Hi,

Maybe it's something I don't get, but I can't get pass the import statement:

ImportError: cannot import name 'ticktock' from 'ticktock' (c:\users\gil\envs\utils\lib\site-packages\ticktock.py)
> c:\dropbox\documents_and_writings\research\digital_humanities\python\stocks\trade_ampf.py(34)<module>()
     32 from shutil import copyfile
     33 from itertools import repeat
---> 34 from ticktock import ticktock

The pip install went fine....

Running on Windows ...

Thanks!

This fixes a bug encountered when using ticktock in a REPL with unnamed tick calls.

In this context, clock.tick_filename does not exist, and name_field_fn fails because tick_name is used before being defined. This was broken recently.

The first commit adds a test to expose the bug, while the second fixes it.

This PR is relatively large, and achieves the following:

• Add the capability to have different formatting strings for different clocks, which adresses the requests in https://github.com/victorbenichoux/ticktock/issues/12 as well as https://github.com/victorbenichoux/ticktock/issues/4
• Simplifies the format string, with a general name field that works in decorators/context managers as one would expect
• Refactor by splitting ticktock.timer into ticktock.clocks with clocks, ticktock.collection with collection, ticktock.timers for decorator/contextmanager interfaces
• Refactor by removing largely unused Clock and tick parameters (e.g. tick_time_ns)
• Make clock identity consistent by never disambiguating clocks on their names
• Better tests of context manager and function decorator functionality
• numerous improvements to the documentation

This PR introduces more formatting options, by enabling the user to choose amongst more keys, for example the tick line, tock line, but also the raw timings.

This added flexibility adds a bit more constraints: in particular, format strings should also contain the clock name part:

set_format("⏱️ [{tick_name}-{tock_name}] {mean} ({std} std) min={min} max={max} count={count} last={last}")

There are also a couple of minor refactorizations.

• Add a global set_format function to set the format of ticktock messages
• Allow setting max_terms through set_format (https://github.com/victorbenichoux/ticktock/issues/4)
• Create a no_update flag in StandardRenderer to avoid print ASCII control chars

This PR fixes https://github.com/victorbenichoux/ticktock/issues/6 by adding another level of precision to the times that are shown by the StandardRenderer.

This can be controlled by setting max_terms on the StandardRenderer, which defaults to 2.

This is a small facade over renderers that we use for e.g. render at the same time to stdout and to external monitoring system. It would be nice to have it in the library

Currently we use StandardRenderer with no_update=True and it is used in a process that uses tqdm (keras training loop). The effect is that the first line from ticktock is appended at the end of the line of current tqdm progress. I can see in https://github.com/victorbenichoux/ticktock/blob/main/ticktock/std.py#L137 there is a new line appended at the end of ticktock output. We would like to add a new line at the beginning as well. WDYT of either including it be default or allowing to customize it by the client?