utime – time related functions

|see_cpython_module| python:time.

The utime module provides functions for getting the current time and date, measuring time intervals, and for delays.

Time Epoch: Unix port uses standard for POSIX systems epoch of 1970-01-01 00:00:00 UTC. However, embedded ports use epoch of 2000-01-01 00:00:00 UTC.

Maintaining actual calendar date/time: This requires a Real Time Clock (RTC). On systems with underlying OS (including some RTOS), an RTC may be implicit. Setting and maintaining actual calendar time is responsibility of OS/RTOS and is done outside of MicroPython, it just uses OS API to query date/time. On baremetal ports however system time depends on machine.RTC() object. The current calendar time may be set using machine.RTC().datetime(tuple) function, and maintained by following means:

  • By a backup battery (which may be an additional, optional component for a particular board).

  • Using networked time protocol (requires setup by a port/user).

  • Set manually by a user on each power-up (many boards then maintain RTC time across hard resets, though some may require setting it again in such case).

If actual calendar time is not maintained with a system/MicroPython RTC, functions below which require reference to current absolute time may behave not as expected.

Functions

utime.gmtime([secs])
utime.localtime([secs])

Convert the time secs expressed in seconds since the Epoch (see above) into an 8-tuple which contains: (year, month, mday, hour, minute, second, weekday, yearday) If secs is not provided or None, then the current time from the RTC is used.

The gmtime() function returns a date-time tuple in UTC, and localtime() returns a date-time tuple in local time.

The format of the entries in the 8-tuple are:

  • year includes the century (for example 2014).

  • month is 1-12

  • mday is 1-31

  • hour is 0-23

  • minute is 0-59

  • second is 0-59

  • weekday is 0-6 for Mon-Sun

  • yearday is 1-366

utime.mktime()

This is inverse function of localtime. It’s argument is a full 8-tuple which expresses a time as per localtime. It returns an integer which is the number of seconds since Jan 1, 2000.

utime.sleep(seconds)

Sleep for the given number of seconds. Some boards may accept seconds as a floating-point number to sleep for a fractional number of seconds. Note that other boards may not accept a floating-point argument, for compatibility with them use sleep_ms() and sleep_us() functions.

utime.sleep_ms(ms)

Delay for given number of milliseconds, should be positive or 0.

utime.sleep_us(us)

Delay for given number of microseconds, should be positive or 0.

utime.ticks_ms()

Returns an increasing millisecond counter with an arbitrary reference point, that wraps around after some value.

The wrap-around value is not explicitly exposed, but we will refer to it as TICKS_MAX to simplify discussion. Period of the values is TICKS_PERIOD = TICKS_MAX + 1. TICKS_PERIOD is guaranteed to be a power of two, but otherwise may differ from port to port. The same period value is used for all of ticks_ms(), ticks_us(), ticks_cpu() functions (for simplicity). Thus, these functions will return a value in range [0 .. TICKS_MAX], inclusive, total TICKS_PERIOD values. Note that only non-negative values are used. For the most part, you should treat values returned by these functions as opaque. The only operations available for them are ticks_diff() and ticks_add() functions described below.

Note: Performing standard mathematical operations (+, -) or relational operators (<, <=, >, >=) directly on these value will lead to invalid result. Performing mathematical operations and then passing their results as arguments to ticks_diff() or ticks_add() will also lead to invalid results from the latter functions.

utime.ticks_us()

Just like ticks_ms() above, but in microseconds.

utime.ticks_cpu()

Similar to ticks_ms() and ticks_us(), but with the highest possible resolution in the system. This is usually CPU clocks, and that’s why the function is named that way. But it doesn’t have to be a CPU clock, some other timing source available in a system (e.g. high-resolution timer) can be used instead. The exact timing unit (resolution) of this function is not specified on utime module level, but documentation for a specific port may provide more specific information. This function is intended for very fine benchmarking or very tight real-time loops. Avoid using it in portable code.

Availability: Not every port implements this function.

utime.ticks_add(ticks, delta)

Offset ticks value by a given number, which can be either positive or negative. Given a ticks value, this function allows to calculate ticks value delta ticks before or after it, following modular-arithmetic definition of tick values (see ticks_ms() above). ticks parameter must be a direct result of call to ticks_ms(), ticks_us(), or ticks_cpu() functions (or from previous call to ticks_add()). However, delta can be an arbitrary integer number or numeric expression. ticks_add() is useful for calculating deadlines for events/tasks. (Note: you must use ticks_diff() function to work with deadlines.)

Examples:

# Find out what ticks value there was 100ms ago
print(ticks_add(time.ticks_ms(), -100))

# Calculate deadline for operation and test for it
deadline = ticks_add(time.ticks_ms(), 200)
while ticks_diff(deadline, time.ticks_ms()) > 0:
    do_a_little_of_something()

# Find out TICKS_MAX used by this port
print(ticks_add(0, -1))
utime.ticks_diff(ticks1, ticks2)

Measure ticks difference between values returned from ticks_ms(), ticks_us(), or ticks_cpu() functions, as a signed value which may wrap around.

The argument order is the same as for subtraction operator, ticks_diff(ticks1, ticks2) has the same meaning as ticks1 - ticks2. However, values returned by ticks_ms(), etc. functions may wrap around, so directly using subtraction on them will produce incorrect result. That is why ticks_diff() is needed, it implements modular (or more specifically, ring) arithmetics to produce correct result even for wrap-around values (as long as they not too distant inbetween, see below). The function returns signed value in the range [-TICKS_PERIOD/2 .. TICKS_PERIOD/2-1] (that’s a typical range definition for two’s-complement signed binary integers). If the result is negative, it means that ticks1 occurred earlier in time than ticks2. Otherwise, it means that ticks1 occurred after ticks2. This holds only if ticks1 and ticks2 are apart from each other for no more than TICKS_PERIOD/2-1 ticks. If that does not hold, incorrect result will be returned. Specifically, if two tick values are apart for TICKS_PERIOD/2-1 ticks, that value will be returned by the function. However, if TICKS_PERIOD/2 of real-time ticks has passed between them, the function will return -TICKS_PERIOD/2 instead, i.e. result value will wrap around to the negative range of possible values.

Informal rationale of the constraints above: Suppose you are locked in a room with no means to monitor passing of time except a standard 12-notch clock. Then if you look at dial-plate now, and don’t look again for another 13 hours (e.g., if you fall for a long sleep), then once you finally look again, it may seem to you that only 1 hour has passed. To avoid this mistake, just look at the clock regularly. Your application should do the same. “Too long sleep” metaphor also maps directly to application behavior: don’t let your application run any single task for too long. Run tasks in steps, and do time-keeping inbetween.

ticks_diff() is designed to accommodate various usage patterns, among them:

  • Polling with timeout. In this case, the order of events is known, and you will deal only with positive results of ticks_diff():

    # Wait for GPIO pin to be asserted, but at most 500us
    start = time.ticks_us()
    while pin.value() == 0:
        if time.ticks_diff(time.ticks_us(), start) > 500:
            raise TimeoutError
    
  • Scheduling events. In this case, ticks_diff() result may be negative if an event is overdue:

    # This code snippet is not optimized
    now = time.ticks_ms()
    scheduled_time = task.scheduled_time()
    if ticks_diff(scheduled_time, now) > 0:
        print("Too early, let's nap")
        sleep_ms(ticks_diff(scheduled_time, now))
        task.run()
    elif ticks_diff(scheduled_time, now) == 0:
        print("Right at time!")
        task.run()
    elif ticks_diff(scheduled_time, now) < 0:
        print("Oops, running late, tell task to run faster!")
        task.run(run_faster=true)
    

Note: Do not pass time() values to ticks_diff(), you should use normal mathematical operations on them. But note that time() may (and will) also overflow. This is known as https://en.wikipedia.org/wiki/Year_2038_problem .

utime.time()

Returns the number of seconds, as an integer, since the Epoch, assuming that underlying RTC is set and maintained as described above. If an RTC is not set, this function returns number of seconds since a port-specific reference point in time (for embedded boards without a battery-backed RTC, usually since power up or reset). If you want to develop portable MicroPython application, you should not rely on this function to provide higher than second precision. If you need higher precision, absolute timestamps, use time_ns(). If relative times are acceptable then use the ticks_ms() and ticks_us() functions. If you need calendar time, gmtime() or localtime() without an argument is a better choice.

Difference to CPython

In CPython, this function returns number of seconds since Unix epoch, 1970-01-01 00:00 UTC, as a floating-point, usually having microsecond precision. With MicroPython, only Unix port uses the same Epoch, and if floating-point precision allows, returns sub-second precision. Embedded hardware usually doesn’t have floating-point precision to represent both long time ranges and subsecond precision, so they use integer value with second precision. Some embedded hardware also lacks battery-powered RTC, so returns number of seconds since last power-up or from other relative, hardware-specific point (e.g. reset).

utime.time_ns()

Similar to time() but returns nanoseconds since the Epoch, as an integer (usually a big integer, so will allocate on the heap).