Leap second day utc2tai interpolation

Hi astropy developers,

I am from North American pulsar time group(Nanograv). When I am processing some data taken in a leap second day(more exactly mjd 56109 ), I realize in that day, the leap second is interpolated. Here is a test I made.

>>> from astropy.time import Time
>>> test_leap = np.linspace(56107,56110,500)
>>> test_time = Time(test_leap, format='mjd', scale='utc')
>>> plt.plot(test_time.mjd, (test_time.tai.mjd - test_time.mjd)*86400.0)

And I have the plot.

My question what is the reason to do an interpolation, and is there any way to shut the interpolation off.

Thank you.

-Luo Jing

[taldcroft]

In theory we can add a warning when anyone uses the jd or mjd attributes of a UTC time object. I’m not generally a huge fan of warnings, but this does seems like a good case of “you are doing something that is likely to get you into trouble because there is a subtle implementation detail you might not be aware of.”

[mwcraig]

In theory we can add a warning when anyone uses the jd or mjd attributes of a UTC time object.

This seems like a good compromise, especially since there is a correct way to calculate time differences in astropy given UTC input (difference the Time objects not the julian dates):

# A time one second before a leap second
pre_leap_second = Time("1994-06-30T23:59:59", scale='utc', format='isot')

# A time one second after the same leap second:
post_leap_second = Time("1994-07-01T00:00:00", scale='utc', format='isot')

# The Wrong Way to get the difference:
bad_interval = post_leap_second.jd - pre_leap_second.jd
print(bad_interval * 86400)
# Prints 1.9999891519546509, which is not 2.

# The Right Way:
good_interval = post_leap_second - pre_leap_second
print(good_interval.jd * 86400)

#prints 1.9999999999939178, which is 2 to machine precision