Round down to a chosen hour-min-sec time at a date

Hello,

I use

now_time = datetime(now,'ConvertFrom','datenum')

Assume it gives me

ans = 
  datetime
   14-May-2019 02:28:24

Then assume I have some chosen time settings hh:mm:ss, eg:

   A = [ 02:00:00 18:00:00 08:00:00 ]

I would like something to go back to the nearest time from my settings, in this case it should give me this (due to above now_time):

   14-May-2019 02:00:00

If the now_time was 14-May-2019 01:59:59, then due to its before 02:00:00, I would like it to go to the time backwards "round down":

   13-May-2019 18:00:00

Another example: Or if now_time is 13-May-2019 08:00:01

I would like:

   13-May-2019 08:00:00

Hope I am clear enourgh and hope for some advise. Thanks in advance

Answer 1

[Walter Roberson]

B = dateshift(now_time, 'start', 'day') + A_in_duration_form;
interp1(B, B, now_time, 'prev')

[Martin]

Hello Walter, thanks a lot for your reply, very appriciated!

It almost works. But if I have A = [ hours(20) hours(2) ].

And if now_time is 14-May-2019 02:30:00 it will correcly say = 14-May-2019 02:00:00.

But if now_time is between 00:00:00 and 01:59:59 it will give the result

ans = 
  datetime
   NaT

Here I would hope for 13-May-2019 20:00:00

[Walter Roberson]

B = dateshift(now_time, 'start', 'day') + A_in_duration_form;
B = [B-days(1), B];

[Martin]

Unfortunately I still opserve those NaT results :/

[Walter Roberson]

   A = [ 02:00:00 18:00:00 08:00:00 ]

Those values need to be in sorted order, especially when you talk about going back days.

"from 2 in the morning until 6pm and then until 8am the next day but go back a day if you need to" is ambiguous: is 03:00 in the 02:00:00 to 18:00:00 range, or is it in the 18:00:00 to 08:00:00 range?

If you want to start from a particular day (for example you might have found midnight Sunday) and proceed on a multiday schedule, then use appropriate offsets, such as [hours(2) hours(18) hours(24+8)

[Martin]

Hi thanks for getting back.. It is always chronologically, so e.g. 14-Mar 02:00:00, 14-Mar 08:00:00, 14-Mar 18:00:00, 15-Mar 02:00:00. or 14-Mar 18:00:00 15-Mar 02:00:00, 15-Mar 08:00:00, 15-Mar 18:00:00

I ran it now, here 14-May, and the results gave me this:

now_time = datetime(now,'ConvertFrom','datenum')
A = [ hours(01) hours(07) hours(13) hours(19) ];
B = dateshift(now_time, 'start', 'day') + A;
B = [B-days(1), B];
interp1(B, B, now_time, 'prev')

now_time = 
   datetime
   14-May-2019 23:14:17
ans = 
   datetime
   NaT

In this case I would wish for the:

ans = 
   datetime
   14-May-2019 19:00:00

Answer 2

[Peter Perkins]

Since you're working in liear time (preserving dates) and no in circular time (caring only about time of day), I would think discretize would be a good way to do this:

>> dt = datetime('today') + hours(sort(48*rand(10,1)))
dt = 
  10×1 datetime array
   04-Jun-2019 01:48:41
   04-Jun-2019 04:44:17
   04-Jun-2019 12:34:11
   04-Jun-2019 17:38:13
   04-Jun-2019 18:59:05
   05-Jun-2019 08:35:54
   05-Jun-2019 14:13:00
   05-Jun-2019 18:29:17
   05-Jun-2019 19:50:15
   05-Jun-2019 23:25:23
>> bins = datetime(2019,6,[4 4 4 4 5 5 5 5],[0 2 8 18 2 8 18 24],0,0)
bins = 
  1×8 datetime array
Columns 1 through 5
   04-Jun-2019 00:00:00   04-Jun-2019 02:00:00   04-Jun-2019 08:00:00   04-Jun-2019 18:00:00   05-Jun-2019 02:00:00
Columns 6 through 8
   05-Jun-2019 08:00:00   05-Jun-2019 18:00:00   06-Jun-2019 00:00:00
>> discretize(dt,bins,bins(1:end-1))
ans = 
  10×1 datetime array
   04-Jun-2019 00:00:00
   04-Jun-2019 02:00:00
   04-Jun-2019 08:00:00
   04-Jun-2019 08:00:00
   04-Jun-2019 18:00:00
   05-Jun-2019 08:00:00
   05-Jun-2019 08:00:00
   05-Jun-2019 18:00:00
   05-Jun-2019 18:00:00
   05-Jun-2019 18:00:00