Reading HH:MM data

New to AMPL
1

Hi,
Im using amplpy and working on a call center scheduling problem where the input data contains shift start and end times and duration, all in the HH:MM format. I also want to be able to use comparative operators like <, > on these time values.

I would appreciate any advice on how to pass this data and use it in an AMPL model. I’m able to read the data into a datetime format in Python, but haven’t figured out how to pass it into AMPL.

Thanks for any pointers!
Atma

1 Like
2

Hi @Atma_Iyer,

You can do that by converting the strings to numbers as follows:

from amplpy import AMPL
import pandas as pd

ampl = AMPL()
ampl.eval("""
set EVENT;
param start{EVENT};
param end{EVENT};
""")
event_df = pd.DataFrame(
    [
        (1, "10:15", "10:30"),
        (2, "11:45", "13:10"),
        (3, "10:04", "10:15"),
        (4, "12:05", "12:40"),
    ],
    columns=["EVENT", "start", "end"],
).set_index("EVENT")
timestr_to_int = lambda v: int(v.replace(":", ""))
timestr_to_minutes = lambda v: int(v[:v.find(":")])*60 + int(v[v.find(":")+1:])
event_df["start"] = event_df["start"].apply(timestr_to_int)
event_df["end"] = event_df["end"].apply(timestr_to_int)
ampl.set_data(event_df, "EVENT")
ampl.eval("display start, end;")

This produces the following output:

:  start   end     :=
1   1015   1030
2   1145   1310
3   1004   1015
4   1205   1240
;

You can choose between converting the strings directly to integers (timestr_to_int) or to minutes (timestr_to_minutes) in case you also want to calculate differences between times. If you had complete timestampts for the events you could use unix timestamps.

3

Hi Filipe,
Thanks so much for the great help. I have successfully used the timestr_to_minutes and the program runs well. We also implemented this model in Python/Pyomo and that is able to handle this data without the need for any conversion.
This may be a useful enhancement in a future AMPL release - supporting time data / timestamp values, especially for scheduling problems.

Once again, I really appreciate your responsiveness and tremendous help.
Regards,
Atma

1 Like
4

Hi @Atma_Iyer,

Thank your for your feedback! For performance reasons AMPL only allows comparisons <, <=, >, >= between numbers. Comparing strings can have substantial performance impacts on huge models (millions of variables and constraints). We are looking into a way to allow this, eventually by making a strcmp function available by default initially which would allow writing conditions like strcmp(time1, time2) < 0. I have just opened an issue for this feature request.

Please let us know of anything else you notice that may help making AMPL easier to use!

5

Thanks so much for the consideration and follow up. The performance consideration is huge - great point. Maybe a numeric representation like timestamp would be powerful. AMPL is incredibly efficient and a key differentiator in my mind, so I’m sure if and when something is done it will be great.
Ona separate note I was also able to use the symbolic parameter option recently and it is fantastic.

Thanks again
Atma

1 Like