>>>I was trying to think of a way to short circuit the loops once the sum of a group of numbers exceeds the target value...
>>
>>Don't know the practical side of the problem, but this reminds me of almost the same problem which we tried to solve (well, at least as a mental experiment) thirteen years ago: the customer was a poultry slaughterhouse (but they omitted the "poultry" when answering the phone :). There was a packing line, where a bunch of individually packed chicken would hang, ready to be dropped into a carton. The given data are: the weight of each of the chickens, the target weight of the carton, and the allowable error. Usually the carton would hold 20kg, and the error margin was +/- 50g.
>>
>>The problem was similar to the problem of fitting a choice of music to a tape to have minimal length of silence in the end, except that this allowed the error to go either way (below or over), while on tape you weren't supposed to cut the last song.
>>
>>We never really solved any of these.
>
>Dragan --
>
>Did they?

I think they're still doing it manually. Or they had engaged some science institute to solve this, but those guys were way too expensive...

>Man, a .25% tolerance is pretty tight.

Thirteen years, I say. Probably was more like 150g on 20kg, or even +/- 1%. Anyway, knowing them, they did give it a lot of slack, usually by making the error to top it over a bit, so nobody would complain.

>It seems like it would be more work to identify and pick the proper ones ("will the 2.45 kg chicken in room 3, aisle b, box 49, level 3, 5th from the right and 7th from the top please come forward"), than use human judgment and give that a little slack.

There were about 30 hooks on the transporter, one chicken hanging on each, and they would have to press a button with the hook number or so.

The main reason we never tried this was the temperature... first we thought watching them do it manually would get us some insight and bring ideas, but it's freezing there. We gave it up and they didn't push it.