I have this scenario. Start with an integer value e.g. 365000. Over a period of time (usually a year) I will add one record a day with a proportional amount (so in this very basic example I would add one record each day of 1000 for the next 365 days). The total of these records should add up to the initial integer value. So each day I calculate how much of the initial total remains, and how many days are left to allocate it, and add a new record with the correct proportional amount.

Now this example is a gross oversimplification as a number of variables can change (including the initial value and the period), but the golden rule still applies - the sum of the records at the end of the period should equal the initial total. It goes without saying, I rarely have a nice easy integer value like 365,000 to start with!

As you might imagine, there will be all sorts of precision issues here. I have chosen to store my calculated data as float, as I believe this gives me the best chance of all my records adding up to the correct amount (or as near as possible!). Storing with any other type (decimal, money etc.) will necessarily compromise on precision and therefore adding up the records will almost certainly mean I will get a larger difference than if I store my results as floats.

I have read advice earlier in this thread suggesting I not store my data as floats. Given this scenario does that advice still stand? I don't think so but would like to hear any thoughts people have.

Thanks,

Ash

p.s. this is financial data 😛