Hi, Biggest con: BigInteger are slower by an order of magnitude and use up much more CPU, think processing files with millions of lines, which users of POI do.
Also it is more complex to handle and you still need to decide how you round numbers when you decide to store them as strings. Dominik On Fri, Oct 18, 2019, 05:08 Andreas Reichel <andr...@manticore-projects.com> wrote: > Hi Andi and Team, > > thank you for prompt response. > > On Thu, 2019-10-17 at 20:11 -0700, kiwiwings wrote: > > > this issue pops up every now and then. Please have a look at > > > http://apache-poi.1045710.n5.nabble.com/Floating-point-behaviour-difference-between-POI-and-Excel-td5715765.html > > I will read through that. > > > So Excel rounds to 15 digits. I'm undecided about the BigDecimal API, > > so Ileave it to the others to decide. > > Pardon my curiosity: what are the Cons? I might be biased by my > profession, but I believe floating point arithmetic sucks. > > > Regarding the DB inserts, why do you store numbers in varchars(12)? > > ... Iwould expect the jdbc API to truncate/round the doubles and use > > the maximumnumber type. > > Well, we hit an exceptional field here: Base Rate can be a number (like > 5% p.a.) or a reference to a yield curve like LIBOR or EURIBOR. > Would we not hit exactly the same problem with NUMBER(3,9) or > DECIMAL(3,9) when 0.10669129999999999 exceeds the precision/scale? > > Thank you again and cheers > Andreas >
