[seqfan] Re: conventions

[Maximilian Hasler]

That's nice work !
For the rounding issue, one should obviously recommend  to compute a
bunch of  digits more than needed. (I'm not even sure that all CAS
guarantee the last digits to be correct, when computations are made
sharply with the required (rather: requested) accuracy...)

Of course round(n+1/2) should be = n+1 for any integer n, on OEIS like
eveywhere else (except maybe in banking / finance & related...)

Maximilian

On Mon, Mar 30, 2009 at 9:18 AM, Douglas McNeil <d.mcneil at qmul.ac.uk> wrote:
>
>
> So I've been playing with parsing the description of a sequence to
> automatically verify parts of the OEIS.  Obviously most truly interesting
> sequences are out of reach, but then again truly interesting sequences are
> likely to be getting more attention anyway: the idea here is to babysit
> constants, recurrences, sequences with specified formulae, and sequences
> of the "n such that prime(n^2)-3 == 5 mod 7" sort.
>
> Found a fair number of minor errors this way.  Two questions came up:
>
> (1) One thing I've noticed is that in some of the decimal expansions, the
> last entries in the sequence correspond not to a slice of the infinite
> decimal expansion but to a version rounded to the number of digits given.
>
> E.g. A003881 = pi/4.  The OEIS sequence ends 9,2,6,7,0, which is its
> rounded version of the group of digits 92669955.
>
> This rounding seems a little strange to me, as it makes sequences
> dependent on the number of entries.  When possible (which it usually is)
> I'd have expected the terms to be the same as in the infinite case.
> Is this intended, or an accidental side effect of doing "Digits:=100;
> evalf(Pi/4);" in maple or the equivalent?
>
> (2) In OEIS-world, what do round(1/2) and round(3/2) correspond to?
>
>
> Doug
>
>
> _______________________________________________
>
> Seqfan Mailing list - http://list.seqfan.eu/
>