OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

# office-formula message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

• From: "David A. Wheeler" <dwheeler@dwheeler.com>
• To: office-formula@lists.oasis-open.org
• Date: Wed, 13 Sep 2006 19:31:48 -0400 (EDT)

```A few comments, based on some latest submissions.

First, in the syntax section, you MUST indicate the type of every
parameter, parameter names should be in lowerCamelCase and should
figure out which parameter is which). So don't do this for function FOO:
do this:
GOOD Syntax: FOO(Number total; Number chosen)

Describe all the details of what the parameters mean, etc.,
in the "Semantics" section.  In particular, I expect to see the mathematical
equation that rigorously defines EXACTLY what the function is
(use Insert->object->formula) - I suggest the CRC books as
good, reputable & easily found sources (any library should have them),
though another reputable source is fine too.

In the test case results, say "Error" if it's an error (or NA if it must be
specifically NA).  Don't be more specific, not all spreadsheets have the
same error values.

There still seems to be some confusion about +/-epsilon in the results.
If a function is supposed to return an EXACT INTEGER, then DO NOT USE
+/- and epsilon. So given test "=1+4", the expected value is 5,
NOT 5 +/- epsilon.  The +/- epsilon is used when an exact answer
is NOT required.  Some functions may have to be solved interatively,
and it's reasonable to use +/- some value there even if the answer is exact,
but please don't use +/- in cases where a function SHOULD produce an
exact value.

Conversely, you MUST use +/- and some value (say epsilon) if the
result will NOT be exact.
So =SIN(0) should return EXACTLY 0, and NOT have a +/- marker on it,
because it's quite reasonable to expect that implementations should
manage to get that exact (indeed, some spreadsheet documents
may DEPEND on that).  The expected result for
SIN(x) when x<>0 will have a +/- in it, because those won't be exact.

--- David A. Wheeler

```

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]