[ Date Index ][
Thread Index ]
[ <= Previous by date /
thread ]
[ Next by date /
thread => ]
On Wednesday 21 July 2004 18:12, Simon Waters wrote:
John Daragon wrote: | On Wednesday 21 July 2004 10:24, Adrian Midgley wrote: |>Perhaps a reasonable design aim might be to separate |>the display format from the underlying data, which would avoid this |> entire class of accidents being possible. | | No it *wouldn't*. That's exactly what Excel does in this case, and exactly | what these users failed to expect. SEPT02 is interpreted as 02SEP04 and | stored as 38232 (from whence it's pretty trivial to reconstruct SEPT02 if you | need to). The irreversible corruption in the paper was referring to data Excel incorrectly guessed to be floating point numbers, not dates.
I can't imagine that's much harder. Perhap's that's a failure of imagination :-)
However the reconstruction of dates would only work if there was a one to one mapping between string and date, this will not always be the case (as many strings map to the same date). So there is potential for data loss from this transform as well, although less likely. Presumably a preferred behaviour is to do no transform till someone explicitly identifies data via a format command, i.e. make the user do the right thing (TM). Or perform transforms only when an appropriate operation is applied - such as "typeless" (or whatever the term is) programming languages do.
It may be. But I *still* don't see what this has to do with whether the code is Open Source or not. jd -- John Daragon argv[0] limited john@xxxxxxxxxx Lambs Lawn Cottage, Staple Fitzpaine, Taunton TA3 5SL, UK (house) 01460 234537 (office) 01460 234068 (mobile) 07836 576127 (fax) 01460 234069 -- The Mailing List for the Devon & Cornwall LUG Mail majordomo@xxxxxxxxxxxx with "unsubscribe list" in the message body to unsubscribe.