Re: Sorting a Multi-occurrence data structure

[Hans Boldt <boldt@xxxxxxxxxx>]

Jon Paris wrote:
> One other comment on this topic.  If you have duplicate keys in the MODS (or
> array for that matter) that you are sorting - there is no guarantee as to
> what sequence they will be in.  Similarly if bsearch is used to locate an
> item in a set with duplicates, there is no guarantee that you will locate
> the first match in the list.  You have to "walk" backwards up the list to
> find the first occurrence if it is important to do so.
>
> Wonder if %LOOKUP comes up with the first one?  or just the first hit?  Hans
> ?  Barbara ?
Huh, uh, wha? (Sorry, I haven't been paying much attention to this
discussion, and it's still early.)

%LOOKUP seems to be a favorite topic these days. As I mentioned in
previous notes, %LOOKUP (like opcode LOOKUP) performs a sequential
search for unordered arrays. Thus, it works just like the opcode.

For ordered arrays, it assumes the programmer keeps his promise that
the data is in the proper order, and then performs a binary search.
But it doesn't just stop at the first match found. In order to match
the semantics of LOOKUP, the algorithm then walks up or down through
the matching elements to locate the appropriate target. For example,
in a %LOOKUPLT() operation, once a match is found, the built-in then
walks up the array to find the element closest to, but just below,
the search target.

Cheers!  Hans