Helge Hafting <[EMAIL PROTECTED]> writes:
>
> Andreas Vox wrote:
> > ...
>
> That's the part I don't really get. How could you possibly know the
> context for any
> word in that list? Do you plan on showing some context for every word
> in your list?
Yes, but I now see that this won't be feasable without modification
(Just rechecked the Wordlist for the Userguide). You can tell I know
index generation only from theory :-)
Ok, imagine a description environment. The label is the index key. After that
there are "ReverseIndexInsets" for the first occurance of this word in any
numbered subsection (this is my modification, don't want to have more than
100 occurances in the index buffer -- I know, you'd still have to prune them
down to 0-9).
A ReverseIndexInset would be collapsable, with the section number as its
label, eg. "30.1.2") When expanded, it would show the context, ie. the
sentence containing the occurance. You could change the style of a
ReverseIndexInset ("see XY", "off", or some charstyle).
If there is other text between the index key and the first ReverseIndexInset,
it will be used for rendering in the final index.
Changing the nesting of the description items would create subitems.
There are three update methods:
1. create index buffer from existing index entries
2. create index buffer from existing index entries and word list. Previously
not existing index entries will get the style "off"
3. merging the index buffer back into the text. Any deleted
ReverseIndexitems or ones with style "off" will be removed.
All others are updated with new parameters.
...
> but the current suggestion
> seems to imply that most of the work will lie in *removing* lots and
> lots of not indexworthy words,
> and then lots and lots of indexworthy words because I don't want to
> index anywhere near every
> occurence of the indexworthy words.
>
> By all means, build that word list with frequencies and every word in
> the document. But I think
> we'll be better off if the default choice is that the words are just
> listed, but no index entry added for
> any of them. The user should explicitly mark (or click, or whatever)
> the words to go in the index,
> because that will less work than removing all the words that should
> *not* go in the index.
...
> a nice jump function to look at context will be immensely helpful. But
> still, no words should actually
> go into the index unless selected by the author.
Ok, I take your point. I hope the "off" style gets your consent.
I also like some of GÃnters ideas:
* include Figure/Table numbers in the labels
* a "muted" style for index entries
Anyway, all this wont come before 1.5.0. or later.
The wordlist and the Regexp-Find-and-Index function will probably be
first because they don't need much change.
/Andreas
