On 04/30/2012 10:36 AM, Jonathan Morgan wrote:
Hi DM,
On Tue, May 1, 2012 at 12:00 AM, DM Smith <dmsm...@crosswire.org
<mailto:dmsm...@crosswire.org>> wrote:
On 04/30/2012 09:37 AM, Daniel Owens wrote:
On 04/30/2012 06:54 AM, Chris Little wrote:
On 4/30/2012 4:39 AM, David Troidl wrote:
Hi Chris,
I'm certainly no expert on your TEI dictionaries, but
wouldn't it make
sense to have the first key be one that would sort
properly, and present
the dictionary in true alphabetical order? I'm
thinking of Middle
Liddell, as well as the Hebrew. This key wouldn't even
necessarily have
to be shown to the user. The second key, the title,
could then maintain
the proper accents for display, without hindering
sorting, searching or
navigation.
I confess, I don't understand what you're proposing this
as an alternative to.
In the example Karl cites, there's just one actual key per
entry. It is an uppercased version of the entryFree's n
attribute. This is the key that is sorted.
The un-uppercased version from the n attribute is being
rendered as part of the entry text via the TEI filters.
This is the part I'm proposing we retain, but render
somewhere else, e.g. right-justified at the bottom of the
entry.
We also render all the text of the entry, which in these
cases includes the text from a title element.
I don't know what 'true alphabetical order' means, but if
you mean localized sort order, it's not possible with the
current implementation of this module type.
--Chris
I think David's concern is something that needs to be dealt
with. A number of possibilities could be pursued, some of them
together:
1. The current implementation is to sort by unicode code
points. This works particularly well with numeric keys. A
quick solution for languages for which such sorting is not
alphabetical would be to follow David's suggestion of using
keys that the user does not even see. This has the advantage
of providing a workable solution right away, but there are
some problems with this. First, we could create a new
"strongs" standard because the current implementation does not
actually hide keys. That could be solved by making the keys so
obscure that no one would remember them. Second, any future,
more robust solution would require reworking all modules keyed
to it. I have toyed with this solution, and it might be the
pragmatic way forward, but it is not ideal.
2. A localized sort order, which I think this is what David
means by true alphabetical order, would be a better long-term
solution.
3. In addition, using genbooks for lexica would work for
lexica that are sorted by root, with subentries nested in a
hierarchy, just like in the Hesychius module and BDB. I have
been working with Troy on this. Unfortunately, front-ends do
not recognize the Feature=HebrewDef option in the conf file
and allow genbooks as lexica. I can send anyone an example
lexicon if you are interested in working on this. In that
case, instead of @n as the key, */x-entry/@osisID would be the
key.
Any thoughts?
I think there is a problem with the sorting of entries in
dictionaries where the keys are not ascii. I don't remember the
details, but I seem to remember it having been discussed here.
For JSword, we'll be building a Lucene search index for the key,
the term and the whole entry. A user lookup will be normalized and
the search will return the key with which lookup will proceed
internally as it does today. ICU provides the ability to create a
localized sort key (not at all suitable for display) that can be
used to sort dictionary entries for the end-users locale. I'm
thinking that for TEI dictionaries the representation of the key
should not be shown at all.
BPBible, and I believe some other frontends as well use binary search
on the original module order to locate a key in a virtual list. This
provides very noticeable speedups on large dictionaries like ISBE. I
think this would require the original module creation to place a
module in localised key order if we really wanted to order by that,
not just have a lookup which as I understand it would only be done
when actually looking for a key? It also really means that a module
can be sorted in one and only one way.
Then again, I'm not even sure we can guarantee any kind of binary
search on localised keys.
A related issue for English dictionaries is allowing mixed-case
dictionary keys (and I think I have heard similar comments about Greek
and maybe other languages). At the moment I think SWORD requires
dictionary keys to be upper-case to ensure that they sort correctly,
but really "Aaron's Rod" looks much better than "AARON'S ROD".
BPBible now attempts to automatically and heuristically turn keys to
mixed case, which I think looks a lot better, but ideally this would
be done in the same way as for other languages: separating sort order
from codepoint order in some way.
The idea given above is to have an index to the SWORD index. It can be
built to be ordered and accessed in whatever way is needed to solve the
problems.
As you note, the problem is that SWORD makes severe assumptions about
the order and nature of the keys. Unless care is taken uppercasing is
not always appropriate. For example in Turkish the uppercase of 'i' is
not 'I'.
In Him,
DM
_______________________________________________
sword-devel mailing list: sword-devel@crosswire.org
http://www.crosswire.org/mailman/listinfo/sword-devel
Instructions to unsubscribe/change your settings at above page