Re: [dict-beta] Doubts and some ways to work arround

[Diego N. <die...@ya...>]

Hi All,

> Until the manpage is corrected, it will be
> necessary to read the source code to see where
> the header ends for each option.

Ok. I haven't read the "formal documentation" for
dictfmt, yet. :-)

> > and dupplicating the last word on the index file.
> > I verified the "last word problem" with wn too.
> A patch to fix that bug would be welcome.

I will see  what I can do. The dict-wn  distribution
has the
word "Zyrian"  repeated at wn.index, though.  Maybe we
could
just remove it manually until we have a patch.

Concerning multilanguage support, here  is my point of
view,
regardless of the plans for full UTF8 support:

Supporting LC_CTYPE  is very usefull because  it
enables non
english locales to use whichever characters they
consider to
be  "isalpha",  and also  enables  the  system to 
correctly
convert  between upper  and lower  case characters  by
using
standard "tolower" and "toupper" macros.

I understand  the index  file has to  be sorted,  of
course.
What I don't  see is enough reasons to sort  the index
under
different locales. For the binary search to work, it
doesn't
really matters  which characters come before  which,
as long
as the order used in the index file is the same as the
order
the searching  algorithm expects. For that,  one could
stick
to pure strcmp (which uses the C locale, period).

I  think  this would  simplify  the  code in 
dictd/index.c,
dictfmt.c  and yet  make  the system  more  robust.
The  way
things are at  the present, changing the locale  used
by the
dictd daemon  can break a  dict/index pair: the 
server will
fail to  find words that are  in the database because 
it is
expecting  a different  order.  Mandating  the C 
LC_COLLATE
locale (or just using strcmp) will solve this.

The signed/unsigned problem is a different story...

Running the following program

    #include <locale.h>
    #include <ctype.h>
    #include <stdio.h>

    int main(void)
    {
        setlocale(LC_CTYPE, "en_US");
        printf("%d %c %d %c\n", á

__________________________________________________
Do You Yahoo!?
Yahoo! Health - Feel better, live better
http://health.yahoo.com