Hi Bill,

On 1 Oct 2000, at 17:17, Bill Moseley wrote:

> At 09:36 AM 09/28/00 -0700, jmruiz@boe.es wrote:
> >- Last unstable version: swish-e-2.1.0.tar.gz.
> 
> Using the 2.1.0 swish-e binary it hangs when trying to index.  It doesn't
> get very far before eating all CPU time:
> 
> S   PID  PPID %CPU %MEM  RSS   VSZ COMMAND
> R 30925 30924 96.9  1.0 1376  2232 swish-LII -c swish.cfg
> 

Well, All the work done in 2.1.X is related to memory. I think that if 
there is an idle CPU, the best thing we can do is using it. In the other 
hand, if there are other proccess that can suffer the impact of the  
CPU eating of swish-e, we can change its priority using UNIX 
commands. 

Can you try 2.1.1 to check the problem? The fix can affect also the 
your index problem. I have tried it with several file-systems with no 
problems.

> A gdb backtrace does help much.  (Any tricks for getting better debugging
> info?).  I've included the gdb backtrace and the config file below.
> 
> >3. BumpPositionCounterCharacters option. See user.config in conf 
> 
> Is there a way to say, for example, a character will bump if at the end of
> a word, but not in the middle of a word?
> 
>    The location was www.swish.com. 
>                        ^         ^
>                    no bump      bump
> 

Sure, documentation needs an update. I think that we must get 1.3.X 
documentation and update it with all the new features. A FAQ is also 
a good idea.
Any volunteers?

> >4. New option -x to write more info ("extended header") of the index 
> >files while searching.
> 
> Can WordCharacters, Ignore and all the index-related settings be displayed
> per index?  And then how hard would it be to print an extra field after the
> file name to indicate which file the result was found?
> 

It can be added, but I have to change the way it is working.
Now, more or less, this how it works with the header info:

Read header indexfile 1
total_header=header_indexfile_1
while there_is_indexfile
{
total_header=merge_headers(total_header,header_indexfile_n)
}

So, the header indexfiles are lost. Only the merged one is preserved.

Writing the indexfilename as an extra field is easier.

Give me a couple of days to implement the modifications.

cu
Jose