Re: [MiNT] toshyp update (toshyp search)

[Paul Wratt <paul.wratt@gmail.com>]

On Sat, Jun 18, 2011 at 10:05 AM, Helmut Karlowski
<helmut.karlowski@ish.de> wrote:
> Am 17.06.2011, 23:20 Uhr, schrieb Paul Wratt <paul.wratt@gmail.com>:
>
>> If people like it, Gerhard can have the extra pages and set up the
>> search engine (it is simple) and I will redirect my url, otherwise it
>
> Quite handy, but I have a problem:
>
> I enter e.g. appl_init and get:
>
> ------------------------------------
>
> AND Search      OR Search
> Phrase Search   Show  results per page
>
> Search: Only in "en'" All Languages
> Categories
>
>
>  Displaying results 1 - 3 of 3 matches (0 seconds)
> 1. [100.00%]  The documentation for TOS: Signals
>  void ) { long ssp; appl_init (); printf("My ProcID is %d.\n", Pgetpid ());
> Psignal (SIGUSR1 , handler); if (setjmp(env)) Cconws ("Coming from
> longjmp.\r\n"); else Cconws ("Coming from
> http://toshyp.isource.net.nz/en/005007.html - 24.2kb (search again)
>
> 2. [100.00%]  The documentation for TOS: GEM
>  in the global field by the appl_init call. The VDI , on the other hand, has
> actually no version number of its own, the more so as the behaviour of the
> individual VDI -functions is determined principally by the device drivers
> used which are,
> http://toshyp.isource.net.nz/en/002002.html - 9.4kb (search again)
>
> 3. [100.00%]  The documentation for TOS: Index
>  of appl_getinfo_str appl_init appl_options appl_read appl_search appl_tplay
> appl_trecord appl_write appl_xbvset appl_xgetinfo appl_yield Apportionment
> of resources with threads APP_ACCESSORY APP_APPLICATION
> http://toshyp.isource.net.nz/en/indexudo.html - 295.4kb (search again)
>
>  Result page: 1
>  Powered by
>
> -------------------------------
>
> None of these points to the actual definition of appl_init. Why is that?
>
>
> --
> Helmut Karlowski
>
short answer:
it was not fully indexed

----------------------------------------------------------------------------------------------

OK I just re-checked, the indexing was getting about 60 pages each in
EN + DE, but there are 186 useful pages (no chapter 1), so I
re-indexed both urls

the search engine is a generic web/document indexer. Toshyp is a very
specific type of web page, I had to turn off a few "features" because
they would just make search results less acurate. Also page weighting
is generally calculated by occurrences, and where they occur (title,
url, text, links, etc).

That being said, no search engines index the true keys of toshyp, the #links

in toshyp some html pages have names instead of chapter numbers, which
also affect search results. also some of those page names are
mis-leading unless you know what they apply too, ie Application.html
refers to AES Application Library. Also some #links are erroneous too,
ie #Application_20library => 8.3 Application library


the technical answer is hinted at in my original post, the exact link is:
Application.html#appl_init

I notice that there are 9 hits in the auto complete, yet only 3
results returned. I have noticed that 2 results are missing from the
1st page of results (when more than 20), so I believe there is an
issue with an incomplete list of results atm. It appears the auto
complete is also lacking, as there are 12 occurrences on
Application.html. I will re-catalog with a greater depth, that may
help

The first thing to get TOS.HYP search results accurate is to add #link
indexing, which would be weighted at 100%. This is merely an option in
the site indexer (in admin), and integrating a new weight.

Because this is an open source package I can also get some issues
fixed as well as adding the new stuff, but apart from what is
discussed here, it is already more powerful than most regular search
engines


Paul