Re: [Bibdesk-develop] Indexing for full text search.

[Sven-S. P. <ssp...@ea...>]

>> Christiaan's reply suggests that BibDesk isn't actually doing  
>> superfluous indexing in those situations but just verifies the  
>> existing index.
>
> ...and indexing any files that weren't in the index.

I quit and restarted BibDesk after one such 'indexing' run to be sure  
that everything was indexed the last time. So (modulo my ignorance of  
how this feature actually works) there shouldn't be any new indexing  
going on.

>> If that is the case, perhaps it would make sense to use different  
>> status strings for indexing vs. index verification. That way users  
>> will not get the impression that BibDesk is wasting their time be  
>> reindexing stuff and should have more confidence that the search  
>> results displayed are up-to-date.
>
> That makes some sense, but I'm curious as to how long it's taking to  
> verify the index.

About 45 seconds.

> How many files do you have, and are any of them on a network volume?

I don't have an exact count but my guess is around 750. All of them  
are on my local hard drive.

> Have you tried profiling the verification phase to see if it can be  
> sped up?  Too frequent UI updates can hammer the index and really  
> slow things down.


I mostly clueless about profiling, so I just ran Shark for 30 seconds  
along with BibDesk (about 25s or which were indexing progress). The  
only item standing out in the result of that was 45% of the time being  
used in sha1_block_data_order called by +sha1SignatureForFile: in  
NSData(BDSKExtensions). The next biggest item was listed with 3%.

		Sven

-- 
Sven-S. Porst . http://earthlingsoft.net/ssp . AIM: cv47al
Pass as best inventor!