On Thu, Jan 12, 2012 at 1:46 PM, Jean-Marc Desperrier <[email protected]>wrote:

> Ted Mielczarek a écrit :
>
>> >  There is a note in the source code that points to a test that David
>>> Humphrey
>>> >  did with respect to full text search; the current approach was an
>>> order of
>>> >  magnitude faster than glimpse (IIRC).
>>>
>> This is documented 
>> here:http://vocamus.net/dave/?**p=138<http://vocamus.net/dave/?p=138>
>>
>
> Thanks, I follow Dave's blog, but apparently did not back in 2008. However
> "the explanation is back there in some old blog post" is not a great
> documentation format :-)
>

"The explanation is not correct since it's old" is not a strong argument
either.  ;-)

It would be interesting to see up to date data on this.  Although I must
say that I have no reason to believe that the results are any different.


> I'm not really satisfied with what he wrote in this blog, because I would
> like to know *why* grep on a concatenated file performs better than
> glimpse, in order to guess when it will stop being true.
>

Perhaps because grep is more efficient?

I used glimpse for a project some time ago, and it was ridiculously slow.
My hand-coded indexer/searcher which I spent about a day on written in PHP
was faster.  ;-)


> And indeed in his test, 0.335s to find the matches was good but today
> Lanedo gets 9.1s, not 0.335s, so either the good result was too easily
> destroyed in a different setup, or he underestimated the future
> consequences of source code size increase, which glimpse might have handled
> better.
>

Has anybody measured the performance of glimpse on our current code base?
Given what I see using mxr, it is not promising.


> But in any case "glimpse has a bad licence, and glimpse performs
> weirdly/badly in some cases" is a good enough reason to dump it in favor of
> FTS inside sqllite.
>

Bad license really kills it.  Bad performance is just the last nail in the
coffin!

Cheers,
--
Ehsan
<http://ehsanakhgari.org/>
_______________________________________________
dev-static-analysis mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-static-analysis

Reply via email to