Sunday, 9 January 2011

Pinot 0.97

After a long break, I am releasing a new version that brings improvements to crawling, memory management and the indexing of acronyms. Several bugs were fixed and translations were massively updated.
I would like to thank all translators involved, and especially Nikolay Kachanov for his feedback. Спасибо большое !

As usual, the NEWS file has the full details and the source can be downloaded here.

Finally, it's come to my attention that Pinot is featured in the Desktop Search section of Datamation's Ultimate List of Open Source Software ! Another good reason to celebrate this New Year ;-)


Martin Dengler said...

Hi - I built 0.97 for Fedora 14, and have been using it for a few days now. I love the search results. The memory usage, not so much:

# ps axuwww | grep 2005
martin 2005 0.8 16.7 2982308 637940 ? SNl May04 13:02 pinot-dbus-daemon

Is there a good way to start figuring out what's causing the memory usage to get so high? Here are some stats:

$ du -hcx ~/.pinot/index/
48K /home/martin/.pinot/index/

$ du -hxc ~/.pinot/daemon/
2.3G /home/martin/.pinot/daemon/

$ du -kxc ~/.pinot/daemon/* | sort -n | tail
4 /home/martin/.pinot/daemon/record.baseB
4 /home/martin/.pinot/daemon/termlist.baseA
32 /home/martin/.pinot/daemon/spelling.baseA
32 /home/martin/.pinot/daemon/spelling.baseB
372 /home/martin/.pinot/daemon/record.DB
47708 /home/martin/.pinot/daemon/termlist.DB
82092 /home/martin/.pinot/daemon/position.DB
341956 /home/martin/.pinot/daemon/postlist.DB
1885636 /home/martin/.pinot/daemon/spelling.DB
2357856 total

Fabrice Colin said...

Hi Martin,
Is this on the first index pass ? Or was the daemon restarted after the index was first built ?
Take a look at the FAQ (installed in /usr/share/doc/pinot-0.97) for hints.
If you want to reduce the size of the daemon's index, you could turn off spelling support (see the README, section 15).
Feel free to post on pinot-discuss.

Martin Dengler said...

Hi Fabrice,

Thanks for the README pointer; it's got useful stuff. It mentions the potentially large disk size of the spelling db; it would be good to mention the RAM size as well, especially if it's only the first time. I will try to get a patch to you in case it's useful.

I am certain those stats weren't from the first time I ran pinot-dbus-daemon (with the directories set) but the daemon had definitely been killed a few times by the OOM-killer (I have 4GB of ram) so perhaps it was the first time, effectively.

I will post on pinot-discuss, thanks.

Martin Dengler said...

Fedora's Pinot is going to be at 0.97 soon - thanks for your work.

Fabrice Colin said...

Hi Martin,
Thanks for this, I really appreciate.