query performance CLI vs. webui

[sottolski]

Hi there,

nice litte tool. However, any idea why it takes about 20-30 secs to answer a query, instead of like 0.1 s on the CLI? The funny thing is, if the result is returned, even the GUI shows those "0.01 s" with the result. So it very much looks like the web server part takes a long time (and 100 % CPU while processing the request).

In case it matters, I indexed a huge maildir (112 GB, over 1 Mio. Messages, resulting in 18 GB in the index directory).

The effect is the same with the standalone as well WSGI w/ apache2.

Any idea (or debug hints) what the CPU cycles are spent for?

Oh, I should add, tested this on a wheezy box with a git clone of the webui, and the recoll-1.22.3 from the recoll ppa.

Thanks

Sascha

[sottolski]

uh, should have done this before, started the standalone again, and observed it with strace. seems the webui is consuming every file in the indexed maildir - which as I said are well about a million of them. this already happens without doing a query, even for just presenting the query interface...

[sottolski]

got something: If I change the Folder depth (number of levels of the folder dropdown) from the default 3 to 1, the UI is super quick!

BTW, took me some time to figure out how to set the initial config path for the recoll config I wanted to use, an ended up with setting

os.environ["RECOLL_CONFDIR"] =

couldn't figure out how to "inject" the env via apache2, SetEnv didn't work for me.

[KlausItem]

Hey sottolski,

can you tell me where you found this Setting "Folder Depth"???

Thanks