Time |
Nick |
Message |
05:00 |
pinesol_green |
News from qatests: Test Success <http://testing.evergreen-ils.org/~live> |
05:36 |
|
dbs joined #evergreen |
06:40 |
|
rlefaive joined #evergreen |
07:13 |
|
rjackson_isl joined #evergreen |
07:22 |
|
agoben joined #evergreen |
07:35 |
|
Dyrcona joined #evergreen |
08:40 |
Dyrcona |
So, my naive tsquery search for ISBNs on metabib.real_full_rec is taking on average 13.2 seconds. |
08:41 |
Dyrcona |
I'm just pulling 020$a from incoming records, stripping out everything that doesn't look like part of the ISBN, and then joining with ' | '. |
08:41 |
Dyrcona |
That gets passed into the query as a parameter that expects a tsquery. |
08:42 |
Dyrcona |
Since Postgres is logging all queries that execute longer than 1 second, I get a nice dump of the original query and the parameters, as well as the duration. |
08:42 |
Dyrcona |
When there are more than two ISBNs, the tsquery string often looks strange. |
08:43 |
tsbere |
(x | (y | z)) like syntax? |
08:44 |
Dyrcona |
Yeah, I was just about to paste an example. |
08:44 |
Dyrcona |
I don't think it matters, but it just seems curious. |
08:44 |
Dyrcona |
$3 = '( ''9780292766631'' | ''9780292766631'' ) | ''9780292766648''' |
08:44 |
Dyrcona |
Oh, nice. That one has the same ISBN twice... |
08:45 |
Dyrcona |
I wonder if using the search indexes would be faster.... |
08:51 |
Dyrcona |
identifier_field maybe? |
09:02 |
Dyrcona |
Nine hours to process 10,000 records seems a bit long, eh? |
10:09 |
dbs |
Yeah, seems likely that using the identifier_field infrastructure for normalizing your incoming ISBNs and matching against the corresponding cmife where cmf.name = 'isbn' would be a lot faster |
10:15 |
Dyrcona |
Yeah. I'll try that later. I've got other things going on and might as well let the current run finish. |
10:16 |
Dyrcona |
Hopefully, it'll be done before New Year's..... ;) |
10:16 |
dbs |
something like EXPLAIN ANALYZE SELECT * FROM metabib.identifier_field_entry mife INNER JOIN config.metabib_field cmf ON cmf.id = mife.field AND mife.index_vector @@ (SELECT plainto_tsquery(translate_isbn1013) FROM translate_isbn1013('270 730 249 X') LIMIT 1) LIMIT 1; --works |
10:16 |
dbs |
(well, hyphens instead of spaces but you get the idea) |
10:17 |
Dyrcona |
Thanks! |
10:18 |
dbs |
add a WHERE cmf.name = 'isbn' in there too :) |
10:19 |
Dyrcona |
:) |
11:11 |
|
Christineb joined #evergreen |
11:34 |
Dyrcona |
Ah, yeah... That's the problem with using metabib.identifier_field_entry..... |
11:34 |
Dyrcona |
ISBN includes $z and I only want to match on $a. |
11:39 |
dbs |
Hmm. May still be faster to create a custom 020$a-only cmife and match against that |
11:46 |
Dyrcona |
Yeah, maybe, but I'll need to discuss that with others. |
12:16 |
|
bmills joined #evergreen |
12:23 |
|
jihpringle joined #evergreen |
13:08 |
|
Shae joined #evergreen |
14:06 |
Dyrcona |
Ran into Archive::Zip not being installed on Debian Wheezy again today. |
14:07 |
* Dyrcona |
wonders if it is worth fixing the prerequisite installer at this late date. |
14:09 |
Dyrcona |
I mean, you really shouldn't be setting up on Wheezy, now, without good reason. :) |
14:13 |
Dyrcona |
Archive::Zip is needed for Excel::Writer::XLSX and the latter just assumes it is installed. |
14:30 |
Dyrcona |
Bleh....Trying to setup everything before putting the servers into place.... |
14:30 |
Dyrcona |
I'm going to have to give this some more thought. |
15:46 |
|
mmorgan joined #evergreen |
15:46 |
|
jvwoolf joined #evergreen |
16:50 |
|
mixo joined #evergreen |
16:50 |
mixo |
hello |
16:52 |
mixo |
I have enabled autosuggestion of search, but suggestion appears only for English titles |
16:57 |
mixo |
but I want to it to appear for russian too. |
17:00 |
|
bmills joined #evergreen |
17:02 |
pinesol_green |
News from qatests: Test Success <http://testing.evergreen-ils.org/~live> |
17:03 |
|
mmorgan left #evergreen |
17:11 |
|
jvwoolf left #evergreen |
18:17 |
|
Dyrcona joined #evergreen |