Time |
Nick |
Message |
00:59 |
|
Newziky joined #evergreen |
03:53 |
|
berick joined #evergreen |
07:28 |
|
rjackson_isl joined #evergreen |
08:33 |
|
Dyrcona joined #evergreen |
08:37 |
Dyrcona |
berick | kmlussier: Should I load the branch on lp 1564685, instead of working/user/berick/browser-patron-reg-3b-bugs-wip? |
08:37 |
pinesol_green |
Launchpad bug 1564685 in Evergreen "Angular patron editor post-2.10 misc. bugs" [Undecided,New] https://launchpad.net/bugs/1564685 |
08:37 |
* Dyrcona |
assumes so, but you know what assuming can do. |
08:37 |
kmlussier |
Dyrcona: Yes, please load that one instead. |
08:38 |
kmlussier |
Dyrcona++ |
08:38 |
Dyrcona |
OK. |
08:43 |
|
mmorgan joined #evergreen |
08:53 |
Dyrcona |
kmlussier: There's a conflict with the working/user/dyrcona/lp1564402-evergreen-isbn13-function and current master. |
08:53 |
Dyrcona |
The conflict is easy to resolve and you have a choice. |
08:53 |
* kmlussier |
gets her coin ready for flipping |
08:53 |
Dyrcona |
There's a console.debug() statement that was commented out in master. |
08:53 |
Dyrcona |
It is still active in the wip branch. |
08:53 |
Dyrcona |
I can choose either option. |
08:54 |
Dyrcona |
If it is left active, the JavaScript console will get a lot of messages about what is filling grids added to it. |
08:54 |
Dyrcona |
Dunno if you want that while testing or not. |
08:55 |
kmlussier |
Dyrcona: I think you can leave it commented out at this point. |
08:55 |
Dyrcona |
OK. |
09:00 |
Dyrcona |
The branch is ready. Time to build it, and upgrade the database. |
09:03 |
|
Newziky joined #evergreen |
09:09 |
Dyrcona |
csharp++ # For ambiguous reasons. ;) |
09:11 |
* Dyrcona |
brushes up on adding a limit to a circ matrix matchpoint while waiting on upgrade scripts. |
09:16 |
Dyrcona |
Ah...that's interesting. |
09:16 |
Dyrcona |
They have a limit. Apparently, they just want it changed. |
09:42 |
|
mllewellyn joined #evergreen |
09:43 |
|
collum joined #evergreen |
09:48 |
|
yboston joined #evergreen |
09:54 |
Dyrcona |
git commit --patch is nice when you want to split recent changes into multiple commits, except when what you want to split is in the same patch hunk. |
10:04 |
Dyrcona |
Ok, I take that back. The e command in git commit --patch takes care of that rather well. |
10:15 |
csharp |
git++ |
10:28 |
Bmagic |
where is the setting that makes naco_normalize_keep_comma instead of standard naco_normalize? |
10:37 |
kmlussier |
Bmagic: config.metabib_field_index_norm_map maps config.metabib_field definitions to the normalizer that should be used. |
10:37 |
Bmagic |
thanks! |
10:53 |
Bmagic |
kmlussier: ok, so, the config.index_normalizer ID#3 specifies "NACO Normalize -- retain first comma" - in your config.metabib_field_index_norm_map - do you have rows for field 31 and 32 ? |
10:56 |
kmlussier |
Bmagic: nope |
10:57 |
Bmagic |
hmmm, without a norm mapping for field 31 ("title";"browse") and 32 ("series";"browse") it seems like it is stripping the comma |
10:58 |
|
Christineb joined #evergreen |
11:20 |
csharp |
acq question: is there any direct link between fund debits and PO line items? if the PO activation errors out at any point, it needs to be "reset" and I'm trying to figure out how to remove (or offset) fund debits that should no longer exist |
11:21 |
csharp |
it doesn't appear that there is a link between the two from my initial investigation |
11:22 |
berick |
csharp: the link is between acq.lineitem_detail's |
11:26 |
Bmagic |
kmlussier: I'm not sure stripping the comma is my problem. When the indexes are sorted, it's not respecting the space character even when the comma is removed. |
11:26 |
Bmagic |
kmlussier: for example this is the order that the db returns "wood alix" "woodall" "wood amanda" |
11:26 |
kmlussier |
Bmagic: Yeah, I think I said the other day I didn't think it was related to the comma |
11:27 |
kmlussier |
But I don't know why it's sorting differently for you than it is for me |
11:27 |
Bmagic |
select sort_value from metabib.browse_entry where id in(select entry from metabib.browse_entry_def_map where source in(1360318,856019,1287379))order by sort_value |
11:27 |
Bmagic |
the database is sorting them... er wrong* |
11:28 |
Bmagic |
what causes postgres to ignore the space character during the "order by" clause ? |
11:30 |
Dyrcona |
Bmagic: Are you sure there is a leading space? |
11:31 |
Bmagic |
its the space between "wood" and "alix" |
11:31 |
Dyrcona |
Oh, spaces in the middle.... |
11:31 |
csharp |
berick: ah - thanks - I'll look there |
11:31 |
Dyrcona |
I've never noticed that behavior, but I've never looked. |
11:32 |
Bmagic |
yep, that query returns that order "wood alix" "woodall" "wood amanda" |
11:32 |
Bmagic |
"woodall" should go after "wood amanda" don't you think? |
11:32 |
Dyrcona |
If is it an asciibetical sort, yes, I would expect that. |
11:33 |
Dyrcona |
oy.... it is not is it. |
11:33 |
Bmagic |
? |
11:33 |
|
bmills joined #evergreen |
11:34 |
Bmagic |
do I have the wrong character set on the DB or something? |
11:34 |
Dyrcona |
Doubt it would make much difference. |
11:35 |
Dyrcona |
LC_COLLATE is probably the main driver here, but I could be wrong. |
11:36 |
Dyrcona |
Ours is LC_COLLATE = 'C' |
11:36 |
Dyrcona |
I'm going to run your query out of curiosity. Granted my records with those ids won't be the same, so the results are likely meaningless. |
11:38 |
Dyrcona |
As I thought, my records with those ids don't throw up any flags. |
11:39 |
Dyrcona |
Closest that I get is 'churchill a life' followed by 'churchill winston' which does what is expected. |
11:39 |
Dyrcona |
I'll see if I can find records with those same names in them. |
11:40 |
kmlussier |
Dyrcona: I looked at the browse sorting for all the MassLNC sites a couple of days ago and didn't see the same behavior Bmagic had encountered. |
11:40 |
|
Callender joined #evergreen |
11:40 |
Bmagic |
bash# locale LC_COLLATE="en_US.UTF-8" |
11:40 |
Dyrcona |
Bmagic: No, the LC_COLLATE for the database. |
11:41 |
Dyrcona |
You can check that in PgAdmin by clicking the database name and looking in the SQL Pane to the lower right. |
11:41 |
Bmagic |
CREATE DATABASE evergreen WITH OWNER = postgres ENCODING = 'UTF8' TABLESPACE = pg_default LC_COLLATE = 'en_US.UTF-8' LC_CTYPE = 'en_US.UTF-8' CONNECTION LIMIT = -1; |
11:41 |
|
bmills joined #evergreen |
11:42 |
Dyrcona |
Bmagic: LC_COLLATE and LC_CTYPE should be 'C' at least that is my understanding. |
11:43 |
Bmagic |
so, I have to drop the DB and create it again? |
11:43 |
Dyrcona |
You might be able to change it on an existing database. I don't know. |
11:43 |
Bmagic |
en_US.UTF-8 seems legit though |
11:44 |
Dyrcona |
Well, it is legit, but I believe there is a good reason to use C with Evergreen. |
11:44 |
Dyrcona |
I'm a bit fuzzy on the details. |
11:44 |
Bmagic |
no kidding, wow, how the heck did this happen |
11:45 |
|
Callender joined #evergreen |
11:46 |
Dyrcona |
CREATE DATABASE eg_main WITH OWNER = superuser ENCODING = 'UTF8' TABLESPACE = pg_default LC_COLLATE = 'C' LC_CTYPE = 'C' CONNECTION LIMIT = -1; |
11:46 |
Dyrcona |
That's from our production database. |
11:47 |
Dyrcona |
I vaguely recall discussing it back in 2010/2011 when I asked about using en_US.UTF-8, myself, but I could be making that up. |
11:48 |
Bmagic |
I see the code right here Open-ILS/src/sql/Pg/create_database_extensions.sql |
11:48 |
Bmagic |
CREATE DATABASE :db_name TEMPLATE template0 ENCODING 'UNICODE' LC_COLLATE 'C' LC_CTYPE 'C'; |
11:48 |
Dyrcona |
Yeah. It is possible to create the database in other ways, though. |
11:49 |
Bmagic |
sheesh, when we upgraded postgres, I pg_dumped the DB and restored it. |
11:49 |
Bmagic |
I wonder if it was like that before the dump |
11:49 |
Dyrcona |
Most likely. I've never had those settings change as a result of restoring a dump. |
11:50 |
Dyrcona |
Except for the name and owner, my development database is the same as production, and I restored it this morning from last night's production dump. |
11:51 |
Bmagic |
looking back at my dump.sql file - CREATE DATABASE evergreen WITH TEMPLATE = template0 OWNER = postgres; |
11:51 |
Bmagic |
it didn't specify those switches during the creation |
11:52 |
Dyrcona |
Ah, could your template0 has them. |
11:52 |
* Dyrcona |
cannot English. |
11:52 |
Dyrcona |
;) |
11:52 |
Bmagic |
template0.... that is a file somewhere? |
11:55 |
Bmagic |
well, this has all been enlightening, I've got some work to do |
11:55 |
Bmagic |
Dyrcona++ kmlussier++ |
11:56 |
Dyrcona |
I was just looking, and I don't recall where the templates live. Somewhere in the postgres admin database, I think. |
11:57 |
tsbere |
template0 is normally the "nothing loaded by default" template. Template1 is the "includes more than the bare pieces" possibly including "global" customizations template. |
11:57 |
tsbere |
Also, any *database* can also be treated as a template, as far as I know... |
11:59 |
Dyrcona |
Right. I believe I copied production once by using it as a template just to see if that worked. |
11:59 |
Dyrcona |
That said, I didn't think that template0 and template1 are databases. |
11:59 |
Dyrcona |
I also thought they could found in pg_catalog somewhere, but i'm not seeing them. |
12:00 |
Dyrcona |
Someone in #postgres is bound to know. |
12:00 |
jeffdavis |
our template0 uses en_US.UTF-8 for collate and ctype but restored pg dumps of our Evergreen db have collate & ctype = 'C' |
12:00 |
jeffdavis |
psql -l should show them |
12:00 |
csharp |
I've experience transaction wraparound warnings for template0/1 before, so they *are* databases as far as I understand |
12:01 |
csharp |
you can vacuum them |
12:01 |
Dyrcona |
jeffdavis++ |
12:01 |
csharp |
(which I do to avoid the xact wraparound issue) |
12:01 |
Dyrcona |
So, mine show en_US.UTF-8 also, but the database is restored with C. |
12:02 |
Dyrcona |
Yeah, looks like they are databases and PgAdmin hides them from you. |
12:03 |
|
brahmina joined #evergreen |
12:03 |
Dyrcona |
Cool. I (re)learned something. :) |
12:03 |
Dyrcona |
Anyway, gonna get some lunch. |
12:07 |
jeffdavis |
pg_dump -i -Fd -Z 9 -f /path/to/dumpfile --serializable-deferrable evergreen |
12:07 |
jeffdavis |
^ that's basically the command we use to create a dump of our db - note that it does not output a .sql file |
12:08 |
jeffdavis |
for restore, we create a new database using Evergreen's create_database_extensions.sql script, then use pg_restore to load the dumped data into it |
12:09 |
Bmagic |
jeffdavis: yeah, I was going to basically do those steps in testing today |
12:10 |
Bmagic |
Which switches do you use for eg_db_config.pl ? |
12:11 |
jeffdavis |
perl Open-ILS/src/support-scripts/eg_db_config --update-config --service all --create-offline |
12:11 |
jeffdavis |
plus user, password, host, port, and database |
12:12 |
Bmagic |
right |
12:12 |
Bmagic |
ok, thanks, that is helpful |
12:13 |
jeffdavis |
happy to help, good luck! |
12:18 |
|
jihpringle joined #evergreen |
12:28 |
|
jihpringle joined #evergreen |
12:40 |
miker |
Bmagic: you can change the collation of specific columns, or even just of indexes built on columns. no need to blow away your whole db |
12:41 |
miker |
http://www.postgresql.org/docs/9.1/static/indexes-collations.html |
13:28 |
|
DPearl joined #evergreen |
13:32 |
|
ericar joined #evergreen |
13:39 |
Bmagic |
miker: Sounds shakey. It would be more comforting knowing that the whole db is setup as 'C' and not just some* columns. As I am not 100% sure which columns are important and what affects what |
14:09 |
|
abowling joined #evergreen |
14:44 |
|
ericar joined #evergreen |
14:45 |
|
ericar joined #evergreen |
14:50 |
|
ericar joined #evergreen |
14:52 |
|
ericar joined #evergreen |
14:54 |
|
ericar joined #evergreen |
15:17 |
kmlussier |
berick: We have one more 2.8 release before it becomes security only, right? |
15:18 |
berick |
kmlussier: yes |
15:19 |
kmlussier |
Thanks! |
15:19 |
berick |
well, at least one more. i forget what the policy is exactly. |
15:19 |
rhamby |
have a good weekend all |
15:19 |
* Dyrcona |
asked about that earlier this week, 'cause I wasn't sure, either. |
15:19 |
Dyrcona |
rhamby: You, also! |
15:21 |
kmlussier |
This email says one month after two subsequent stable releases - http://list.georgialibraries.org/pipermail/open-ils-dev/2011-November/007746.html |
15:22 |
kmlussier |
Google doesn't show me any updates to the policy in that email. |
15:23 |
* kmlussier |
should look at old meeting minutes to see if it was actually voted upon and approved. |
15:23 |
Dyrcona |
Well, that sound to me like 2.8.8 next month would be the last. |
15:27 |
berick |
thanks kmlussier |
15:27 |
berick |
i guess we'll know when 2.10.2 is cut |
15:32 |
miker |
Bmagic: sorry, meetings. Not sure what's shakey about it ... it's a standard feature of postgres, and for other sort-ish columns we do other thing -- like normalize values with padding and such. using a locale-aware collation is important for sorting anything but 7bit ascii correctly, and, indeed, for data integrity more generally beyond sorting. but, that's just one DBA's opinion. we could, of course, instead simply normalize the browse sort column |
15:32 |
miker |
to account the known difference between ascii and unicode sorting in a locale-aware environment. |
15:36 |
kmlussier |
OK, that information was outdated. The revised policy was posted to the list a month later - http://markmail.org/message/53umdzope7nobtqb. It looks like a release gets bug fix support for 12 months after its release, at which point it moves to security release. |
15:36 |
* kmlussier |
is going to put this on a wiki page |
15:36 |
kmlussier |
If it's not there already |
15:39 |
kmlussier |
It looks like 2.8.0 was released on March 31, 2015, so, yes, it looks like we get one more bug-fix release. |
15:39 |
kmlussier |
Right back where we started. |
15:55 |
* Dyrcona |
sings "Back where we started....He we go 'round again...." |
15:56 |
Dyrcona |
heh. typo-fingers strike again. |
15:58 |
|
jlitrell joined #evergreen |
16:20 |
|
wjr joined #evergreen |
16:21 |
|
geoffsams joined #evergreen |
16:25 |
kmlussier |
http://wiki.evergreen-ils.org/doku.php?id=dev:release_process:schedule |
16:28 |
Dyrcona |
kmlussier++ |
16:29 |
Bmagic |
miker: so, in this one paticular case, I could drop the index and create it again with the correct locale option? |
16:33 |
pinesol_green |
[evergreen|Mike Rylander] LP#1373601: Consider relevant characters before using word-boundary checks - <http://git.evergreen-ils.org/?p=Evergreen.git;a=commit;h=e7c8785> |
17:16 |
|
mmorgan left #evergreen |
17:36 |
miker |
Bmagic: I think you'd want to set the collation on the column and then reindex it ... that should be enough, I believe. a test server could tell. that said, I do think normalizing the sort value differently would be a good step to avoid this altogether ... like, replace spaces with +s or some such (more thought than I've just given it would be warranted, of course) |
17:37 |
|
maryj joined #evergreen |
17:39 |
Bmagic |
miker: thanks again! |
17:49 |
Bmagic |
I guess I have to drop the column and add it again. alter table metabib.browse_entry alter column sort_value SET NOT NULL COLLATE 'C'; does not work. errors on "COLLATE" |
18:16 |
miker |
bmagic: not at a computer, but there's a spell if you change the data type. just change it to be the same. |