| Time |
Nick |
Message |
| 06:56 |
|
collum joined #evergreen |
| 08:41 |
|
mmorgan joined #evergreen |
| 09:26 |
|
Dyrcona joined #evergreen |
| 10:07 |
gmcharlt |
csharp_: I shall do what I want... and not ignore yoU! |
| 10:12 |
Dyrcona |
"Testing. Please delete." :) |
| 10:27 |
Dyrcona |
Are 19 parameters too many? Someone is going to complain they are not enough. |
| 10:27 |
Dyrcona |
I guess really there are only 17 that the user can control. The other two are given by the framework. |
| 10:48 |
goood |
and in the end, you only need one: query :) |
| 10:49 |
|
BDorsey joined #evergreen |
| 10:50 |
Dyrcona |
True, but..... |
| 10:52 |
csharp_ |
gmcharlt: thank you for your service |
| 10:52 |
Dyrcona |
Well, here's a surprise: I've added the first two boolean parameters. |
| 10:53 |
csharp_ |
one thing we can't ignore is that a few mail reputation sites had marked list.evergreen-ils.org as bad |
| 10:53 |
csharp_ |
pretty sure it's password-reset generated spam |
| 10:56 |
Dyrcona |
goood: For Boolean parameters, will it accept anything other than true or false? |
| 10:57 |
Dyrcona |
"It" being the openapi_server and apparatus. |
| 10:59 |
Dyrcona |
I'm wondering if I can get away with using OpenILS::Application::Apputils::is_true or if I should implement my own version for OpenAPI booleans? |
| 11:00 |
goood |
I believe it expects JSON-ish values. sec... |
| 11:01 |
goood |
yep: https://swagger.io/docs/specification/v3_0/data-models/data-types/#boolean |
| 11:01 |
goood |
just, exactly, true and false |
| 11:01 |
Dyrcona |
OK, so this should work: $docache = $U->is_true(lc($docache)); |
| 11:02 |
Dyrcona |
Probably don't need the lc. |
| 11:02 |
goood |
if it's nullable, obviously null is allowed and can be considered either false or "some third value", but .. yeah, we treat it as false |
| 11:03 |
Dyrcona |
It's optional. |
| 11:04 |
Dyrcona |
Should I add a default_value of false? |
| 11:05 |
Dyrcona |
I'll do that. Makes sense anyway that it's false if it is missing, at least for these tow options. |
| 11:05 |
goood |
meh, if it's not true, it's false. I think that ends up being a JSON::Boolean (or similar) |
| 11:06 |
Dyrcona |
No documentation found for "JSON::Boolean". |
| 11:06 |
goood |
(I don't have a strong opinion on default values, unless it serves a functional purpose) |
| 11:08 |
goood |
ah, it's just JSON::true and JSON::false. for testing in stringy or int-y ways automatically: https://metacpan.org/pod/JSON#JSON%3A%3Ais_bool |
| 11:09 |
goood |
anyway, they're safe to test directly in openapi-side code, looks like |
| 11:13 |
Dyrcona |
I see: "They are overloaded to act almost exactly like the numbers 1 and 0." Guess I'll just go with what I've got and see what happens. Hopefully, I'll make enough progress to test today. |
| 11:14 |
Dyrcona |
See if this works: $docache = {docache=>$docache}; |
| 11:16 |
Dyrcona |
goood: It looks like the parameters are passed to the controller function in the order that they're listed in the endpoint table entry. |
| 11:16 |
goood |
yes, exactly |
| 11:19 |
goood |
method_params lets you define the opeapi param mapping to the perl backend, so, for instance, you can pass a static value to the handler to change it's behavior. there are some that are reused because they do the same work with different params in different positions. like the verify-by-barcode and verify-by-username, IIRC |
| 11:20 |
goood |
mainly, it's the named-parameter to positional-parameter mapping we need to be able to pass through to opensrf methods |
| 11:20 |
goood |
which makes it familiar to opensrf developers, and understandable to openapi developers |
| 11:22 |
goood |
(though, similar to opensrf where you get the method object as an invocant (and the client handle), you get the controller object as the invocant for Advanced Stuff(tm)) |
| 11:23 |
Dyrcona |
Yeah, so this: my ($c, $ses, $title, $author, $subject, $series, $keyword, $identifier, $isbn, $issn, $org_unit, $depth, $limit, $offset, $sort, $sort_dir, $tag_circulated_records, $query, $docache) = @_; |
| 11:26 |
goood |
that sure is exercising every possible option! :D |
| 11:27 |
goood |
tbh, I don't recall top-level isbn and issn, but if the code says it works... |
| 11:27 |
Dyrcona |
Well, the ones that make sense. It's not possible to have arbitrary options is it? like ?search_format=X |
| 11:28 |
goood |
no, you have to define them all, but you could add (say) a facet param and then include information about the facet type and value in the param value |
| 11:29 |
Dyrcona |
What should I do with string values in a limited range? Should I return a 400 if they're out of range? |
| 11:30 |
goood |
and it is possible (at the openapi level) to define structured params that get exploded into their parts, like facet={type:foo,value:[bar,baz]}, but our openapi param parser doesn't handle that yet |
| 11:30 |
goood |
strings in a range? |
| 11:30 |
goood |
like, an org_unit that doesn't map to a shortname? |
| 11:30 |
Dyrcona |
sort_dir is either asc or desc, or sort can be 1 of puddate, author, or title. |
| 11:31 |
goood |
ah! well, there's an enum type. not sure if I've used that (or support it yet, tbh) ... sec |
| 11:32 |
goood |
(oh, and sort can be those 3, OR any record attribute that is marked as a sorter, or some special values like "rel" and "poprel") |
| 11:33 |
Dyrcona |
Well..... Dunno if I wan to get into that. I'm not parsing the query string where I think you can also specify sort. |
| 11:34 |
Dyrcona |
I've never tried to figure out what happens if you try to use both the arghash and query string sort. |
| 11:34 |
goood |
ah, no enum support yet in the schema datatype table. so, I think just make sure it's a simple string and let the backend figure it out |
| 11:35 |
goood |
for sort, I mean |
| 11:35 |
Dyrcona |
Would take a lot of code to implement enum support? I'm not sure where to begin, tbh. |
| 11:37 |
goood |
it would not be trival. it might be simpler to define a regex for string-type params that can be used to validate it |
| 11:38 |
Dyrcona |
OK, I'll use a regex in my code for now. We can figure out refactoring it to the database and openapi_server later. |
| 11:38 |
goood |
+1 |
| 11:39 |
Dyrcona |
I'm trying to get this done so I can put it on a test server by Wednesday. |
| 11:39 |
goood |
that would be 99% in support of the openapi description doc and schema we deliver, for client code generation |
| 11:39 |
goood |
(and docs/"explanations" can cover that for us here) |
| 11:40 |
Dyrcona |
That was something I noticed. I didn't see anyway to add a description to parameters for the Swagger docs. Is that not part of the OpenAPI specs? |
| 11:43 |
goood |
it is supported by v3.0, yes. the schema has "example" and "examples/summary" properties for that: https://swagger.io/docs/specification/v3_0/adding-examples/ |
| 11:44 |
goood |
we could add cols to openapi.endpoint_param to hold those |
| 11:45 |
goood |
(and enums might be less hard than I first thought, but still db and code changes that aren't NECESSARY today. maybe as part of v1 promotion?) |
| 11:45 |
Dyrcona |
Maybe in time for 4.0? |
| 11:46 |
goood |
oh, almost definitely not |
| 11:47 |
Dyrcona |
ok |
| 11:48 |
goood |
but if a vendor's willing to start using your local additions now, v1 doesn't seem like a hard requirement |
| 11:48 |
Dyrcona |
I was thinking by the Evergreen 4.0 release in the fall. You don't think we could knock out enums and descriptions this summer? |
| 11:50 |
Dyrcona |
I also wonder what kind of validation/scrubbing I should do on the inputs. For most of the strings, I probably need to remove punctuation and maybe duplicate spaces. |
| 12:02 |
|
jihpringle joined #evergreen |
| 12:07 |
goood |
(meeting time) anything's possible, but I can't guarantee time or a completion target |
| 12:13 |
goood |
IMO, we shouldn't modify user input, just (at most) validate and error if it's not to our liking |
| 12:14 |
goood |
(that DOES suggest enum, min/max, and regex support, but we can do it in the handler for now) |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:19 |
jihpringle |
à |
| 12:20 |
csharp_ |
keyboard cat! |
| 12:20 |
jihpringle |
it was indeed :) |
| 12:20 |
csharp_ |
jihpringle++ |
| 12:20 |
* csharp_ |
watches jeff reholster his moderator guns |
| 12:39 |
|
collum joined #evergreen |
| 13:16 |
Dyrcona |
heavy use was made of skeletons and keyboard macros during the production of this code: https://www.gnu.org/software/emacs/manual/html_node/autotype/index.html |
| 13:53 |
* Dyrcona |
listens to The Steve Miller Bad in honor of updating to Linux Kernel 7.0.7 this morning. |
| 14:03 |
Bmagic |
lol |
| 14:28 |
Dyrcona |
Path: /search/multiclass?title=alchemist&author=coelho |
| 14:28 |
Dyrcona |
{"errors":[{"message":"Internal Server Error.","path":"\/"}],"status":500} |
| 14:31 |
Dyrcona |
And that's about all I have to go on. Nothing in the nginx error.log |
| 14:35 |
* Dyrcona |
decides to stop logging to syslog on this test vm. |
| 14:40 |
goood |
Dyrcona: you may see more useful info in the console where you start the server, not certain. but you def won't see anything useful in the nginx log if you get to the point of seeing that error structure |
| 14:41 |
Dyrcona |
goood: Not really: [2026-05-15 14:41:06.17516] [3212] [error] [gWs5kphvWvj9] Exception! at /usr/share/perl5/Error.pm line 484. |
| 14:41 |
goood |
that's almost certiainly bubbling up through basic_handler_wrapper in openapi_server |
| 14:42 |
goood |
ok, so it is likely a problem in the handler, or behind it, that's throwing an error or calling die() |
| 14:43 |
Dyrcona |
grepping gWs5kphvWvj9 turns up not much. I can check the log table in the db. |
| 14:43 |
goood |
that is just the request id |
| 14:43 |
goood |
it's a random string |
| 14:43 |
Dyrcona |
Yeah, but there's associated log entries. |
| 14:44 |
Dyrcona |
Nothing indicating an error. |
| 14:44 |
goood |
is the wip branch more or less up to date? |
| 14:45 |
Dyrcona |
grep ERR just turns up a bunch of osrfStringArrayAdd messages. |
| 14:45 |
Dyrcona |
Yeah, the wip branch is what I'm using. Relevant stuff is all in the top commit. |
| 14:49 |
goood |
+1 |
| 14:52 |
goood |
I see a couple typos: open-ils.search.biblio.multiclasse (note the "e" at the end), and a single "|" instead of "||" for ORing things together |
| 14:52 |
Dyrcona |
goood++ |
| 14:52 |
goood |
(not sure that's what we want to do there, regardless ... |
| 14:52 |
Dyrcona |
I figured it had to be a typo somewhere. |
| 14:53 |
goood |
you /should/ be seeing the "no method called multiclasse" in the opensrf logs, though |
| 14:53 |
goood |
modulo exact words, obv |
| 14:55 |
Dyrcona |
Nope. :/ |
| 14:55 |
goood |
womp womp |
| 14:58 |
goood |
I know the opensrf method definition says docache is supposed to be an object, but ... it doesn't look like that's true |
| 14:58 |
Dyrcona |
Ok. |
| 14:58 |
goood |
it's just a bool (flag) param |
| 15:00 |
Dyrcona |
Y'know. I thought I saw that yesterday, then looking at the doc comments this morning I thought, "I must have been mistaken." |
| 15:01 |
goood |
it's dealer's choice if that defaults to true ... IIRC, we send 0 for that in searches with the #staff modifier, and some special cases re metarecord searching, but otherwise send a 1 |
| 15:01 |
goood |
default to true /here/ I mean |
| 15:02 |
|
mmorgan1 joined #evergreen |
| 15:02 |
Dyrcona |
OK. It's just code, we can always change it. :) |
| 15:03 |
Dyrcona |
After I get a successful search, I'll change the default in my database and restart the openapi server. |
| 15:05 |
Dyrcona |
Meh. Still an interal server error. Do I have to restart all services? |
| 15:05 |
goood |
no, just openapi_server |
| 15:07 |
Dyrcona |
So, I'll just start dumping the params. They'll show up in the console, won't they? |
| 15:08 |
goood |
unsure ... they'll show up somewhere! you should be able to use the standard logger to log to the opensrf logs, I think? |
| 15:09 |
goood |
but, lemme see what your result stuff looks like. the bare call returns an arrayref of arrayrefs, I think |
| 15:09 |
Dyrcona |
so, I should probably have array for the result type. I'll bet that's my problem. |
| 15:10 |
goood |
yeah, it does validate result structure |
| 15:10 |
Dyrcona |
can I do schema_type 'array' array_items 'array'? |
| 15:11 |
* Dyrcona |
tries it. |
| 15:11 |
goood |
you could just say `return {foo => $U->simple...}''` |
| 15:11 |
goood |
yes |
| 15:12 |
Dyrcona |
I'll try the double array. If that doesn't work, I'll go back to object with {results => ....} |
| 15:12 |
Dyrcona |
I want this to be as thin a wrapper as possilbe. |
| 15:12 |
Dyrcona |
possible, even... |
| 15:16 |
Dyrcona |
Meh. still not working... |
| 15:17 |
Dyrcona |
I'm going to set validate to false to see if that makes a difference. |
| 15:19 |
Dyrcona |
Nope. There's a problem elsewhere still. |
| 15:28 |
Dyrcona |
"syntax OK" |
| 15:34 |
Dyrcona |
Weird. If I do query= and nothing else. I get to a line that I added that dumps \%arghash. If do the title&author one. I don't get there. |
| 15:35 |
Dyrcona |
I added another line above that to dumpe the searches hash. I don't get there, either. |
| 15:45 |
Dyrcona |
Something is wrong with add_to_hash, but i don't see it, and Perl's not giving me any hints. |
| 15:46 |
Dyrcona |
So, we'll go back to the repetitious version. |
| 15:51 |
Dyrcona |
OMG! Another typo... |
| 15:51 |
Dyrcona |
Grr. |
| 15:52 |
Dyrcona |
Perl should have complained about a misspelled function name. |
| 15:53 |
goood |
ha! join_paramaters ... just saw that |
| 15:53 |
Dyrcona |
No 'use strict; use warnings;' that's why no errors form perl -c... |
| 15:54 |
Dyrcona |
That's what you get when you own 2 tow trucks: a par'amaters... |
| 15:54 |
goood |
yeah, perl -wc didn't warn me either |
| 15:55 |
Dyrcona |
still not working. |
| 15:55 |
Dyrcona |
ugh I made a typo fixing the typo... |
| 15:55 |
Dyrcona |
I need to slow down. |
| 15:58 |
Dyrcona |
Ok. I'm not getting the fields with default values: offset, limit, etc. |
| 16:00 |
Dyrcona |
think I'll check for errors, etc. |
| 16:04 |
|
jihpringle joined #evergreen |
| 16:09 |
Dyrcona |
OK. Yet another typo.... Dunno why Perl doesn't catch more of these. |
| 16:10 |
Dyrcona |
If it weren't typos, I'd have no type at all.... Another one after fixing that one. |
| 16:12 |
Dyrcona |
I had multiple typos in the "my $U = 'OpenILS::Application::AppUtils'; line... preventing $U->simplereq from working/being called, but Perl didn't notice other than just blowing up at runtime. |
| 16:12 |
csharp_ |
@quote add < Dyrcona> If it weren't typos, I'd have no type at all.... |
| 16:12 |
pinesol |
csharp_: The operation succeeded. Quote #259 added. |
| 16:14 |
Dyrcona |
Now, the quetion becomes can we run multiclass search from simplereq, because it appears to not work. |
| 16:15 |
Dyrcona |
Looks like mojolicious timed out. |
| 16:16 |
Dyrcona |
Attempting to build a client session as a server Session ID |
| 16:16 |
Dyrcona |
yeahp. gonna have to do this differently, I think. |
| 16:17 |
goood |
heh... seems like it should work, though. it works for other handler wrappers |
| 16:18 |
Dyrcona |
I'm going to remove my print Dumper code and see what happens. I think the search just may be too slow, but this machine had a pingest run recently. Maybe I need to reindex metabib tables? |
| 16:19 |
goood |
:shrug: |
| 16:20 |
Dyrcona |
The Isbn retrieval was fast. |
| 16:20 |
goood |
gather (used in simplereq) has an explicit 60s timeout, if that matters |
| 16:21 |
Dyrcona |
I got a line from mojolicious about 120 seconds in the console. After that, I got the message I pasted above, that I assume was what I got back from simplereq, but maybe it wasn't. |
| 16:24 |
Dyrcona |
[2026-05-15 16:23:27.17192] [4394] [error] Worker 4397 has no heartbeat (50 seconds), restarting (see FAQ for more) |
| 16:24 |
Dyrcona |
[2026-05-15 16:23:27.17223] [4394] [info] Stopping worker 4397 gracefully (120 seconds) |
| 16:24 |
Dyrcona |
Attempting to build a client session as a server Session ID [1778876554.14091134 |
| 16:24 |
Dyrcona |
Again with those messages. |
| 16:26 |
Dyrcona |
I think it is just timing out. |
| 16:40 |
goood |
well, if need's be, it's easy to adjust the timeout (we could make it configurable). in opensrf_server, change the ->start line to something like: app->secrets(scalar time)->heartbeat_timeout(180)->start; |
| 16:40 |
Dyrcona |
i'm reindexing using some code I swiped from Bmagic. :) |
| 16:43 |
Dyrcona |
I may stop services on the other vm using this db server, too, but I don't think it is doing anything. |
| 16:45 |
Dyrcona |
It should not take 2 minutes for an author/title search to return. |
| 16:46 |
Dyrcona |
Adjusting the timeout is a bad idea. |
| 16:54 |
Dyrcona |
This reindex isn't going to finish before 5:00 PM. |
| 16:55 |
goood |
you could watch pg_stat_activity and see if your search is running. if so, I'd say push the typo-free version for others to look at |
| 16:56 |
Dyrcona |
goood: The typo-free version is pushed. That's how I was putting it on the test machine. |
| 16:57 |
goood |
woot |
| 16:57 |
Dyrcona |
I plan to make a clean branch before sharing for real. |
| 16:57 |
goood |
+1 |
| 16:57 |
goood |
again, /very/ happy someone's adding some API fun! |
| 16:57 |
goood |
thank! |
| 16:57 |
Dyrcona |
I have the db upgrade in a different directory for now. |
| 16:57 |
goood |
Dyrcona++ |
| 16:57 |
Dyrcona |
Aw, shucks.... |
| 16:59 |
Dyrcona |
it doesn't look like the search is still running. I'll look for the OPAC or other code that runs search. Maybe it needs to stream or something? |
| 17:00 |
Dyrcona |
But that might wait until Monday. |
| 17:00 |
Bmagic |
Dyrcona++ |
| 17:01 |
|
mmorgan1 left #evergreen |
| 17:01 |
Dyrcona |
I'm signing out for now. |
| 18:39 |
|
jihpringle joined #evergreen |
| 22:55 |
|
book` joined #evergreen |
