09:37 <mrgreenfur> Is there a simple way to add a computed column to a model? I'd like to show averages of a related model without re-querying each time it's requested

10:37 <sigurd-> Hello, does sequel work with ruby 1.9.3? I'm getting ArgumentError: argument out of range (Sequel::InvalidValue) from time.rb:202:in `make_time' when calling .first on a result

11:02 <sigurd-> I tried installing the latest version from github, but I'm still getting the same error

11:04 <leucos> sigurd-: may be this comes from your db adaptor.? Sequel is doing great for me with 1.9.3. Have a traceback ?

11:05 <sigurd-> yeah, here: http://pastie.org/3264041

11:05 <Caesium> is it a mysql datetime which is set to zero or something?

11:06 <sigurd-> yes, that it posible, I have some values I'm not using

11:06 <sigurd-> it worked just fine earlier though

11:07 <Caesium> could be 1.9.3 is stricter about zero dates

11:08 <sigurd-> yeah, I'll try to add some dates just to see

11:14 <jeremyevans> mrgreenfur: Usually what I'd do in your case is add a dataset method that uses select_append to add an additional column to return.

11:15 <jeremyevans> sigurd-: If you are using invalid dates/times in MySQL, you probably want to set DB.convert_invalid_date_time = nil

11:16 <leucos> When Chuck Norris has an ORM problem, he always asks jeremyevans

11:17 <sigurd-> ok, thanks. But I'm gonna try fiksing it by inserting the dates I'm actually going to store there first

11:17 <jeremyevans> sigurd-: Certainly working with valid dates is preferable :)

11:23 <sigurd-> It's working fine now. Thanks for the help. I think the problem was that I/Rails might have changed the table to NOT NULL on the datetimes

11:25 <jeremyevans> sigurd-: And I know with certain MySQL configs (maybe even the default), that means if you attempt to insert a NULL value, it will automatically change it to a bogus value

11:25 <sigurd-> yeah, it listed it as 0000-00-00 00:00:00

11:26 <mrgreenfur> jeremyevans: thanks, i understand the dataset method part, but is there a way to use a subquery in select_append? I'd like to get a computed value from a related table

11:26 <jeremyevans> sigurd-: I require more sane behavior from the databases I use in production, which is why I use postgres :)

11:27 <jeremyevans> mrgreenfur: Sure: select_append(DB[:table].select{sum(foo)}.where(:bar=>'baz'))

11:27 <jeremyevans> mrgreenfur: with an .as(:column) at the end so you get a consistent column alias

11:29 <sigurd-> jeremyevans: Yeah, I've used postgres in classes, but I'm used to mysql so I've been using it for personal projects

11:30 <jeremyevans> sigurd-: It's never to late to switch. Postgres is better in almost every way.

11:30 <sigurd-> I'll remember that next time :)

11:30 <mrgreenfur> jeremyevans: thanks!

11:41 <mrgreenfur> jeremyevans: is there a place to put that in the model definition? Maybe when defining the on_to_many?

11:43 <jeremyevans> mrgreenfur: def_dataset_method/dataset_module can be called anywhere in the model definition

11:47 <mrgreenfur> hrm, this is getting complicated. How bad of an idea is it to save computed values in a table and updated them in code? :)

11:56 <jeremyevans> mrgreenfur: Not bad at all, if you do it correctly in a database trigger

11:56 <jeremyevans> mrgreenfur: Doing it in an after hook in a model is much less desireable

11:57 <mrgreenfur> jeremyevans: Ahh, awesome. I was afraid I would violate a normal-form or somesuch. I'll drop a trigger in when the related table gets an insert/delete.

11:58 <jeremyevans> Well, it might actually violate a normal form, but denormalization for performance reasons is considered good sense my most DBAs

11:58 <mrgreenfur> i can live with that. Browseing the whole database and having it do a bunch of calculations on each element is killing me.

11:59 <mrgreenfur> err whole table

11:59 <mrgreenfur> plus, i've never used a trigger, this'll be fun

11:59 <jeremyevans> mrgreenfur: Are you doing something simple like a counter or a total field?

12:00 <mrgreenfur> jeremyevans: an average of ratings for an associated model that saves individual ratings

12:03 <jeremyevans> mrgreenfur: What database are you using?

12:05 <mrgreenfur> jeremyevans: mysql; I've got a page that shows the elements sorted by their average rating and it was probably poorly coded (sorting in ruby and not in mysql)

12:07 <jeremyevans> mrgreenfur: Basically, you'll want an after insert/update/delete trigger on the associated model table that updates the current model table with the average

12:07 <jeremyevans> mrgreenfur: It'll make inserts/updates/deletes more expensive, but if your access pattern is mostly read only, that's probably not a big deal

12:07 <mrgreenfur> jeremyevans: yeah! then I can use .order(:average_rating) and it'll be much faster

12:19 <mrgreenfur> jeremyevans: this works great. Thanks much for your help!

13:48 <sts> hello folks. is there a way to override a models setter methods to eg. encrypt a password before saving a dataset?

13:53 <Caesium> you could override the = operation?

13:53 <Caesium> def password=(password)

13:53 <Caesium> in the model?

13:54 <Caesium> hm probably best to call the argument something else though ;)

13:58 <jeremyevans> Caesium: Just define the method and call super

13:59 <sts> Caesium: ok and how do i assign the new value? Is super(new_value) still the way to go?

14:00 <Caesium> jeremyevans: so it'd be something like self[:password] = dostuffto arg \n super

14:00 <Caesium> or just super(dostuffto arg) ?

14:04 <sts> thank you

14:05 <jeremyevans> Caesium: super is preferable

14:06 <jeremyevans> Caesium: self[:password] = dostuff will work in most cases, but if you have a plugin that overrides password=, then things will probably break

14:39 <GitHub190> [sequel] xaviershay opened issue #427: Generating impossible queries. https://github.com/jeremyevans/sequel/issues/427: http://git.io/v-ZYBg

15:34 <msch> is there a way to have a common base class that inherits from Sequel::Model where I can put some common stuff?

15:35 <msch> problem is, sequel tries to read from a table for this "abstract" class

15:42 <jeremyevans> msch: BaseClass = Class.new(Sequel::Model)

15:42 <jeremyevans> msch: but you can just add methods to Sequel::Model itself

15:43 <jeremyevans> msch: If all of your classes inherit from the same base class, I'd just modify Sequel::Model itself

15:43 <msch> jeremyevans: well it's not all :)

15:43 <jeremyevans> If only some do, then the base class approach works. Either that, or put it in a module and include that module in all the classes

15:44 <msch> so does Class.new approach skip the "inherited" callback, or how does this work?

15:44 <jeremyevans> msch: The inherited stuff looks for the name of the class. If the class has no name (and anonymous classes don't), then it doesn't associate it with a dataset

15:45 <msch> clever :)

15:57 <msch> hm... i tried overwriting MySubclass.new to include some Module sometimes, and it gets called for .new and for .create, but not when I'm fetching via e.g. .all

15:58 <jeremyevans> msch: That's expected. You probably want an after_initialize hook instead

15:59 <msch> jeremyevans: it's weird but i need to change the instantiated model's class at runtime

15:59 <msch> jeremyevans: because if i include a Module in each new object ruby will soon run out of memory

16:00 <jeremyevans> msch: Can you paste what you are doing. Because you shouldn't need to be including a module with every object

16:01 <msch> jeremyevans: my entities can have tags and certain tags should mix in code

16:01 <msch> jeremyevans: e.g. Person is tagged with Evil, then overwrite lend_money

16:02 <jeremyevans> msch: Personally, I tend to use conditionals for that, but I suppose if you have a lot of different tags, that can get out of hand

16:02 <msch> jeremyevans: yep we've got like 20 tags.

16:03 <jeremyevans> msch: I don't think it should run ruby out of memory though. You should just be doing an extend(Evil) inside after_initialize

16:03 <msch> ok that sounds nice

16:03 <jeremyevans> msch: Basically, put each tagged behavior in a separate module, and extend the classes in after initialize based on the tags

16:03 <jeremyevans> extend the instances I mean

16:07 <msch> jeremyevans: yeah that sounds good. i was just afraid that ruby would create a new shadow class each time that never gets GCed

16:08 <jeremyevans> msch: Nope. It won't be great in terms of performance, but it shouldn't break the GC

16:11 <msch> wouldn't post_load be a better place to put it?

16:11 <jeremyevans> msch: No, that's only called in a specific case

16:11 <msch> ok

16:11 <msch> thanks a lot by the way!

16:22 <chris__> The after_initialize hook in my model blows up if I define it as private, but the other hooks work just fine this way. Is this a bug?

16:23 <jeremyevans> chris__: hooks should be public methods

16:23 <jeremyevans> chris__: that's how Sequel defines them

16:23 <jeremyevans> chris__: The reason it fails is that Sequel now uses allocate instead of initialize for objects retreived from the database

16:25 <chris__> jeremyevans: Hmm, ok. It makes more sense to me for hooks to be private, since it feels like other classes shouldn't be calling them, but if the implementation requires it I guess it's not a big deal.

17:49 <GitHub73> [sequel] jeremyevans pushed 1 new commit to master: http://git.io/5VhFbw

17:49 <GitHub73> [sequel/master] Change IN/NOT IN handling of empty arrays (Fixes #427) - Jeremy Evans