Rails - Oct 2008 - Polymorphic Associations in Rails 2 not configurable

This has been bugging me for some time.

The implementation for polymorphic associations in Rails is *still*
using the record_type, record_id model where record type has to be the
entire written out class name.

All of this logic is hardcoded deep in ActiveRecord, making it very
difficult to deviate from this.

For example, what if I wanted to have the "type" column be a tinyint
unsigned that mapped with the name of the class in some hash? Or what if
I wanted to infer the type from a different column all together?

Repeating a long classname for every row in a database is repetitive and
can eat up database size. For example, "SomeLongModuleName::ModelName"
could be replaced with "1", which would then map to that name. That
would be 1 byte for the TinyInt(3) implementation vs N bytes for
actually spelling out the class name *every* time.

Whats frustrating me is the inability to easily configure this, or even
to patch this to get it to work properly. Is anyone also having issues
with this or is there anything out there to give you more control over
polymorphic associations?

Aryk Grosz
Mixbook.com
-- 
Posted via http://www.ruby-forum.com/.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

You could just build your own associations solution instead of trying to
patch the polymorphic associations.

Also, if you have indexes for the columns, the time "searching" for
the row
isn''t really different than it would be for a number instead of a
varchar.
The database should hit the result row in the same time.

On Tue, Oct 7, 2008 at 11:07 PM, Aryk Grosz <
rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org> wrote:
>
> This has been bugging me for some time.
>
> The implementation for polymorphic associations in Rails is *still*
> using the record_type, record_id model where record type has to be the
> entire written out class name.
>
> All of this logic is hardcoded deep in ActiveRecord, making it very
> difficult to deviate from this.
>
> For example, what if I wanted to have the "type" column be a
tinyint
> unsigned that mapped with the name of the class in some hash? Or what if
> I wanted to infer the type from a different column all together?
>
> Repeating a long classname for every row in a database is repetitive and
> can eat up database size. For example,
"SomeLongModuleName::ModelName"
> could be replaced with "1", which would then map to that name.
That
> would be 1 byte for the TinyInt(3) implementation vs N bytes for
> actually spelling out the class name *every* time.
>
> Whats frustrating me is the inability to easily configure this, or even
> to patch this to get it to work properly. Is anyone also having issues
> with this or is there anything out there to give you more control over
> polymorphic associations?
>
> Aryk Grosz
> Mixbook.com
> --
> Posted via http://www.ruby-forum.com/.
>
>
-- 
Maurício Linhares
http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ (en)
João Pessoa, PB, +55 83 8867-7208

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

Yes you could just build your own for this task or that task, but then 
you start adding function after function until you realize why 
activerecord made associations in the first place (getters, setters, 
callbacks, reflections). =)

I tried patching polymorphic associations, its not easy, because you 
have to change the behavior in like 10 different places. Its not pretty.

Im not worried about the time searching for the record, Im frustrated 
about using 30 bytes to store "ThisLongClassNameBlahBlah" vs 1 byte to
store a number that maps to that name for every record.

Also, I''d have to add a "_type" column even though you could
infer the
classname based off another column, thus making things anything but DRY.


Maurício Linhares wrote:
> You could just build your own associations solution instead of trying to
> patch the polymorphic associations.
> 
> Also, if you have indexes for the columns, the time "searching"
for the
> row
> isn''t really different than it would be for a number instead of a 
> varchar.
> The database should hit the result row in the same time.
> 
> On Tue, Oct 7, 2008 at 11:07 PM, Aryk Grosz <
> rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org> wrote:
> 
>> For example, what if I wanted to have the "type" column be a
tinyint
>> to patch this to get it to work properly. Is anyone also having issues
>> with this or is there anything out there to give you more control over
>> polymorphic associations?
>>
>> Aryk Grosz
>> Mixbook.com
>> --
>> Posted via http://www.ruby-forum.com/.
>>
>>
> 
> --
> Maurício Linhares
> http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ 
> (en)
> João Pessoa, PB, +55 83 8867-7208
-- 
Posted via http://www.ruby-forum.com/.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

On Tue, Oct 7, 2008 at 11:38 PM, Aryk Grosz <
rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org> wrote:
> Also, I''d have to add a "_type" column even though you
could infer the
> classname based off another column, thus making things anything but DRY.
>
>
>
Could you explain this a little bit more?

I guess i don''t get the idea you''re trying to reach here. How
could you
infer the classname based on another column without having the classname in
that other column?

-- 
Maurício Linhares
http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ (en)
João Pessoa, PB, +55 83 8867-7208

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

Sure Mauricio, here''s an example.

Lets say you have a "products" table. You have product_type table
column
in that table that you use for single table inheritance. So lets say you 
have the value "book" as the product_type.

Now you want to have a  pricing for this product, so you do:

 belongs_to :pricing, :polymorphic => true

So now you need a pricing_type with a class name of "BookPricing", or
in
other words "#{product_type.classify}Pricing". You will always follow 
this convention since I wouldnt have BookPricing for a product_type of 
"poster"

But wait! There''s more. Now I want another polymorphic association for 
"details" for that product. So I do:

 belongs_to :detail, :polymorphic => true

So, now I need a "detail_type" column of "BookDetail" or in
other words
"#{product_type.classify}Detail".

Now I have two extra string columns that are not needed and are making 
things much less DRY.

Aryk

Maurício Linhares wrote:
> On Tue, Oct 7, 2008 at 11:38 PM, Aryk Grosz <
> rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org> wrote:
> 
>> Also, I''d have to add a "_type" column even though
you could infer the
>> classname based off another column, thus making things anything but
DRY.
>>
>>
>>
> Could you explain this a little bit more?
> 
> I guess i don''t get the idea you''re trying to reach here.
How could you
> infer the classname based on another column without having the classname 
> in
> that other column?
> 
> --
> Maurício Linhares
> http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ 
> (en)
> João Pessoa, PB, +55 83 8867-7208
-- 
Posted via http://www.ruby-forum.com/.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

Well, that''s definitely a "complicated" example :)

I guess you should try to figure out with the core rails guys where
you could hack the code to do just that (what would be infer the name
of the real association from something else), i haven''t looked at the
code myself, but it shouldn''t be that hard with the right guidance.

Also, if you want to roll your own solution, it isn''t that hard and
there''s also plenty of work already written about the matter, one of
the best options is Martin Fowler''s "Patterns of Enterprise
Application Architeture" (
http://martinfowler.com/eaaCatalog/index.html ), there are 3 sections
about object-relational mapping tool structure.

On Wed, Oct 8, 2008 at 1:30 AM, Aryk Grosz
<rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org>
wrote:
>
> Sure Mauricio, here''s an example.
>
> Lets say you have a "products" table. You have product_type table
column
> in that table that you use for single table inheritance. So lets say you
> have the value "book" as the product_type.
>
> Now you want to have a  pricing for this product, so you do:
>
>  belongs_to :pricing, :polymorphic => true
>
> So now you need a pricing_type with a class name of
"BookPricing", or in
> other words "#{product_type.classify}Pricing". You will always
follow
> this convention since I wouldnt have BookPricing for a product_type of
> "poster"
>
> But wait! There''s more. Now I want another polymorphic association
for
> "details" for that product. So I do:
>
>  belongs_to :detail, :polymorphic => true
>
> So, now I need a "detail_type" column of "BookDetail"
or in other words
> "#{product_type.classify}Detail".
>
> Now I have two extra string columns that are not needed and are making
> things much less DRY.
>
> Aryk


-- 
Maurício Linhares
http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ (en)
João Pessoa, PB, +55 83 8867-7208

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

Thanks for the input,

I wouldnt say this is exactly a complicated example. Its basically where 
you could infer the polymorphic class from STI type sort of thing. Im 
pretty sure many other people run into it.

The bottom line is that there is no customizablility.

The problem is that the code has sql strings generated with 
"#{variable}_type = X which makes it very hard to replace with a 
function to control the outputted polymorphic sql condition.

I wonder if there is anybody else that has run into the same problems.


Maurício Linhares wrote:
> Well, that''s definitely a "complicated" example :)
> 
> I guess you should try to figure out with the core rails guys where
> you could hack the code to do just that (what would be infer the name
> of the real association from something else), i haven''t looked at
the
> code myself, but it shouldn''t be that hard with the right
guidance.
> 
> Also, if you want to roll your own solution, it isn''t that hard
and
> there''s also plenty of work already written about the matter, one
of
> the best options is Martin Fowler''s "Patterns of Enterprise
> Application Architeture" (
> http://martinfowler.com/eaaCatalog/index.html ), there are 3 sections
> about object-relational mapping tool structure.
> 
> On Wed, Oct 8, 2008 at 1:30 AM, Aryk Grosz
> <rails-mailing-list-ARtvInVfO7ksV2N9l4h3zg@public.gmane.org> wrote:
>>
>> So, now I need a "detail_type" column of
"BookDetail" or in other words
>> "#{product_type.classify}Detail".
>>
>> Now I have two extra string columns that are not needed and are making
>> things much less DRY.
>>
>> Aryk
> 
> 
> 
> --
> Maur�cio Linhares
> http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/ 
> (en)
> Jo�o Pessoa, PB, +55 83 8867-7208
-- 
Posted via http://www.ruby-forum.com/.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Talk" group.
To post to this group, send email to
rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
To unsubscribe from this group, send email to
rubyonrails-talk+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-talk?hl=en
-~----------~----~----~----~------~----~------~--~---