This is odd... I have a few models that after some time just start doing SELECT "my_tablemy_belongs_to_table".* sort of thing. (Almost like it is trying to reference a join table, but missing a _ between the table names.)
In other words, things work for hours, days or weeks then suddenly goes from SELECT "my_table".* to SELECT "my_tablemy_belongs_to_table".*
There are a few activerecord settings (table_name, table_name_prefix maybe some others) that could do this. Might these be getting changed by accident (perhaps because your model has a method whose name clashes with an active record method) ?
Fred
The associated FROM clause also changes to match. Hundreds of exceptions per minute then ensue and all heck breaks loose. ;')
I perhaps am running into a namespace issue or something(?)... how does a model get triggered into changing it's associated table name in runtime? Our temp fix it to just restart the service/app but that's obviously not sustainable. We had to do it several times today.
Thoughts?
ruby 2.5.0p21 (2018-02-18 revision 62472) [x86_64-linux]
rails 5.1.6
pg 1.0.0
Thanks!
Phil
Thank you Fred and Colin.
Somehow table_name seems to be getting set to something else... I can simulate how it breaks but I don't know where/why it is breaking in the production code...
UserAccount.table_name = "blah"
=> "blah"
UserAccount.first
Traceback (most recent call last):
1: from (irb):2
ActiveRecord::StatementInvalid (PG::UndefinedTable: ERROR: relation "blah" does not exist
LINE 1: SELECT "blah".* FROM "blah" ORDER BY "blah"."id" ASC LIMIT ...
^
: SELECT "blah".* FROM "blah" ORDER BY "blah"."id" ASC LIMIT $1)
Something in Rails 5.1.x is appending to 'table_name' in the model during some condition?
I can't easily reproduce. I don't see anywhere in the code where I'm operating on table_name directly.
(still digging, I hope everyone is having a good Saturday, btw :')
Phil
Ah, I've solved this. We've migrated code that used arrays, to strings. We were still using "<<", though, to concatenate strings. Let me give you some examples to understand the issue....
>> a = MyItem.table_name
=> "my_items"
>> b = a
=> "my_items"
>> c = b
=> "my_items"
>> d = c
=> "my_items"
>> d += ", other_items"
=> "my_items, other_items"
>> d
=> "my_items, other_items"
>> c
=> "my_items"
All good, right? Now, let's use '<<':
>> a = MyItem.table_name
=> "my_items"
>> b = a
=> "my_items"
>> c = b
=> "my_items"
>> d = c
=> "my_items"
>> d << ", other_items"
=> "my_items, other_items" <---- Same result! Looks right.
>> c
=> "my_items, other_items"
>> MyTable.table_name
=> "my_items, other_items" <--- Oh noes, the result was inherited all the way down to the table_name of the model!
Now the MyItem model is borked until the app is restarted.
The malformed 'from' clause was a bit of a red herring and that was easily resolved, btw. (A 'join(', ')' was skipped because the construction of the from-clause was a string not an array any more in our code.)
The '<<' method appears to work the same as the String method ".concat(something)" . It also makes all other variables inherit the same value.
Another possible solution is to use '.dup', for what it is worth:
>> d = c.dup
=> "my_items"
>> d << ", other_items"
=> "my_items, other_items"
>> c
=> "my_items"
I hope this helps somebody if they run into a similar problem! (I don't claim this is a Ruby bug or such, just perhaps not very well documented feature.)
Phil
