This sets a precident for how we handle `attribute` calls, which aren't
backed by a database column. We should not take this as a conscious
decision on how to handle them, and this can change when we make
`attribute` public if we have better ideas in the future.
As the composed attributes API gets fleshed out, I expect the
`persistable_attributes` method to change to
`@attributes.select(&:persistable).keys`, or some more performant
variant there-of. This can probably go away completely once we fully
move dirty checking into the attribute objects once it gets moved up to
Active Model.
Fixes#18407
It is up to the TransactionManager keep the state of current transaction, so after it commits it needs to copy any remaning record to the next current transaction
This is cropping up all over the place. After a brief dive, I'm really
not sure why we have `arel.bind_values` at all. A cursory grep didn't
reveal where they're actually being assigned (it's definitely in AR, not
in Arel). I'd like to dig further into it, as I'm fairly certain we
don't actually need it, we just need a way for the predicate builder to
communicate merged binds upstream.
Fixes#18414
The reason this bug occured is that we never actually check to see if
this column has changed from it's default, since it was never assigned
and is not mutable.
It appears I was wrong in b301c40224c6d15b539dbcc7485adb44d810f88c, with
my statement of "there is no longer a case where a given value would
differ from the default, but would not already be marked as changed."
However, I chose not to revert the deletion of
`initialize_internals_callback` from that commit, as I think a solution
closer to where the problem lies is less likely to get erroneously
removed. I'm not super happy with this solution, but it mirrors what is
being done in `_update_record`, and a fix for one should work for the
other.
I toyed with the idea of changing the definition of `changed?` on the
type to `changed_in_place?`. If we type cast the raw value, it'll break
a test about updating not modifying the lock column if nothing else was
changed. We could have the definition check if `raw_old_value` is `nil`,
but this feels fragile and less intention revealing. It would, however,
have the benefit of cleaning up old data that incorrectly persisted as
`nil`.
Fixes#18422
When trying to checkout connection from connection pool,
checkout()(and checkout_and_verify) verify whether the connection
is active or not.
And, if the connection is not active, connection adapters try to
reconnect to server. And, if database is down at this moment,
reconnect fails and exception is raised.
(Ex: Mysql2::Error: Can't connect to local MySQL server through socket xxx)
But, ConnectionPool does not catch the exception, but leaks current
disconnected connection to @connection.
So, if database's temporary down happens several times and exceeds
the number of connection pool(5 by default), activerecord will be
no more available, even if database server is already recovered.
This patch fix it by catching exception and releasing connection.
The new test/docs further explain the conflicts that can happen when
mixing `:if`/`:unless` options with `:only`/`:except` options in
`skip_before_action`.
The gist is that "positive" filters always have priority over negative
ones.
The previous commit already showed that `:only` has priority over `:if`.
This commit shows that `:if` has priority over `:except`.
For instance, the following snippets are equivalent:
```ruby
skip_before_action :some_callback, if: -> { condition }, except: action
```
```ruby
skip_before_action :some_callback, if: -> { condition }
```
Test case for using skip_before_filter with the options :only and :if
both present. In this case, the :if option will be ignored and :only
will be executed.
Closes#14549 (the commit was cherry-picked from there).
