These should allow external code to run blocks of user code to do
"work", at a similar unit size to a web request, without needing to get
intimate with ActionDipatch.
When the `source_type` option is passed to a has_many through, the
resulting `Reflection` will be an instance of `PolymorphicReflection`
and not `ThroughReflection`, meaning that it will ignore the scopes that
it needs to apply from the through reflections. This adds an additional
delegation call to remedy this. I've been finding the reflection code
completely impenetrable lately, it could use some major love.
Fixes#22726
Currently `exec_query` raises `NoMethodError` when executing no result
queries (`INSERT`, `UPDATE`, `DELETE`, and all DDL) in mysql2 adapter.
```
irb(main):002:0> conn.execute("create table t(a int)")
(43.3ms) create table t(a int)
=> nil
irb(main):003:0> conn.execute("insert into t values (1)")
(19.3ms) insert into t values (1)
=> nil
irb(main):004:0> conn.exec_query("insert into t values (1)")
SQL (28.6ms) insert into t values (1)
NoMethodError: undefined method `fields' for nil:NilClass
```
Some tests does not work for unprepared statements.
Add `if ActiveRecord::Base.connection.prepared_statements` and fix a
regex for fix tests failure with `prepared_statements: false`.
record.id_was is nil in after_create/after_save, so we should use
id in these cases.
While this logic feels incomplete, the existing update_record uses the same
logic:
2fda4e0874/activerecord/lib/active_record/relation.rb (L83)
This logic was originally added for a similar problem:
updates not working with after_create hook.
See: 482f8c15b1d699c95bfbc3d836f674a09c0d9031
Followup to #23581Fixes#23844
I ran into an issue where validations on a suppressed record were
causing validation errors to be thrown on a record that was never going
to be saved.
There isn't a reason to run the validations on a record that doesn't
matter.
This change moves the suppressor up the chain to be run on the `save` or
`save!` in the validations rather than in persistence. The issue with
running it when we hit persistence is that the validations are run
first, then we hit persistance, and then we hit the suppressor. The
suppressor comes first.
The change to the test was required since I added the
`validates_presence_of` validations. Adding this alone was enough to
demonstrate the issue. I added a new test to demonstrate the new
behavior is explict.
I was encountered remaining `:binary_testings` table by tests failure.
When remaining `:binary_testings` table, never reach `drop_table` due to
`create_table` in the test always fails.
In order to fix issue #17621 we added a check to validations that
determined if a record should be validated. Based on the existing tests
and behavior we wanted we determined the best way to do that was by
checking if `!record.peristed? || record.changed? || record.marked_for_destruction?`
This change didn't make it into a release until now. When #23790 was
opened we realized that `valid?` and `invalid?` were broken and did not
work on persisted records because of the `!record.persisted?`.
While there is still a bug that #17621 brought up, this change was too
drastic and should not be a RC blocker. I will work on fixing this so
that we don't break `valid?` but also aren't validating parent records
through child records if that parent record is validate false. This
change removes the code changes to validate and the corresponding tests.
It adds tests for two of the bugs found since Rails 5 beta2 release.
Fixes#17621
When `ActiveRecord::Coders::JSON` serialization is used and the default of the column returns `''` it raises the following error:
```
JSON::ParserError: A JSON text must at least contain two octets!
```
If MySQL is running in non-strict mode, it returns an empty string as column default for a text column:
```ruby
def extract_default
if blob_or_text_column?
@default = null || strict ? nil : ''
end
end
```
Since `''` is invalid JSON, there shouldn't be an attempt to parse it, it should be treated like nil.
ActiveRecord::Coders::JSON should behave consistently for all possible non-user-set column default values.
Fixes#23645
When you're using an `attr_accessor` for a record instead of an
attribute in the database there's no way for the record to know if it
has `changed?` unless you tell it `attribute_will_change!("attribute")`.
The change made in 27aa4dd updated validations to check if a record was
`changed?` or `marked_for_destruction?` or not `persisted?`. It did not
take into account virtual attributes that do not affect the model's
dirty status.
The only way to fix this is to always validate the record if the
attribute does not belong to the set of attributes the record expects
(in `record.attributes`) because virtual attributes will not be in that
hash.
I think we should consider deprecating this particular behavior in the
future and requiring that the user mark the record dirty by noting that
the virtual attribute will change. Unfortunately this isn't easy because
we have no way of knowing that you did the "right thing" in your
application by marking it dirty and will get the deprecation warning
even if you are doing the correct thing.
For now this restores expected behavior when using a virtual attribute
by always validating the record, as well as adds tests for this case.
I was going to add the `!record.attributes.include?(attribute)` to the
`should_validate?` method but `uniqueness` cannot validate a virtual
attribute with nothing to hold on to the attribute. Because of this
`should_validate?` was about to become a very messy method so I decided
to split them up so we can handle it specifically for each case.
habtm join tables commonly have two id columns and it's OK to make those
two id columns a primary key. This commit eliminates the warnings for
join tables that have this setup.
ManageIQ/manageiq#6713
Add assertions to MySQL `ORDER BY FIELD()` with empty data.
These tests examine to sanitize `ORDER BY FIELD()` with empty data
appropriately.
```ruby
Tag.order(['field(id, ?)', []]).to_sql
# => SELECT "tags".* FROM "tags" ORDER BY field(id, NULL)
Tag.order(['field(id, ?)', nil]).to_sql
# => SELECT "tags".* FROM "tags" ORDER BY field(id, NULL)
```
This resolves a bug where if the primary key used is not `id` (ex:
`uuid`), and has a `validates_uniqueness_of` in the model, a uniqueness error
would be raised. This is a partial revert of commit `119b9181ece399c67213543fb5227b82688b536f`, which introduced this behavior.
- Previously it used to show error message
<"undefined method `limit_value' for {:title=>\"Rails\"}:Hash">
- Now it shows following error message.
>> Post.where.not(name: 'DHH').or(name: 'Tenderlove')
ArgumentError: You have passed Hash object to #or. Pass an ActiveRecord::Relation object instead.
- Fixes#23714.
When passing in an array of different types of objects to `where`, it would only take into account the class of the first object in the array.
PriceEstimate.where(estimate_of: [Treasure.find(1), Car.find(2)])
# => SELECT "price_estimates".* FROM "price_estimates"
WHERE ("price_estimates"."estimate_of_type" = 'Treasure' AND "price_estimates"."estimate_of_id" IN (1, 2))
This is fixed to properly look for any records matching both type and id:
PriceEstimate.where(estimate_of: [Treasure.find(1), Car.find(2)])
# => SELECT "price_estimates".* FROM "price_estimates"
WHERE (("price_estimates"."estimate_of_type" = 'Treasure' AND "price_estimates"."estimate_of_id" = 1)
OR ("price_estimates"."estimate_of_type" = 'Car' AND "price_estimates"."estimate_of_id" = 2))
Saw the `merge!` and had to prove to myself that the parent model's local_stored_attributes was not being changed when stored_attributes is called on a child model. Proved to be working as expected but this test is probably still useful to keep around.
Previously if you used `t.foreign_key` twice within the same
`create_table` block using the same `to_table`, all statements except
the final one would fail silently. For example, the following code:
def change
create_table :flights do |t|
t.integer :from_id, index: true, null: false
t.integer :to_id, index: true, null: false
t.foreign_key :airports, column: :from_id
t.foreign_key :airports, column: :to_id
end
end
Would only create one foreign key, on the column `from_id`.
This commit allows multiple foreign keys to the same table to be created
within one `create_table` block.
In MySQL, PostgreSQL, Oracle and SQLServer, a value over the limit
cannot be inserted or updated (See #23522).
In SQLite3, a value is inserted or updated regardless of the limit.
We should avoid a string value truncation in uniqueness validation.
- Tests on Travis are randomly failing because schema_migrations table
does not exist in teardown block.
- Also checked that all other places where we have used
`ActiveRecord::SchemaMigration.delete_all` we have rescued it, so used
it here also. This failure was not specifically related to the test
added in this PR but to overall compatibility migration tests, so
adding as separate commit.
When changing the PK for a record which has a uniqueness validation on
some other attribute, Active Record should exclude itself from the
validation based on the PK value stored on the DB (id_was) instead of
its new value (id).
- Corrects an incorrect exception message when using accepts_nested_attributes_for
- Removes rescue/reraise behavior introduced in #19077
- Adds has_many & has_one, nested_attributes test case specifying the message that
should be conveyed with an exception raised because one of the nested attributes provided is unknown
`assert_in_delta` in `timestamp_test.rb` causes an intermittent test
failure. It looks like that caused by subseconds truncation in MySQL 5.5.
Example:
```
1) Failure:
TimestampTest#test_touching_many_attributes_updates_them [/home/travis/build/rails/rails/activerecord/test/cases/timestamp_test.rb:125]:
Expected |2016-02-06 09:22:10 +0000 - 2016-02-06 09:22:09 UTC| (1.000837185) to be <= 1.
```
This reverts commit 99801c6a7b69eb4b006a55de17ada78f3a0fa4c1.
Ultimately it doesn't matter whether `add_index` or `t.index` are used
in the schema dumper in any meaningful way. There are gems out there
which hook into the old behavior for things like indexing materialized
views. Since the reverted commit doesn't seem to add much benefit,
there's no reason for us to break these gems.
add some documentation and add 4 tests regarding error vs. warning behavior
fix a typo when referring to the message
go back to default in tests so that ordering is not important. use a constant instead of method. fix assert_nothing_raised call. use self.klass to allow per class configuration
remove logger warn assets as that is tested elsewhere. pass error_on_ignore through find_each and find_in_batches also.
add blocks to the finds so that the code is actually executed
put the setting back to default in an ensure
Add a changelog entry
We had previously updated this to attempt to map over whatever was
passed in, so that additional types like range and array could benefit
from this behavior without the time zone converter having to deal with
every known type.
However, the default behavior of a type is to just yield the given value
to `map`, which means that if we don't actually know how to handle a
value, we'll just recurse infinitely. Since both uses of `map` in this
case occur in cases where we know receiving the same object will
recurse, we can just break on reference equality.
Fixes#23241.
When a `before_commit` callback raises, the database is rolled back but
AR's record of the current transaction is not, leaving the connection in
a perpetually broken state that affects all future users of the
connection: subsequent requests, jobs, etc. They'll think a transaction
is active when none is, so they won't BEGIN on their own. This manifests
as missing `after_commit` callbacks and broken ROLLBACKs.
This happens because `before_commit` callbacks fire before the current
transaction is popped from the stack, but the exception-handling path
they hit assumes that the current transaction was already popped. So the
database ROLLBACK is issued, but the transaction stack is left intact.
Common cause: deadlocked `#touch`, which is now implemented with
`before_commit` callbacks.
What's next:
* We shouldn't allow active transaction state when checking in or out
from the connection pool. Verify that conns are clean.
* Closer review of txn manager sad paths. Are we missing other spots
where we'd end up with incorrect txn state? What's the worst that can
happen if txn state drifts? How can we guarantee it doesn't and
contain the fallout if it does?
Thanks for @tomafro for expert diagnosis!
This reverts commit 9f3730a516f30beb0050caea9539f8d6b808e58a, reversing
changes made to 2637fb75d82e1c69333855abd58c2470994995d3.
There are additional issues with this commit that need to be addressed
before this change is ready (see #23377). This is a temporary revert in
order for us to have more time to address the issues with that PR,
without blocking the release of beta2.
If `AR.primary_key` is called for a table who has composite primary key,
the method returns `nil`. This behavior sometimes generates invalid SQL.
The first time developers notice to invalid SQL is when they execute
SQL. This commit enables developers to know they are doing something
dangerous as soon as possible.
This reverts commit f6db31ec16e42ee7713029f7120f0b011d1ddc6c.
Reason:
Scope names can very easily conflict, particularly when sharing Concerns
within the team, or using multiple gems that extend AR models.
It is true that Ruby has the ability to detect this with the -w option, but the
reality is that we are depending on too many gems that do not care about Ruby
warnings, therefore it might not be a realistic solution to turn this switch on
in our real-world apps.
The commit which originally added this behavior did not consider that
doing `Subclass.new` does not actually populate the `type` field in the
attributes (though perhaps it should). We simply need to not use the
defaults for STI related things unless we are instantiating the base
class.
Fixes#23285.
* 5-0-beta-sec:
bumping version
fix version update task to deal with .beta1.1
Eliminate instance level writers for class accessors
allow :file to be outside rails root, but anything else must be inside the rails view directory
Don't short-circuit reject_if proc
stop caching mime types globally
use secure string comparisons for basic auth username / password
- We don't need the select scope added by user as we only want to max
timestamp and size of the collection. So we already know which columns
to select.
- Additionally having user defined columns in select scope blows the cache_key
method with PostGreSQL because it needs all `selected` columns in the group_by
clause or aggregate function.
- Fixes#23038.
- Using `references` or `belongs_to` in migrations will always add index
for the referenced column by default, without adding `index:true` option
to generated migration file.
- Users can opt out of this by passing `index: false`.
- Legacy migrations won't be affected by this change. They will continue
to run as they were before.
- Fixes#18146
This fixes incorrect assumptions made by e991c7b that we can assume the
DB is already casting the value for us. The enum type needs additional
information to perform casting, and needs a subtype.
I've opted not to call `super` in `cast`, as we have a known set of
types which we accept there, and the subtype likely doesn't accept them
(symbol -> integer doesn't make sense)
Close#23190
When updating an associated record via nested attribute hashes the
reject_if proc could be bypassed if the _destroy flag was set in the
attribute hash and allow_destroy was set to false.
The fix is to only short-circuit if the _destroy flag is set and the
option allow_destroy is set to true. It also fixes an issue where
a new record wasn't created if _destroy was set and the option
allow_destroy was set to false.
CVE-2015-7577
- When relations return no result or 0 result then cache_key should
handle it gracefully instead of blowing up trying to access
`result[:size]` and `result[:timestamp]`.
- Fixes#23063.
Mysql has a weird bug where it cannot index a string column of utf8mb4 if it is over a certain character limit. To get compatibility with msql we can add a limit to the key column. 191 characters is a very long key, it seems reasonable to limit across all adapters since using a longer key wouldn't be supported in mysql.
Thanks to @kamipo for the original PR and the test refactoring.
Conversation: https://github.com/rails/rails/pull/23009#issuecomment-171416629
- Before this patch if we try to find cache_key of a loaded but empty
collection it used to give error because of trying to call `updated_at`
on `nil` value generated by
`collection.max_by(×tamp_column).public_send(timestamp_column)`.
- This commit fixes above error by checking if size is greater than zero
or not.
This reverts commit 224eddfc0eeff6555ae88691306e61c7a9e8b758, reversing
changes made to 9d681fc74c6251d5f2b93fa9576c9b2113116680.
When merging the pull request, I misunderstood `has_secure_token` as declaring a model
has a token from birth and through the rest of its lifetime.
Therefore, supporting conditional creation doesn't make sense. You should never mark a
model as having a secure token if there's a time when it shouldn't have it on creation.
The code that set the from clause was removed in
bdc5141652770fd227455681cde1f9899f55b0b9. I did not give any reason for
doing so. My assumption was that I intended to change it to use the
clause objects, but forgot. We appeared to not have test coverage for
this case.
Fixes#22996
When you are using scopes and you chaining these scopes it is hard to
know which are the values that are incompatible. This way you can read
the message and know for which values you need to look for.
[Herminio Torres]
This reverts commit 5d41cb3bfd6b19833261622ce5d339b1e580bd8b.
This implementation does not properly handle cases involving predicates
which are not associated with a bind param. I have the fix in mind, but
don't have time to implement just yet. It will be more similar to #22823
than not.
While the predicates are an arel equality node where the left side is a
full arel attribute, the binds just have the name of the column and
nothing else. This means that while splitting the predicates can include
the table as a factor, the binds cannot. It's entirely possible that we
might be able to have the bind params carry a bit more information (I
don't believe the name is used for anything but logging), and that is
probably a worthwhile change to make in the future.
However the simplest (and likely slightly faster) solution is to simply
use the indices of the conflicts in both cases. This means that we only
have to compute the collision space once, instead of twice even though
we're doing an additional array iteration. Regardless, this method isn't
a performance hotspot.
Close#22823.
[Ben Woosley & Sean Griffin]
Given a default_scope on a parent of the current class, where that
parent is not the base class, the parent's STI condition would become
attached to the evaluated default scope, and then override the child's
own STI condition.
Instead, we can treat the STI condition as though it is a default scope,
and skip it in this situation: the scope will be merged into the base
relation, which already contains the correct STI condition.
Fixes#22426.
While the commit message (and changelog example) in
5e0b555b453ea2ca36986c111512627d806101e7 talked about sibling classes,
the added test had a child ignore its parent's scoping, which seems less
reasonable.
This PR addresses the issue described in https://github.com/rails/rails/pull/22967#issuecomment-170251635. If the database is non empty and has no new migrations than `db:migrate` will not set the environment. This PR works by always setting the environment value on successful `up` migration regardless of whether or not a migration was actually executed.
When running passing condition assertions in the same test the user had already
been saved at that point.
Split out so we have a not yet persisted user.
Rename condition tests to improve clarity a bit.
Fix the NoMethodErrors introduced in 224eddf, when adding conditional token creation.
The model declarations but the column wasn't added to the schema.
This is an alternate implementation to #22875, that generalizes a lot of
the logic that type decorators are going to need, in order to have them
work with arrays, ranges, etc. The types have the ability to map over a
value, with the default implementation being to just yield that given
value. Array and Range give more appropriate definitions.
This does not automatically make ranges time zone aware, as they need to
be added to the `time_zone_aware` types config, but we could certainly
make that change if we feel it is appropriate. I do think this would be
a breaking change however, and should at least have a deprecation cycle.
Closes#22875.
/cc @matthewd
You cannot use `update_attributes` on models that do not have a primary key. Since SchemaMigration versions are guaranteed to be unique (they have a unique index on them) we can safely use them as a primary key.
This PR introduces a key/value type store to Active Record that can be used for storing internal values. It is an alternative implementation to #21237 cc @sgrif @matthewd.
It is possible to run your tests against your production database by accident right now. While infrequently, but as an anecdotal data point, Heroku receives a non-trivial number of requests for a database restore due to this happening. In these cases the loss can be large.
To prevent against running tests against production we can store the "environment" version that was used when migrating the database in a new internal table. Before executing tests we can see if the database is a listed in `protected_environments` and abort. There is a manual escape valve to force this check from happening with environment variable `DISABLE_DATABASE_ENVIRONMENT_CHECK=1`.
In Pg and Sqlite3, `:text` and `:binary` have variable unlimited length.
But in MySQL, these have limited length for each types (ref #21591, #21619).
This change adds short-hand methods for each text and blob types.
Example:
create_table :foos do |t|
t.tinyblob :tiny_blob
t.mediumblob :medium_blob
t.longblob :long_blob
t.tinytext :tiny_text
t.mediumtext :medium_text
t.longtext :long_text
end
Originally `connection#create` had aliased to `connection#insert` in PG
adapter. But it was broken by #7447. Re-alias `create` to `insert` for
fixing it.
uuid-ossp extension is alreadly enabled on test schema.
And `disable_extension!('uuid-ossp', connection)` can be a cause of test failure.
`ActiveRecord::StatementInvalid: PG::UndefinedFunction: ERROR: function uuid_generate_v1() does not exist`
will happen depending on the execution order.
I realized that `first(2)`, etc. was unnecessarily querying for the
records when they were already preloaded. This was because
`find_nth_with_limit` can not know which `@records` to return because
it conflates the `offset` and `index` into a single variable, while
the `@records` only needs the `index` itself to select the proper
record.
Because `find_nth` and `find_nth_with_limit` are public methods, I
instead introduced a private method `find_nth_with_limit_and_offset`
which is called internally and handles the `loaded?` checking.
Once the `offset` argument is removed from `find_nth`,
`find_nth_with_limit_and_offset` can be collapsed into
`find_nth_with_limit`, with `offset` always equal to `offset_index`.
A `(?:var)?binary` with default '' is a correct definition.
Remove `missing_default_forged_as_empty_string?` method for fixing this
issue because this method is a workaround for older mysql legacy adapter
(19c99ac, f7015336).
- uses instance defined level if no custom local log level defined
- Keeps track of local log level per [ thread + object-instance ]
- prevents memory leakage by removing local level hash key/value on #silence method exit
- avoids the use of Thread local variables
In MySQL 5.7.5 and up, ONLY_FULL_GROUP_BY affects handling of queries
that use DISTINCT and ORDER BY. It requires the ORDER BY columns in the
select list for distinct queries, and requires that the ORDER BY include
the distinct column.
See https://dev.mysql.com/doc/refman/5.7/en/group-by-handling.html
This reverts commit 537ac7d6ade61e95f2b70685ff2236b7de965bab, reversing
changes made to 9c9c54abe08d86967efd3dcac1d65158a0ff74ea.
Reason:
The way we preload associations will change the meaning of GROUP BY
operations. This is illustrated in the SQL generated by the added
test (failing on PG):
Association Load:
D, [2015-12-21T12:26:07.169920 #26969] DEBUG -- : Post Load (0.7ms) SELECT "posts".* FROM "posts" LEFT JOIN comments ON comments.post_id = posts.id WHERE "posts"."author_id" = $1 GROUP BY posts.id ORDER BY SUM(comments.tags_count) [["author_id", 1]]
Preload:
D, [2015-12-21T12:26:07.128305 #26969] DEBUG -- : Post Load (1.3ms) SELECT "posts".* FROM "posts" LEFT JOIN comments ON comments.post_id = posts.id WHERE "posts"."author_id" IN (1, 2, 3) GROUP BY posts.id ORDER BY SUM(comments.tags_count)
We know the query will return exactly one row for each entry in the
`ids` array, so we can do all the limit/offset calculations on that
array, in advance.
I also split our new ordered-ids behaviour out of the existing
`find_some` method: especially with this change, the conditionals were
overwhelming the actual logic.