There are two repositories for the ActiveRecord adapter. The one you're in currently, activerecord-cockroachdb-adapter, is the CockroachDB specific ActiveRecord code. Users install this alongside ActiveRecord then use CockroachDBAdapter to initialize ActiveRecord for their projects.
This adapter extends the PostgreSQL ActiveRecord adapter in order to override and monkey-patch functionality.
In CockroachDB, create two databases to be used by the ActiveRecord test suite: activerecord_unittest and activerecord_unittest2.
CREATE DATABASE activerecord_unittest;
CREATE DATABASE activerecord_unittest2;
It is best to have a Ruby environment manager installed, such as rbenv, as Rails has varying Ruby version requirements. If you are using rbenv, you then install and use the required Ruby version.
(Alternatively, one can use ./docker.sh build/teamcity-test.sh
to run
tests similarly to TeamCity. The database is destroyed between each
test file.)
Install rbenv with ruby-build on MacOS:
brew install rbenv ruby-build
echo 'eval "$(rbenv init - zsh)"' >> ~/.profile
Install rbenv with ruby-build on Ubuntu:
sudo apt-get install rbenv
echo 'eval "$(rbenv init - zsh)"' >> ~/.profile
git clone https://github.com/sstephenson/ruby-build.git ~/ruby-build
sh ~/ruby-build/install.sh
Use rbenv to install version of Ruby specified by .ruby-version
.
rbenv install
rbenv rehash
Using bundler, install the dependencies of Rails.
bundle install
Then, to run the full test suite with an active CockroachDB instance:
bundle exec rake test
To run specific ActiveRecord tests, set environment variable TEST_FILES_AR
. For example, to run ActiveRecord tests test/cases/associations_test.rb
and test/cases/ar_schema_test.rb.rb
TEST_FILES_AR="test/cases/associations_test.rb,test/cases/ar_schema_test.rb" bundle exec rake test
To run specific CockroachDB Adapter tests, set environment variable TEST_FILES
. For example, to run CockroachDB Adpater tests test/cases/adapter_test.rb
and test/cases/associations/left_outer_join_association_test.rb
TEST_FILES="test/cases/adapter_test.rb,test/cases/associations/left_outer_join_association_test.rb" bundle exec rake test
To run a specific test case, use minitest's -n
option to run tests that match a given pattern. All minitest options are set via the TESTOPTS
environemnt variable. For example, to run test_indexes
from CockroachDB's test/cases/adapter_test.rb
file
TEST_FILES="test/cases/adapter_test.rb" TESTOPTS=`-n=/test_indexes/` bundle exec rake test
By default, tests will be run from the bundled version of Rails. To run against a local copy, set environemnt variable RAILS_SOURCE
. Running against a local copy of Rails can be helpful when try to debug issues.
RAILS_SOURCE="path/to/local_copy" bundle exec rake test
test/config.yml
assumes CockroachDB will be running at localhost:26257 with a root user. Make changes to test/config.yml
as needed.
Loading the full test schema every time a test runs can take a while, so for cases where loading the schema sequentially is unimportant, it is possible to use a backup to set up the database. This is significantly faster than the standard method and is provided to run individual tests faster, but should not be used to validate a build.
First create the template database.
bundle exec rake db:create_test_template
This will create a template database for the current version (ex. activerecord_test_template611
for version 6.1.1) and create a BACKUP
in the nodelocal://self/activerecord-crdb-adapter/#{activerecord_version}
directory.
To load from the template, use the COCKROACH_LOAD_FROM_TEMPLATE
flag.
COCKROACH_LOAD_FROM_TEMPLATE=1 TEST_FILES="test/cases/adapters/postgresql/ddl_test.rb" bundle exec rake test
And the activerecord_unittest
database will use the RESTORE
command to load the schema from the template database.
Currently, only a beta version of Rails is tested. This means that the adapter has been modified in to accommodate unreleased changes. In order to run the tests for Rails 5.1 or 4.2, the test changes will need to be cherry-picked back. Conflicts are mostly only expected for tests that have not yet been added.
Sadly, this does mean that we will have to have multiple versions of the driver for the multiple versions of Rails.
A proposal for the CockroachDB adapter versioning would be to follow
ActiveRecord minor versions. For example, if you use Rails 4.2.5, you
would specify the CockroachDB version ~> 4.2.0
.
Currently the fork is set up to run using TeamCity only on the current master branch, with an alpha build of CockroachDB. it would be even better to be able to test multiple versions of the adapter, and do so against different versions of CockroachDB.
As CockroachDB improves, so do the features that can be supported in ActiveRecord. Many of them are gated by conditions the CockroachDBAdapter has overridden. As these features are completed, these gates should be toggled. Something that would help this process would be linking those issues back to this adapter so that part of the feature completing includes updating the adapter.
I have not investigated if this is already possible, but I would assume no.
A possible way to approach this would be to add a shim to cause any tests that use it to fail, and grep the tests that pass and then skip them.
TODO: Expand on this. Jordan is likely the only person with publishing credentials. I'm not sure if there is anything else other than:
gem build ...
gem publish <output file>
When executing the test suite, each test file will reload fixtures. This drops and creates about 200 tables (2 databases, 100 tables each). Currently there are performance problems that rise from having lots of table descriptors around, cockroachdb/cockroach#20753. At best, we can run test files individually, clear out the CockroachDB data, and restart the node to alleviate this.
Currently, annotations have been added to test files to indicate if it is failing, and some brief details on why. Any annotated failures have been skipped right now for further investigation. The pattern is the following:
# FILE(OK)
indicates that the file is currently passing, with no skips
required.
# FILE(BAD)
indicates that there are failures that have been skipped.
These skips will look like skip(reason) if current_adapter?(:CockroachDBAdapter)
.
# FILE(BROKEN)
indicates that there are failures that have not been
skipped. This is often done if the entirety of a test file is
unsupported.
# FILE(NOT DONE)
indicates files that have not yet been executed,
cleaned up, or skipped until passing.
The purpose of these was to make the tests grep-able while going through all the failures.
Some of the skipped failures are:
-
default:
key is not working for columns in table schema definitions. This causes tests to fail due to unexpected data. -
array:
key is not working for columns in table schema definitions. -
"Salary is not appearing in list"
is being raised in a number of places. Likely during fixture setup. -
sum
function seems to result in a different type in ActiveRecord. Instead of returning a Rubyint
, it returns a Rubystring
. It appears that MySQL2 also does this. A suspected cause might be howDecimal
is handled ifsum
consumes integers and return a decimal. -
Potentially fork the PostgreSQL::SchemaDumper to handle anything specific to CockroachDB, like primary keys or bigints.
-
You can call
@connection.create_table(..., id: :bigint)
, but this will not changes things for CockroachDB (I think...), so it would be not allowed. Even better, our adapter could interpret this and generate the appropriate explicit pkey column. Not sure what string pkeys look like... -
string
types are introspected totext
types. -
A user can do an update, delete, and insert on views.
-
Postgres specific bit strings are not properly supported.
Grepping for FIXME(joey)
, TODO(joey)
, and NOTE(joey)
will yeild
most of the touchpoints including test failures and temporary monkey
patches. Some monkey patches were made directly to Rails, which will
need to be cleaned up.
rvm is an environment manager that lets you manage and swap between multiple versions of Ruby and their dependencies.
bundle is dependency manager that uses a projects Gemfile
(and often
<project>.gemspec
) to manage and load dependencies and their required
versions. When using projects commands are prefixed with
bundle exec ...
. Bundle will ensure that all dependencies are fetched
and used.