Fetch from Upstream #1
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[ci skip]
Fix mysql adapter link in README.md.
Maybe it's time to drop support for those versions of Ruby that are no longer being updated? 2.1 is currently end of life, and 2.2 is going to be end of life'd at the end of March.
As of 31/3/2018, Ruby 2.2 and lower are officially unsupported. Also, Rails 4.1 and lower are completely unsupported as well, so I removed those here. Made stuff much cleaner!
Add Ruby 2.5 to travis build matrix
By default, creating a materialized view causes the associated query to be run immediately. The `create view` statement does not complete until the view is populated. While this is a good default, it may be particularly expensive and potentially not desired during deploys. Specifying `no_data: true`, allows the view to be created without a refresh. It will not be usable until it is manually refreshed.
…umper-output Remove extra space from schema dumper output
`pg` 0.21 + has deprecated PGConn in favor of `PG::Connection`. This change fixes the following deprecation: > The PGconn, PGresult, and PGError constants are deprecated, and will be removed as of version 1.0
When we moved to our own organization from the thoughtbot org, we lost the hound-provided rubocop config associated with the thoughtbot org. This change updates to the prefered hound config format that references our own `.rubocop.yml` file. How did I settle on the rubocop rules? Well, I ran rubocop on the codebase and either fixed what came up or added exceptions. We have lots of line length violations that I didn't fix. For the most part, I like the encouragement to stay at 80 characters, but sometimes it's not worth it. I think getting the hound reminder in the PR is sufficient enough for us to have a conversation.
Ruby 2.6 is out, so I've added it to the matrix. Rails 6 (rails edge) requires 2.5+ so we can skip building it with older rubies.
Addresses #258 - cascade failing to identify dependent views when the `view_to_refresh` is a substring of a dependency.
Some folks add `;` to the end of their view definitions to terminate the statement. When this is combined with `no_data: true`, an exception was being raised because `WITH NO DATA` was being interprested as a separate SQL statement. To fix this, we remove the trailing semicolon from the SQL definition before adding `WITH NO DATA`. Fixes #272
Our existing build matrix was a little out of date. We were missing Rails 6 and Ruby 2.7. I could have just added those and called it a day, but our build matrix was already getting unwieldy. I started by removing Ruby 2.3, which is no longer receiving updates. I then removed Rails versions older than 5.2 for the same reason. Finally, I removed the Rubies between 2.4 and 2.7 on the grounds that testing on the oldest currently-supported ruby and the newest is sufficient. Scenic's Rails and Ruby version support remains unchanged, but we're limiting CI to a reasonable combination based on ongoing support from Ruby and Rails. We're saving build time and CO2 emissions.
Prior to this change, `create_view :greetings, version: 2` would error because it was reversed to `drop_view :greetings, version: 2`. The `drop_view` method in our schema statements does not take a `version` argument because that wouldn't make sense. This change removes the `version` argument while leaving others (i.e. `materialized`).
Previous to this change, using the cascading refresh behavior did not also pass through the option to concurrently refresh as one would expect. This meant that while the named view would be refreshed concurrently when desired, none of its dependencies would. Fixes #276
Semantically, I agree that the changelog for a release is the list of every commit and that what we typically refer to as a changelog these days is really a list of newsworthy changes. However, it seems to me that battle is lost and that people expect this file to be named `CHANGELOG.md`. So here it is.
Postgres 14 [introduced] an internal view pg_stat_statements_info, which
we need to exclude from our dumper because including it causes loading
the schema to fail:
PG::DuplicateTable: ERROR: relation "pg_stat_statements_info" already exists
To test this, enable pg_stat_statements in the dummy database. This is a
reasonable addition as the majority of rails apps are using this
extension and it's cleaner and less error-prone than adding a fake view
of the same name in a specific test for this case.
To support this view, we're currently specifically excluding it in the
query used in the Postgres adapter to load views. Rather then excluding
it by name, we could do an extension of our current exclusion for
extension-owned views, `AND c.relname NOT IN (SELECT extname FROM
pg_extension)` by excluding any views with an extension name included
rather than exact matches. I'm not convinced that this is more robust
than excluding exact names has been shown to be by this bug.
Fixes #348
* Fix ERB deprecation warning example warning: ``` lib/generators/scenic/model/model_generator.rb:39: warning: Passing safe_level with the 2nd argument of ERB.new is deprecated. Do not use it, and specify other arguments as keyword arguments. ``` followed Ruby standard library style to make distinction on ERB version check: ruby/ruby@3406c5d
This flag was introduced in #368 but hasn't been in a release yet. I was writing up the next release's changelog and I wasn't thrilled with having an underscore in a CLI flag. There's something to be said for `--replace_view` matching the schema method that the generated method invokes but I think `--replace` is still better.
We're not an active enough project to keep up with master as part of a required build task. I don't want to turn off any contributors who see a red build against an unreleased version of Rails. I think we could rework these as a separate workflow that reports status appropriately but is not required, but I'll leave that for another day.
You cannot use `replace_view` on a materialized view. You can use `replace_view` on any standard postgres view that may or may not depend or be depended on by a materialized view.
REFRESH MATERIALIZED VIEW CONCURRENTLY doesn't work for a view which is not populated, even it has unique index. This method is useful to perform concurrent refresh if it's possible. Real use case in Rails will be like this: * Database is loaded from structure.sql to run specs, but the view isn't populated yet so concurrent refresh is not possible. We want to fallback to normal(non-concurrent) refresh in this case. * After initial population, views are expected to be refreshed concurrently. It's possible because the view is already populated. Co-authored By: Derek Prior <[email protected]>
We had been using Hound CI and a custom RuboCop config to encourage adhering to our style guide. In the last couple of years, Standard has emerged and gained steam as a "set it and forget it" RuboCop configuration for Ruby. Maintaining our own opinionated RuboCop configuration is not something I care to do any longer. This change: 1. Adds Standard 2. Fixes all[^1] off the issues it detects 3. Adds Standard to CI 4. Removes our hound and RuboCop configuration Adding Standard as an explicit CI step will require compliance for passing builds, which is a departure from our previous setup of only commenting via Hound. I think the time is right for this change given the maturity of auto-fixers and integrations within various Ruby workflows. [^1]: We continue to allow the use of `eval` in our specs
Upcoming changes like [topological sorting][1] have the potential to write views to `db/schema.rb` in an order that does not respect dependencies. This adds a test to ensure that views in all schemas are both included, and in the correct dependency order. [1]: #398
This reverts commit d0f2052. It is failing every build on main.
In order to use the model to `find` individual records, the primary key must be explicitly defined. At least that's my experience with PostgreSQL. Also, see https://arjanvandergaag.nl/blog/view-backed-activerecord-models.html
- Reformat FAQ sections to be H3 to improve hierarchy ans so that they can be linked to directly - Add several known blog posts about Scenic
Also remove ~>0.19 constraint on pg gem, which conflicts with ~>1.1 constraint in Gemfile
Upcoming changes like [topological sorting][1] have the potential to write views to `db/schema.rb` in an order that does not respect dependencies. This adds a test to ensure that views in all schemas are both included, and in the correct dependency order. [1]: #398
Co-authored-by: Mykhaylo Sorochan <[email protected]> Co-authored-by: Edward Loveall <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.