rspec-turbo 0.1.0 → 0.1.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 6e75805ec30e8c00d22ab5db21967fb3ddfdeb0ab8f6edabad1047f6b8d3d895
-  data.tar.gz: 3081d8c6f070131b441e4d00c368396b3e19cd48a2d79ed13b60a427a30c0033
+  metadata.gz: 64eb6a3af09b14eb69de00d8e41a4d47ac536b31bb9d3ec4c20e6b27e9f6b7fc
+  data.tar.gz: 5d062a69b01ac90ed24144dfebb7fa295519d6d09c3ec33278801823e01a0871
 SHA512:
-  metadata.gz: 6f29717c3d1da16330b2019c4ff0cd7043f846ac0bc803716ddd86bdc338b382c41037c58ed68b9182a9fcad66ac0406d1ca994f76c3c3e2402e2533bff767b0
-  data.tar.gz: e66d30be4d8d35ba555cd4da8b5d950f19bf88fcdbd5df1c9feff10c14aab6956dbbd244d83cb82e3783264cfaedb58fd2ea80d05a7536f94691e684d808f78e
+  metadata.gz: 0e6909eb453c1488a4f98332667b52f941d609ffd67f415ec81c4bbdd2696258ea2ba10f92470735a9b35925d549229eb4f45f8c8bc5a926f3caa87719e3962b
+  data.tar.gz: 5e3ad70b33cb0008ae5478c5179f86c0b0ab17466dcba4b8154b0802afecdcdc4f2bc86c847b48145c7e0dccdb27a5e29052debb8a961dc79c9099a3f27600b6

data/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Changelog
 
+## [0.1.1] - 2026-06-12
+
+### Added
+- Three entry points: the `rspec-turbo` binary plus a `spec:turbo` Rake task
+  (reachable as both `rake spec:turbo` and `rails spec:turbo`), registered in
+  Rails apps through a Railtie.
+- `coverage:merge` Rake task that collates per-worker SimpleCov result files
+  with `SimpleCov.collate`, emitting JSON on CI (`JSONFormatter`) and HTML
+  locally (`HTMLFormatter`); glob overridable via `RSPEC_TURBO_COVERAGE_GLOB`.
+
+### Fixed
+- Workers and the dry-run now force `RAILS_ENV=test`, so they always target the
+  `*_test` databases `DbSetup` created — instead of relying on the project's
+  `rails_helper` to set it (which broke when launched via `rake`/`rails`, where
+  the parent process boots in development).
+
 ## [0.1.0] - 2026-06-12
 
 Initial extraction from the single-file `turbo.rb` runner into a gem.
@@ -13,12 +29,6 @@ Initial extraction from the single-file `turbo.rb` runner into a gem.
   enabled by default (disable with `RSPEC_TURBO_NO_PROFILE=1`) and safe outside
   Rails (times examples without counting SQL when ActiveSupport is absent).
 - JUnit XML output (`JUNIT_DIR`) and SimpleCov coverage merging (`COVERAGE=1`).
-- Three entry points: the `rspec-turbo` binary plus a `spec:turbo` Rake task
-  (reachable as both `rake spec:turbo` and `rails spec:turbo`), registered in
-  Rails apps through a Railtie.
-- `coverage:merge` Rake task that collates per-worker SimpleCov result files
-  with `SimpleCov.collate`, emitting JSON on CI (`JSONFormatter`) and HTML
-  locally (`HTMLFormatter`); glob overridable via `RSPEC_TURBO_COVERAGE_GLOB`.
 
 ### Fixed (versus the original script)
 - `DbSetup#show_log` referenced an undefined `w` variable on failure.

data/README.md

@@ -14,7 +14,7 @@ dashboard, and a report that tells you exactly which folders are slowing you
 down.
 
 ```sh
-bundle add rspec-turbo --group test
+bundle add rspec-turbo --group "development, test"
 bundle exec rspec-turbo
 ```
 
@@ -96,10 +96,10 @@ does one thing and tunes hard for it:
 
 ## Install
 
-Add it to the `:test` group of your Gemfile:
+Add it to the `:development, :test` group of your Gemfile:
 
 ```ruby
-group :test do
+group :development, :test do
   gem "rspec-turbo"
 end
 ```
@@ -108,6 +108,14 @@ end
 bundle install
 ```
 
+> **Why `:development` too?** `rake` and `rails` commands run in the
+> **development** environment by default, and Bundler only loads gems from the
+> groups for the current environment. The `spec:turbo` / `coverage:merge` tasks
+> are registered by a Railtie, so the gem must be loaded in `development` for
+> `rake spec:turbo` / `rails spec:turbo` to exist. (The `bundle exec rspec-turbo`
+> binary works from any group.) If you must keep it in `:test` only, run the
+> tasks with `RAILS_ENV=test rails spec:turbo`.
+
 ## Usage
 
 ```sh
@@ -165,6 +173,32 @@ parse argv → DbSetup → FileDiscovery → BatchPlanner → Executor (pool)
 - **`simplecov` + `simplecov_json_formatter`** — only when `COVERAGE=1`
   (see [Coverage](#coverage-optional) below).
 
+### Parallel test databases (required)
+
+Each worker runs with its own `TEST_ENV_NUMBER` (`1`, `2`, … up to the worker
+count), so **every worker needs its own database** — otherwise they trample each
+other. Make the test database name include `TEST_ENV_NUMBER` in
+`config/database.yml`:
+
+```yaml
+default: &default
+  adapter: postgresql
+  username: postgres
+  password: postgres
+
+test:
+  <<: *default
+  database: myapp_test<%= ENV["TEST_ENV_NUMBER"] %>
+```
+
+That resolves to `myapp_test1`, `myapp_test2`, … — one per worker. (Plain
+`rails` commands, with no `TEST_ENV_NUMBER`, still use `myapp_test`.)
+
+On the first run — or after a schema change — `rspec-turbo` creates and loads
+all of them for you (`db:drop db:create db:schema:load db:seed`, one process per
+slot, each with its own `TEST_ENV_NUMBER`). Later runs reuse them via the
+schema-fingerprint cache; force a rebuild with `RSPEC_TURBO_FORCE_SETUP=1`.
+
 ## Environment variables
 
 | Variable | Default | Purpose |
@@ -274,6 +308,18 @@ speak RuboCop pick up the same rules; the canonical runner is `standardrb`.
 VS Code is pre-wired (`.vscode/settings.json`) to format on save with Standard
 via the Ruby LSP extension.
 
+## Troubleshooting
+
+**`Unrecognized command "spec:turbo"` (`Rails::Command::UnrecognizedCommandError`)**
+— the gem isn't loaded in the environment your command runs in. `rake`/`rails`
+default to `development`, so move the gem to `group :development, :test` (see
+[Install](#install)), or run `RAILS_ENV=test rails spec:turbo`. The
+`bundle exec rspec-turbo` binary works regardless.
+
+**Workers fail with "database … already exists" / connection clashes** — your
+`config/database.yml` test database name isn't keyed by `TEST_ENV_NUMBER`. See
+[Parallel test databases](#parallel-test-databases-required).
+
 ## Contributing
 
 Issues and pull requests are welcome. Run `bundle exec rake` before opening a

data/lib/rspec_turbo/batch_planner.rb

@@ -59,7 +59,7 @@ module RSpecTurbo
       File.open(Config.dry_run_log, "w") do |err_file|
         IO.popen(
           # COVERAGE=0 keeps SimpleCov from contaminating the JSON on stdout.
-          [{"COVERAGE" => "0", "TEST_ENV_NUMBER" => "1"},
+          [{"RAILS_ENV" => "test", "COVERAGE" => "0", "TEST_ENV_NUMBER" => "1"},
             "bundle", "exec", "rspec", "--dry-run", "--format", "json",
             *@rspec_options, *@files.map { |f| "spec/#{f}" }],
           err: err_file, &:read

data/lib/rspec_turbo/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module RSpecTurbo
-  VERSION = "0.1.0"
+  VERSION = "0.1.1"
 end

data/lib/rspec_turbo/worker.rb

@@ -40,6 +40,10 @@ module RSpecTurbo
 
     def self.env(slot, progress_file)
       {
+        # Force the test env so each worker connects to the *_test database
+        # DbSetup created, regardless of how the run was launched (a Rake/Rails
+        # task boots in development) or whether rails_helper sets it.
+        "RAILS_ENV" => "test",
         "TEST_ENV_NUMBER" => slot.to_s,
         "RSPEC_TURBO_PROGRESS_FILE" => progress_file,
         "COVERAGE" => ENV.fetch("COVERAGE", "0")

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: rspec-turbo
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.1.1
 platform: ruby
 authors:
 - thadeu