splitrb/split
40.4
Weak · 5 November 2026
2.3k
lines of production code
Ruby
primary language
4
measurements over time
What this system is
Features
Add String\#constantize extension for Split
lib/split/extensions/string.rb now provides a String\#constantize method when ActiveSupport is not present, allowing Split to convert string names to their corresponding Ruby constants without requiring the ActiveSupport library.
lib/split/extensions · high confidence
New block randomization, weighted sample, and Whiplash bandit algorithms
Users can now choose from three new experiment allocation strategies in addition to existing options. The Block Randomization algorithm ensures balanced participant distribution by selecting from alternatives with the minimum current count. The Weighted Sample algorithm assigns participants to alternatives based on predefined weights. The Whiplash algorithm implements a multi-armed bandit approach that dynamically favors alternatives with higher completion rates, using a fairness constant to balance exploration and exploitation.
lib/split/algorithms · high confidence
New persistence adapters for cookies, Redis, and dual-context storage
The library now supports pluggable persistence layers beyond the default session storage. Users can store split test results in browser cookies via CookieAdapter (with configurable length and domain), in Redis via RedisAdapter (with configurable namespace, lookup strategy, and TTL), or combine them using DualAdapter to handle different storage strategies for logged-in versus logged-out users. The existing SessionAdapter remains available as the default behavior.
lib/split/persistence · high confidence
Removals
Removed version constant file
The file defining the Multivariate module's version constant has been deleted, removing the explicit version string from the library's codebase.
lib/multivariate · high confidence
Behavioural changes
Dashboard modernization with configurable pagination and improved confidence calculations
The dashboard now supports configurable pagination, allowing users to set the default number of experiments per page via the \dashboard\_pagination\_default\_per\_page\ configuration option. The interface has been modernized by replacing internal structs with dedicated view objects (\DashboardExperiment\, \DashboardAlternative\) for cleaner data presentation. Additionally, the confidence level helper has been updated to handle edge cases like tiny z-scores and numeric strings more robustly, and experiment names are now properly escaped to prevent misnaming issues in winner determination.
lib/split/dashboard · high confidence
Introduce Split gem with configurable Redis persistence and Rails integration
The library now provides a unified entry point (lib/split.rb) that replaces the previous multivariate module, offering configurable Redis connections (accepting URLs, hashes, or instances) and automatic Rails integration via a Railtie. Users can now configure the gem through a standard initializer block, utilize a new caching layer, and benefit from pluggable experiment algorithms and persistence adapters, while the legacy lib/multivariate.rb file has been removed.
lib · high confidence
Modernized Split dashboard with dark mode and filtering
The Split dashboard has been redesigned with a modern UI that includes a light/dark theme toggle (persisted in localStorage), a text-based filter to search experiments by name, and toggles to show/hide active or completed experiments. The visual presentation now features a responsive layout, rounded corners, and distinct styling for headers and experiment cards, improving readability and user interaction.
lib/split/dashboard/public · high confidence
Redesigned Split dashboard with modern UI, cohorting controls, and probability metrics
The Split dashboard has been completely rewritten to provide a modern, responsive interface. Users can now manage experiment lifecycle directly from the UI, including enabling/disabling cohorting, resetting data, reopening finished experiments, and deleting experiments. The view displays detailed metrics for each alternative, including participants, conversion rates, and a new 'Probability of being Winner' calculation alongside the traditional confidence level. The dashboard also supports multiple goals per experiment, allows forcing alternatives for the current user, and displays active browser overrides. A new layout includes a theme toggle and environment indicator.
lib/split/dashboard/views · high confidence
Split A/B testing library rewritten for version 4.0.5
The Split library has been completely rewritten and upgraded to version 4.0.5, introducing a modernized architecture with pluggable persistence adapters (Session, Cookie, Redis, and Dual), a new ExperimentCatalog for centralized experiment management, and a redesigned dashboard built on Sinatra. This major update adds support for combined experiments, configurable algorithms (including a new beta distribution RNG via Rubystats), and enhanced metrics tracking with multiple goals per experiment. The internal storage layer has been refactored into distinct ConfigStorage and RedisStorage components, and the helper methods have been encapsulated to safely mix into Rails controllers and models without polluting the namespace.
lib/split · high confidence
Standardizes development environment with new linting and configuration files
The project now includes dedicated configuration files for code quality and testing, specifically \.rubocop.yml\ (targeting Ruby 3.1), \.codeclimate.yml\, \.csslintrc\, and \eslint.config.js\, alongside a \.rspec\ file to enable profiling. The \Rakefile\ has been updated to include a shebang and \frozen\_string\_literal\ comment, and \.gitignore\ has been expanded to exclude common development artifacts like \coverage/\, \node\_modules/\, and IDE files.
(repo-wide) · high confidence
Test coverage
Added comprehensive test suite for Split core components; Added test coverage for new persistence adapter implementations; Added tests for block randomization, weighted sample, and whiplash algorithms; Added tests for dashboard view objects and pagination helpers.
Dependencies
Modernize Ruby dependencies and add JavaScript linting
The project has updated its core Ruby dependencies to require Ruby 3.1+ and added explicit runtime requirements for \redis\ (\>= 4.2), \sinatra\ (\>= 1.2.6), \rubystats\ (\>= 0.3.0), \matrix\, \bigdecimal\, and \cgi\. Development dependencies have been refreshed, including \rack-test\ (\~\> 2.0), \rspec\ (\~\> 3.7), and \rake\ (\~\> 13), while the \multivariate\ gemspec was removed. Additionally, a new \package.json\ introduces ESLint 10.8.0 and related packages for JavaScript linting, and the Gemfile now pins \concurrent-ruby\ (\< 1.3.9) and \json\ (\< 3) to ensure compatibility with newer Rails versions.
(dependencies) · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 41 → 40 (-0.7)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.
Lenses
- Code Health 60 → 60 (-0.4)
- Architecture 93 → 93 (-0.0)
- Maturity 50 → 55 (+4.4)
- Readiness 57 → 57 (+0.1)
- Security 35 → 32 (-2.8)
- Accessibility 35 → 35 (+0.0)
Resolved (5)
- Duplicate intent with different names. Both methods appear to handle the logic of finding an existing experiment or creating a new one if it doesn't exist. In Ruby conventions, find_or_create is standard, while find_or_initialize usually implies returning a new unsaved object. Having both suggests either redundant functionality or confusing semantic differences not immediately obvious from the signature.
- Inconsistent naming strategy for override sources. override_alternative is generic, while the others explicitly state the source (by_params, by_cookies). It is unclear if override_alternative is a fallback, a dispatcher, or a duplicate of one of the specific methods. This creates ambiguity for the caller about which method to use for a general override lookup.
- Inconsistent setter naming and granularity. set_p_winner and set_completed_count are specific to statistical data. set_field and set_prob_field are generic and vague. It is unclear what set_field sets compared to set_prob_field or set_p_winner. This lack of specificity makes the API harder to use correctly.
- Off-boarding risk: anonymized user #1
- Outdated: codeclimate-test-reporter
New (1)
- Off-boarding risk: anonymized user #1
Changes since last survey
- 11 commits — 10 feature/other, 1 fixes
By area
- .github/workflows — 4 commits
- (repo) — 3 commits
- lib/split — 3 commits
- (root) — 1 commit
Notable commits
- fix: Merge pull request #769 from splitrb/fix-ci-matrix
- change: Build Ruby 3.1 on ubuntu-24.04
- change: Build Ruby 4.1 preview
- change: Check the recalculation interval before reading any counts
- change: Load a dashboard experiment's alternatives in minimum roundtrips to Redis
- change: Merge pull request #770 from splitrb/dashboard-roundtrips
- change: Merge pull request #771 from splitrb/remove-split-cache
- change: Only pin Rails and json when RAILS_VERSION is set
- change: Pin each CI job to its Rails minor
- change: Remove Split::Cache and make config.cache a no-op
- change: Update actions/checkout to v5
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
splitrb/split was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 5 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 1f7fecdd00dfb69431fc657f4a2328a8fe987038 — the exact code this score is about.
- Scored under rubric-2026.11.2 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.2+preprod-5c4e66d9318b.