Skip to main content

What's new in fast-check 4.6.0?

· 9 min read
Nicolas Dubien
Nicolas Dubien
fast-check maintainer

Until now, stringMatching had no built-in way to cap the length of the produced strings except applying a manual post-filter on the generated values. This release adds a maxLength constraint to ensure that values stay within bounds by construct without relying on excessive and costly filtering.

Our published bundle is also lighter by 17% as it went from 1618 kB to 1344 kB.

Continue reading to explore the detailed updates it brings.

Bound stringMatching with maxLength​

Before 4.6.0, stringMatching could produce strings of arbitrary length depending on the regex. For open-ended patterns such as /[a-z]+@[a-z]+\.[a-z]+/, the generator had no upper limit on the number of characters it could emit. The only limit was the internal notion of size.

When testing code that expects bounded inputs, the only workaround was to add an external .filter() call. Unfortunately, such a call hurt generation efficiency as it required throwing away many already-generated values.

Starting with 4.6.0, you can pass maxLength directly:

fc.stringMatching(/[a-z]+@[a-z]+\.[a-z]+/, { maxLength: 50 });

Under the hood, the regex AST is rewritten before generation. Unbounded quantifiers like * and + are turned into explicit ranges and repetition counts in alternatives are distributed so that the overall length budget is respected. This means most generated candidates already satisfy the constraint without filtering. Even with this optimization, the generator still falls back to post-filtering to ensure invalid values are never exposed.

This approach makes bounded-value generation faster by reducing post-filtering and producing smaller candidate values.

As an example, the call to fc.stringMatching(/^[a-z]+@[a-z]+\.[a-z]+$/, { maxLength: 50 }) will result in the regex /^[a-z]{1,47}@[a-z]{1,47}\.[a-z]{1,47}$/. This example clearly shows that post-filtering is still needed but that the adapted regex significantly reduces the number of invalid values generated.

Lighter bundle​

This release migrates our build pipeline from the TypeScript compiler to Rolldown. We also benefit from improvements in our underlying random number generator library, pure-rand, by upgrading to its latest major version.

These changes reduce the size of our published bundle from 1618 kB down to 1344 kB. It also cuts the file count from 1331 to just 11.

A smaller bundle also translates into faster import times. On our side, we measured import times improved by 1.75× to 2.35× depending on the environment (e.g. GitHub Codespace, Windows workstation).

Deprecation of Random#next(n) and Random#nextInt()​

The method Random#next(n) and the no-argument Random#nextInt() have been marked as deprecated. Calls to these methods can be replaced by calls to Random#nextInt(min, max).

In the long run, we plan to eventually deprecate the whole Random class in favor of the instances directly coming from pure-rand. By doing so, we hope to remove an unnecessary indirection layer for faster random value generation. We also want to let users choose the distribution they use in their arbitraries.

Changelog since 4.5.0​

The version 4.6.0 is based on version 4.5.3.

Features​

  • (PR#6599) Add basic maxLength support to stringMatching
  • (PR#6600) Better clamp on regexes when maxLength on stringMatching
  • (PR#6687) Deprecate Random::next(n) and Random::nextInt()

Fixes​

  • (PR#6502) Bug: Bad d.ts import in BuildInversedRelationsMapping
  • (PR#6578) Bug: Don't crash when stringifying detached ArrayBuffers
  • (PR#6700) Bug: Fix object unmapper and depth computation for special keys
  • (PR#6432) CI: Move all dependencies to dev on examples/
  • (PR#6443) CI: Migrate to Docusaurus v4 configuration format
  • (PR#6456) CI: Enable persist-credentials in add-contributor workflow
  • (PR#6501) CI: Bump module in tsconfig to node18
  • (PR#6548) CI: Fix zizmor ignore config line numbers
  • (PR#6554) CI: Drop tests against Node 20
  • (PR#6563) CI: Fix check_publish status to be success on no error
  • (PR#6565) CI: Add create release workflow
  • (PR#6610) CI: Rework pnpm configuration
  • (PR#6619) CI: Add PR template enforcement workflow
  • (PR#6622) CI: Skip Netlify doc publish on PRs
  • (PR#6625) CI: Run PR template check without approval
  • (PR#6623) CI: Skip PR template check for Renovate bot
  • (PR#6638) CI: Bundle fast-check using rolldown
  • (PR#6662) CI: Rework configuration of examples
  • (PR#6683) CI: Add Claude Code GitHub Action workflow
  • (PR#6684) CI: Add configuration for pre and post tool Claude hooks
  • (PR#6690) CI: Refine GH Action triggering CLAUDE
  • (PR#6693) CI: Configure another Claude model
  • (PR#6703) CI: Add top-level permissions: {} to workflows missing it
  • (PR#6704) CI: Refactor Claude workflow custom instructions configuration
  • (PR#6705) CI: Add SessionStart hook to ensure dependencies are installed
  • (PR#6597) Clean: Drop runkit file
  • (PR#6640) Clean: Drop unused "tsd" in package.json
  • (PR#6441) Doc: Release note for 4.5.0
  • (PR#6442) Doc: Replace generic blog tags with feature-specific taxonomy
  • (PR#6458) Doc: Add adamni21 as doc contributor
  • (PR#6496) Doc: Refine npm keywords
  • (PR#6514) Doc: Skill for JavaScript testing expert
  • (PR#6516) Doc: Add note to avoid overusing filter and fc.pre
  • (PR#6517) Doc: Update testing skill to recommend mimicking existing test structure
  • (PR#6523) Doc: Add PR template requirement to Copilot instructions
  • (PR#6522) Doc: Add note on complementary testing approaches
  • (PR#6524) Doc: Add snapshot vs screenshot guidance
  • (PR#6527) Doc: Push to install missing tooling
  • (PR#6530) Doc: Clearer guidelines for constraints of arbs in skill
  • (PR#6526) Doc: Add AI-powered testing documentation
  • (PR#6529) Doc: Add testing-library and browser testing part in skill
  • (PR#6528) Doc: Add bigint type preference for integer computations in skill
  • (PR#6531) Doc: Add TDD fashion thinking in skill
  • (PR#6553) Doc: Add josephjunker as doc contributor
  • (PR#6561) Doc: Add page on "What is property-based testing" and modify "Why property-based testing"
  • (PR#6605) Doc: Drop Snyk link on Readme
  • (PR#6603) Doc: Update CONTRIBUTING.md for AI
  • (PR#6572) Doc: Rework issue templates
  • (PR#6613) Doc: Update PR template
  • (PR#6634) Doc: Rework bug-report template
  • (PR#6635) Doc: Rework regression-report template
  • (PR#6652) Doc: Update Readme to point to npmx
  • (PR#6659) Doc: Update home to link to npmx
  • (PR#6696) Doc: Add rushelex as code contributor
  • (PR#6448) Performance: Optimize RunDetailsFormatter array allocations
  • (PR#5718) Performance: Import less from pure-rand
  • (PR#6679) Performance: Bump pure-rand to v8
  • (PR#6446) Performance: Replace loose equality by strict one
  • (PR#6444) Performance: Slightly faster code for RunExecution
  • (PR#6437) Refactor: Replace fileURLToPath patterns with import.meta.*
  • (PR#6567) Refactor: Remove ErrorWithCause, use Error directly
  • (PR#6621) Refactor: Replace glob package with native Node.js fs.glob
  • (PR#6675) Refactor: Drop @rollup/plugin-replace for rolldown builtin
  • (PR#6550) Security: Fix zizmor template-injection findings
  • (PR#6472) Test: Provide cause when doc generation fails
  • (PR#6381) Test: Migrate test-types to vitest
  • (PR#6507) Test: Filter ESM-only packages from CommonJS mode checks
  • (PR#6453) Typo: Fix typo for dictionary arbitrary constraint maxKeys
  • (PR#6552) Typo: Replace flatMap with chain in error message