fix: Add .js extensions to relative imports in ESM TypeScript projects

[mohammedahmed18]

Summary

• Fixes ERR_MODULE_NOT_FOUND errors in generated tests for ESM TypeScript projects
• Adds .js extensions to relative imports (e.g., ./module → ./module.js)
• Only affects ESM projects; CommonJS projects unchanged

Problem

Generated tests were importing from relative paths without .js extensions:

import TreeNode from '../../injector/topology-tree/tree-node';

This causes ERR_MODULE_NOT_FOUND in Node.js ESM, which requires explicit extensions for relative imports, even when source files are .ts.

Solution

Added add_js_extensions_to_relative_imports() function that:

• Adds .js to relative imports without extensions
• Preserves imports with existing extensions
• Leaves node_modules imports unchanged
• Integrated into test generation pipeline (ESM projects only)

Test Plan

• Added 7 unit tests for the new function
• All 35 module_system tests pass
• All 315 JavaScript language tests pass
• Manually verified fix resolves the issue for trace 17751b8f-fa61-48bc-bdee-b924f0c7afc4

Affected Trace IDs

17751b8f-fa61-48bc-bdee-b924f0c7afc4, 3b985200-a906-4c54-a685-df40361d6b2c, 91795877-3ccf-482c-86bd-748834b76f6e, 0298c59c-8980-4aed-b05d-b94940a6544f, ec2864a4-0de0-4ce9-9ec8-b545c82a4f53

🤖 Generated with Claude Code

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Codeflash Bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[claude]

Claude finished @mohammedahmed18's task in 5m 31s —— View job

---

PR Review Summary

• Triage: LARGE (59 lines of production code)
• Lint and typecheck: prek passes; mypy errors are pre-existing (unrelated to this PR)
• Stale threads: none
• Code review
• Duplicate detection
• Optimization PRs evaluated

---

Prek Checks

All ruff checks pass. No issues to fix.

Code Review

Regex misses import type syntax (medium)

import type { Foo } from './module' is a common TypeScript pattern. The regex's (?:\w+) alternative matches type as a default import name, but then \s+from\s+ fails to match since { Foo } from follows — so these imports never get .js added.

module_system.py:546 — the pattern could be extended:

# Before
r"(import\s+(?:(?:\{[^}]*\})|(?:\*\s+as\s+\w+)|(?:\w+))\s+from\s+)..."
# After
r"(import\s+(?:type\s+)?(?:(?:\{[^}]*\})|(?:\*\s+as\s+\w+)|(?:\w+))\s+from\s+)..."

Fix this →

Mixed default+named imports not matched (low)

import Default, { named } from './module' won't get .js added. Less common in generated tests but a gap worth noting.

Inconsistent import placement in support.py (low)

ModuleSystem is added to the existing from codeflash.languages.javascript.module_system import (...) block, but add_js_extensions_to_relative_imports is imported separately inside the if block at line ~2040. Both should be in the same import statement.
Fix this →

Docstrings on new functions (low)

add_js_extensions_to_relative_imports and the inner add_extension both have docstrings. Per project conventions (CLAUDE.md): "Do not add docstrings to new or changed code unless the user explicitly asks for them."

Duplicate Detection

import_resolver.py has extension-related logic but it resolves file paths on disk, not import strings. No duplicates detected.

Test Coverage

7 new unit tests added and all pass. Missing coverage for:

• import type { Foo } from './module' (the regex gap above)
• import Default, { named } from './module' (mixed imports)

---

Optimization PRs

• PR ⚡️ Speed up method TestResults.timing_coefficient_of_variation by 338% in PR #1949 (cf-1082-benchmark-noise-floor) #1955 (TestResults.timing_coefficient_of_variation +338%): ✅ Merged
• PR ⚡️ Speed up method TestResults.total_passed_runtime by 20% in PR #1949 (cf-1082-benchmark-noise-floor) #1954 (TestResults.total_passed_runtime +20%): Conflicting after ⚡️ Speed up method TestResults.timing_coefficient_of_variation by 338% in PR #1949 (cf-1082-benchmark-noise-floor) #1955 merge — will auto-resolve once rebased
• PR ⚡️ Speed up function speedup_critic by 10% in PR #1949 (cf-1082-benchmark-noise-floor) #1956 (speedup_critic +10%): CI still pending
• PR ⚡️ Speed up function fmt_delta by 11% in PR #1941 (cf-compare-copy-benchmarks) #1943 (fmt_delta +11%): CI failures are in trace benchmark tests, appear to be flaky test pollution unrelated to the optimization change — left open

---

Last updated: 2026-04-01

[Saga4]

Not sure if this is existing, the question remains that we are not following some conviction on function name and claude code end up creating the same functions quickly in some other helper file than searching for existing logics