Microsoft Typespec Typespec

Test Emitter Skill

Builds and tests @typespec/http-client-python to validate local changes. Automatically regenerates if the generated code is stale.

Paths

All paths are relative to the http-client-python package root: ~/Desktop/github/typespec/packages/http-client-python

Workflow

Step 1: Build the emitter

cd ~/Desktop/github/typespec/packages/http-client-python
npm run build

Check for TypeScript compilation errors. If the build fails, report the errors to the user and stop — do not proceed to testing.

Step 2: Check if regeneration is needed

Regeneration is needed if ANY of these conditions are true:

1. Generated folder doesn't exist or is empty:

   cd ~/Desktop/github/typespec/packages/http-client-python
   
   if [ ! -d "generator/test/unbranded/generated" ] \
     || [ -z "$(ls -A generator/test/unbranded/generated 2> /dev/null)" ]; then
     echo "Unbranded generated folder is empty - regeneration needed"
   fi
   
   if [ ! -d "generator/test/azure/generated" ] \
     || [ -z "$(ls -A generator/test/azure/generated 2> /dev/null)" ]; then
     echo "Azure generated folder is empty - regeneration needed"
   fi
   

2. No regenerate marker file exists:

   if [ ! -f ".last-regenerate" ]; then
     echo "No regenerate marker - regeneration needed"
   fi
   

3. Emitter source files are newer than the marker:

   if [ -n "$(find emitter/src -newer .last-regenerate -type f 2> /dev/null | head -1)" ]; then
     echo "Emitter source changed since last regenerate - regeneration needed"
   fi
   

Step 3: Regenerate if needed

If any condition from Step 2 is true:

cd ~/Desktop/github/typespec/packages/http-client-python
echo "Regenerating test clients..."
npm run regenerate

After successful regeneration, update the marker:

touch .last-regenerate

If regeneration fails, report the error and stop.

If regeneration is NOT needed, inform the user:

"Generated code is up to date (source unchanged since last regeneration). Skipping regenerate."

Step 4: Run the generator test suite

cd ~/Desktop/github/typespec/packages/http-client-python
npm run ci

This runs:

1. test:emitter - vitest unit tests for the emitter
2. ci:generator - pytest tests against generated code

The full CI can take several minutes. Keep the user informed that it's running.

Step 5: Report results

• If tests pass: Confirm success and summarize any notable output (warnings, skipped tests, etc.)
• If tests fail: Show the failing test names and error messages clearly. Offer to help debug or investigate specific failures.

Integration with diff-upstream

Both skills share the regeneration marker system:

• diff-upstream always regenerates and updates .last-regenerate
• test-unbranded-emitter checks this marker and skips regeneration if fresh

Efficient workflow:

diff-upstream → test-unbranded-emitter

The second command will skip regeneration since diff-upstream just did it.

Standalone workflow:

test-unbranded-emitter

Will regenerate only if source changed since last run.

Notes

Regeneration takes time

npm run regenerate can take 2-3 minutes. The staleness check avoids this delay when the user is iterating on test fixes (not emitter changes).

Test types

This skill runs npm run ci (full CI). If the user wants just unit tests or just pytest, they can ask specifically.

Known flaky test

The test server teardown sometimes fails with ProcessLookupError - this is a pre-existing infrastructure issue, not a test failure. All actual tests may pass even if you see this error at the end.

Force regeneration

If the user suspects stale generated code despite the marker, they can force regeneration:

rm ~/Desktop/github/typespec/packages/http-client-python/.last-regenerate

Then run the skill again.

Gitignore

Ensure this entry is in the package's .gitignore:

.last-regenerate

Source: microsoft/typespec — distributed by TomeVault.