P1b/refactor

[gavinlin-cloud]

P1B: Refactor (packages/httpapi-codegen/src/index.ts:810): Function with high complexity (count = 31): assertPortable

Use this pull request template to briefly answer the questions below in one to two sentences each.
Feel free to delete this text at the top after filling out the template.

1. Issue

Link to the associated GitHub issue:
#51
Full path to the refactored file:
packages/httpapi-codegen/src/index.ts
What do you think this file does?
(Your answer does not have to be 100% correct; give a reasonable, evidence‑based guess.)
It builds standalone TypeScript API clients from an Effect HttpApi definition. The generated client is its own file and can't use server runtime objects, so the compiler checks first that every schema is "portable", meaning it can be written using only effect types.
What is the scope of your refactoring within that file?
(Name specific functions/blocks/regions touched.)
assertPortable() at line 810, plus the visitCurrent helper nested inside it.

Which Qlty‑reported issue did you address?
(Name the rule/metric and include the BEFORE value; e.g., “Cognitive Complexity 18 in render()”.)
Function with high complexity (count = 31) in assertPortable. Qlty also flagged many returns (count = 14) on it, and the nested visitCurrent was separately flagged at complexity 20 with 10 returns.

2. Refactoring

How did the specific issue you chose impact the codebase’s maintainability?
The function was 47 lines doing three different jobs: tracking visited nodes with a cache and cycle guard, recursing into child nodes based on node kind, and applying a stricter rule for tagged error classes. Changing one job meant reading all three. The helpers were also nested functions, which pushed every branch deeper and inflated the score.
What changes did you make to resolve the issue?
I pulled out three helper functions to module scope: astChecksPortable, portableChildren, and taggedErrorUnportable.
How do your changes improve maintainability? Did you consider alternatives?
Each job is now its own function, and the list of node kinds lives in one place instead of being spread through an if-chain. assertPortable went from complexity 31 to no complexity smell, returns dropped from 14 to 9, and visitCurrent's smells are gone. I looked at also moving visit and visitCurrent out to module scope, which would cut the number further. I decided against it because it means passing the visiting set and portable cache around as arguments, and that felt risky on recursive code with a cycle guard when the smell was already fixed.

3. Validation

How did you validate that the change is correct?
The tests in packages/httpapi-codegen/test/generate.test.ts already cover this code. They check exact error text and exact pass/fail results, so any change in behavior would break them. All 66 tests pass, same as before the refactor.
Attach a screenshot of the test coverage showing the lines were executed by the tests.

Attach a screenshot showing the tests that cover the change passing during CI

Attach a screenshot of qlty smells --no-snippets <full/path/to/file.ts> showing fewer reported issues after the changes.

local test passing