Add tests to compile generated code

[Weakky]

Tries to compile every tests to make sure there are no errors

• Typescript
• Flow

Fixes #69

[jasonkuhrt]

Started onboarding myself into this pr tonight. A small dep change to fix a yarn warning allowed the tests to pass locally. They pass on ci too, but then (it seems) jest never exits and so the ci times out. Aside from ramping up, a difficulty here is the lack of feedback. To help narrow the issue, I may try to incrementally temporally eliminate test code until the ci problem goes away.

[jasonkuhrt]

I didn't observe carefully before. Looking again at the tests running on c, I noticed that the two "large-schema" tests are missing from the output, e.g. from local:

 PASS  src/tests/flow/large-schema.test.ts (8.686s)
 PASS  src/tests/typescript/large-schema.test.ts (13.524s)

When I isolate the large-schema tests they run ok though (build link).

Looking at past recent failures it seems there is a random-like failure among "basic" and "large" schema tests across both TS and FT.

I would ssh into the build container but I don't have permission:

[jasonkuhrt]

jest --runInBand is one solution. It increases the test time however. Will continue looking for a root cause.

[jasonkuhrt]

Related:

• https://discuss.circleci.com/t/babel-7-jest-results-in-timeout/25383/5
• https://vgpena.github.io/jest-circleci/
• https://circleci.com/ideas/?idea=CCI-I-578

[jasonkuhrt]

Eventually figured out that the disk contention was not the root cause (probably not implicated at all). There are quite a few existing issues on the inerwebs about cci/travis not handling jest auto-cpu-count detection well. Capping worker count on CI worked here too.

I found one flow test that was commented out and needed finishing, default name. What I found was that flow wouldn't compile because (not type name mismatch):

export interface Mutation_AddMemberData {
  email: string;
  projects: string[];
}

export interface Mutation_Args_AddMember {
  data: AddMemberData;
}

I now append prefixes to the types of field arguments (in the above case... Mutation_) to comply with the pattern in our snapshots. I do not really understand the flow strategy, so, I'm not 100% confident with my commit here; in a nutshell: every field arg type gets a prefix except enum, input, object (which correspond to at least in some tests to GQL ID type).

[jasonkuhrt]

This module contains bug-fix code in response to the new compilation step we do.

[timsuchanek]

Thanks a lot both @Weakky and @jasonkuhrt for all this work! This will add soo much stability to our releases. Besides the two lets, the PR looks amazing 🙌

[jasonkuhrt]

@timsuchanek let me know if there are any other blocking issues in this PR from your point of view.

[jasonkuhrt]

@timsuchanek given that you seemed pretty satisfied already and I've fixed the two style issues you noted, going to merge, keep things rolling. And these tests like you said will start helping immediately : )