Hi all 🙂 I had an interesting discussion at my company yesterday and wanted to know your point of view. A group of developer argued that Pact contact testing doesn’t provide much value when combined with Typescritp/Graphql because with Graphql we can use the schema to generate the type and use those types on our clients to ensure they are in sync with the API.

anyone have any experience utilizing the graphql-codegen library to help generate the contracts Pact can use in a mock server? 🙂 been looking at lots of examples and what not, but haven't found anything around in particular with codegen yet.

Is state handlers something we can't utilize from the GraphQLInteraction? adding

.given()

does work, but it also creates two interactions for the same test apparently adding the given() call to my interaction. I don't know if having both uponReceiving and given in one interaction is creating those two tests, but based on the type defs of uponReceiving that wouldn't make sense to me.

Wondering what are strategies people using to test Apollo’s

useQuery

hooks with Pact? The only working one I found so far is to write component tests and that does mean that we are tying pact to our UI implementation. Trying to make pact covered layer as close as possible to the direct http calls.

Something my mind is still a little fuzzy on is the benefit (specifically in GraphQL world) of verifying the new GraphQLInteraction added with a call to a mocked server and expecting the response to match what we would expect it to be. I think the hurdle my mind is currently trying to jump over is when would that expectation reasonably fail ... it seems like it would be a failure because I wrote the expectation wrong itself but failure wouldn't happen otherwise(?). I think to highlight the general usecase a bit more, here is the last bit of a GraphQLInteraction and what I expect it to respond with:

new GraphQLInteraction()
.......
.willRespondWith({
    status: 200,
    headers: {
        'Content-Type': 'application/json',
    },
    body: {
        data: {
            addTime: {
                updatedRecord: {
                    id: like('62edadccafea8bf4d0d17bbd'),
                    currentTime: like(40),
                },
            },
        },
    },
});

Calling then a mocked Pact server, I am essentially testing that the response would also equal that response defined in the GraphQLInteraction ... so when would my expectation within the test file itself actually fail other than me intentionally (or unintentionally) not writing correctly the object to match against the response. Seems like even the unintentional failure would be more prohibitive than beneficial. Totally could be misunderstanding the whole point (it's actually highly likely) ... but just wanted further documentation to unfuzzy this concept, and/or community discussion to help bridge this understanding in my head. Seems like, as I'm typing this, this would be more broad than just GraphQL itself and maybe it's a concept higher above I'm just not yet grasping, so would 1,000% love some insight and guidance to this scenario.

Is there documentation available on how to generate a graphql provider contract for bi-directional testing?

We don’t support GraphQL for BDCT, unless you can document a GraphQL API as an OAS (which I don’t think you can do?)

We were looking at a conversion tool such as this, https://www.npmjs.com/package/graphql-to-openapi, but it would mean another layer for potential failure.

hey, not sure if this has been answered already, but is there update for PactV3 support for graphql?