I am building a GraphQL service to act as a endpoint-to-rule-them-all and GraphiQL has helped me sell the concept very well. It is much easier to talk about how queries work when a developer can just jump right in and start playing with it without having to setup any tooling or whatnot.
Interactively querying REST APIS is a massive pain, doing anything useful requires multiple dependent queries, postman doesn't really help with that.
Being able to query a graph interactively is actually useful, I've used graphiql to explore the data graph to figure out the query I need for the data requirements of a component before I start writing the actual component itself.
That really depends on your API! We have an orchestration layer that exists specifically so our front end does not have to run multiple queries all the time. Most of what our front end does is by running a very small handful of queries.
I think orchestration layers are something most growing/larger orgs strive for, in my experience.
If it's actually a REST API then it's going to take multiple queries. Certainly you can put an non REST API in front of it to reduce the queries. Every site does it and every site does it in a different way. GraphQL is built to solve this exact problem in a standard way.
We use postman extensively at work. We even have non-devs using it to hack with API's to get things done without having to wait for a UI to be built. It's a great tool. I especially like how you can serialize all of your saved endpoints and send them to other devs.
Am I the only one who designs endpoints to spit back usage instructions if you call them without arguments? This goes a long way toward helping devs use those endpoints, in my experience - and they can use normal tools (in particular, the network pane of Chrome Dev tools). This is not to take away from GraphiQL, but it doesn't seem like that much more than you get with such a convention.
I've not seen this practice very much in my experience; most API's I've worked with go the "automated documentation generation" route and host the docs somewhere separate instead. This does mean you don't need tooling just to find out how to use the API.
This is distinct from throwing exceptions when an endpoint is called with e.g. invalid parameters.
No you're not the only one. But I do fear we may be some sort of "silent minority". I always felt that having human and machine readable instructions was part of trying to achieve "proper" REST, the full HEOTAS or whatever the acronym is. I admire the principle and the outcomes it strives for, but very much dislike the "wankery" that goes on over the topic sometimes.
I am not sure why editors think a mark up is a good experience. Recently I have seen several editors that have markup on one side, and a formatted/stylized output on the editor. The output only sometimes feeds back to the markup, and usually does not have full functionality. This ends up being slow most of the time, and clicking/dragging/expanding text is not a good experience. They only thing they provide is validation feedback. But you still have worry about closure, indentation, formatting, etc. I usually default to using a light text editor and a shell watching validation.
Is everyone so sick of Windows forms that we can only edit text now? No GUI?
