How experimental the API really is?

[paluh]

Hi,

First things first - this library seems awesome after a closer look. I really love the API around records with nested routes and ability to generate client. Thanks a lot for building it!

I want to try your library in our project but I'm a bit scared by the claim from the README that it is still experimental so I want to ask if this is really the case.

• Should I expect really huge changes regarding the API? Do you have an approximation of what is possibly coming?

• Have you used this lib in production? Do you know anyone who is using it too?

• It seems (just by reading module names) that you are building efficient Trie structure for routing. Am I right? Have you done any benchmarking etc.?

P.S. I had my own attempt somewhat failed attempt of building this kind of router in the past: https://github.com/paluh/purescript-boomboom so I'm even more impressed how far you have pushed this API!

[hoodunit]

Hi! Thank you for your kind words. So here's where my head is at regarding this project at the moment. The main bigger design-affecting feature I'm hoping to get in there soon is for auto-generating API documentation in an extendable way. This is mostly fleshed out and "might" have some effect on the API, though I don't expect dramatic changes. So far all that has changed I believe is to use curly braces instead of triangle braces to match the more common convention (e.g. "/{id}" instead of "/"). After that I'm hoping to take some time to do proper benchmarking, round out various nitty gritty issues, and ideally declare a 1.0 at some point. I have other ideas on the horizon, but that would already fulfill the three pillars of servers, clients, and documentation (plus reasonable performance) that I have been targeting.

The project is not to my knowledge used in production, outside of one little project of mine for a small client (and various toy projects).

As mentioned I would like to do benchmarking slightly down the line to see what the performance is, but at least design-wise I don't believe there are significant blockers to making it performant (and some decisions have been made partly with this in mind). But the benchmarks will tell, of course.

[paluh]

Thanks for the quick response.

@hoodunit Don't you think that the above indicates that statement from the README could be transformed into a "status" section where we can put list of these upcoming features and plans? Maybe it is just time to encourage early adopters like me to do some battlefield tests ;-)

P.S. By the way @vladciobanu has done a short review/exploration of servant like solutions (it starts around 00:50) for PS recently and was somewhat afraid that this lib is somewhat stale... ;-)

[hoodunit]

You may be right. I'll have to think about this a bit. But I'm picking this project up again here so there should be some activity. Thanks for the pointer to the video review, the live video walk-through is a really nice way to get some feedback.