[00:00.000 --> 00:16.760]  Okay, please welcome our next speaker, Ivan, to talk about contracts for free.
[00:16.760 --> 00:17.760]  Hi everyone.
[00:17.760 --> 00:25.360]  It's a really quick talk about the Akitoi project and some unpopular opinion about
[00:25.360 --> 00:29.000]  using runtime types.
[00:29.000 --> 00:35.040]  So the thing is, there is an old issue, maybe you know it, in NICS, which was closed since
[00:35.040 --> 00:40.120]  because it will never be implemented, which is a bit embarrassing about the fact that
[00:40.120 --> 00:50.600]  NICS have no any static type system, and the thing is, I changed my mindset relatively
[00:50.600 --> 00:57.840]  recently about that, but I will talk about it later, because what's the issue about the
[00:57.840 --> 00:58.840]  type system?
[00:58.840 --> 01:04.360]  It's the lack of a static type system, it's the error you'll get, you'll get them at
[01:04.360 --> 01:05.360]  the last moment.
[01:05.360 --> 01:11.240]  It's really inconsistent regarding where the mistakes were made, and when you read your
[01:11.240 --> 01:17.760]  stack trace, you get out time to find what was the actual error, and that is really not
[01:17.760 --> 01:22.760]  helpless for, for example, beginners trying to work on NICS expression or people trying
[01:22.760 --> 01:27.080]  to debug NICS expression right then by other person.
[01:27.080 --> 01:36.600]  So I changed my mindset because NICS is a really simple language, and it should be a
[01:36.600 --> 01:42.560]  simple language, and it invites us to build construct in library space, and if you look
[01:42.560 --> 01:47.280]  at NICS built-ins, for example, it's really few built-ins, and a lot of things that could
[01:47.280 --> 01:52.920]  be built-ins are actually implemented as NICS by Quage, LibSync, and there is a NICS by
[01:52.920 --> 02:00.760]  Quage, LibTypes, maybe without NNES, which is what you use when you define NICS option
[02:00.760 --> 02:09.200]  to say the option could be that or that and everything, and so I changed my mindset because
[02:09.200 --> 02:16.200]  NICS is made in two steps, you know, so it's the first step that you evaluate the language
[02:16.200 --> 02:23.160]  and produce things in NICS to review, symbols in NICS store, and after you build, you made
[02:23.160 --> 02:29.400]  your Quage, but the instantiate part is guaranteed to terminate, you know, you will, the computation
[02:29.400 --> 02:35.400]  will always terminate, which is what you expect for a type system actually, and which is not
[02:35.400 --> 02:42.240]  the case of some type system or most of macro systems that are not expected to terminate,
[02:42.240 --> 02:51.720]  so actually it's a good guarantee, and that's why we could do a lot of things in NICS, you
[02:51.720 --> 02:58.120]  can see it as, I changed my mind because I see that NICS is not really about the runtime
[02:58.120 --> 03:04.720]  and programming language, so we could do a check at runtime, and I wrote a thing while
[03:04.720 --> 03:11.320]  a vacation and realized later that other folks already did it, which is Tajjin, maybe you
[03:11.320 --> 03:18.280]  know him, he works on Twix, Twix right now, and he also writes NICS OnePager, one of the
[03:18.280 --> 03:25.720]  best NICS starting guide at some point, and the thing is, it's really simple, you define
[03:25.720 --> 03:32.880]  the data you want, just validate your function, so you say, I want something like that, and
[03:32.880 --> 03:38.160]  you check at this point of your NICS expression what it evaluates, that's its respect to a
[03:38.160 --> 03:46.080]  contract, a validator function, and that's help actually there to fail early where you
[03:46.080 --> 03:54.200]  want it to fail, there is differences in implementation, I will not talk really about, but I really
[03:54.200 --> 03:59.480]  simply didn't rely on NICS package, and it's fully compatible with NICS exception, so you
[03:59.480 --> 04:11.720]  can define a thing, and the last thing is, those composes, I go back inside, you can
[04:11.720 --> 04:17.480]  define types, you can define validator, instruct, that became validator, and you can compose
[04:17.480 --> 04:23.720]  really complex things, because it's really expressive, and also the nice things with
[04:23.720 --> 04:30.720]  runtime types is you can produce recoverable error, which is the case for types errors that
[04:30.720 --> 04:36.920]  exist in NICS itself, so for example, this will be unrecoverable, but if you use the
[04:36.920 --> 04:44.960]  runtime type, you would recover this situation. Last thing is, I try to advertise this kind
[04:44.960 --> 04:53.880]  of technique, because it helps a lot to debugs your application and to have assertion about
[04:53.880 --> 05:02.680]  this, respect that and everything, it doesn't really solve the problem, because it's not
[05:02.680 --> 05:09.920]  a real type system actually, but if you're not starting something from scratch, if you
[05:09.920 --> 05:14.600]  actually working on an existing NICS expression, NICS expression could be actually really large,
[05:14.600 --> 05:19.680]  I think personally it helps, if you're starting a new thing from scratch, there's really cool
[05:19.680 --> 05:24.960]  things like Q, dial, NICL, generate JSON, and you use that live in NICS package, so you
[05:24.960 --> 05:31.360]  can use it really easily, and there is also, I like personally plural NICS, because it
[05:31.360 --> 05:42.320]  generates NICS output results and JSON, okay, I think I said most of what I would say, so
[05:42.320 --> 05:47.320]  don't hesitate if you have a question or want to share it with you. You don't have time?
[05:47.320 --> 06:15.340]  Thank you, it was really good presentation.