sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
<sharkdp>
The type inference & list datatype changes have been merged! The latest version is also online at https://numbat.dev/.
<triallax>
fantastic!
<sharkdp>
As far as I can tell, there are only very minor regressions (worse error messages in some edge cases and some details in pretty printing).
<sharkdp>
And those should be solvable as well.
<sharkdp>
I will create a few follow-up tickets later today.
<triallax>
i'll try it out in a bit
<sharkdp>
I wrote a lot of new tests. All of the previously-existing ones are passing. But I'm sure we will still find a few bugs here and there, so please feel free to play around with this. And if you have some existing Numbat code, please check if it still works.
<sharkdp>
There's also a breaking change concerning the definition of generic functions, most of which will now need `Dim` bounds. See https://numbat.dev/doc/example-numbat_syntax.html => my_sqrt for an example.
<sharkdp>
I think there are also some cool follow-up tasks, like giving `quadratic_equation` a `List<…>` return type
<sharkdp>
.. or turning the numerical_diff.nbt example into actual functionality within the prelude
Charbot9000 has quit [Remote host closed the connection]
<triallax>
and it feels strange to be seeing a haskell-style function signature in numbat
<achin>
it's clearly says it's a function, but it also re-uses the traditional [brace] syntax for listing the "dimension", which seems a little confusing
<triallax>
you have a point
<triallax>
also, would be nice if it preserved the names of the generics
<achin>
seeing [DateTime] isn't confusing to me, because there's just one thing there. but there's a whole lot of syntax and symbols in the function example that somehow set it apart for me
<achin>
but i do understand the desire to use consistant syntax. so i'm not sure if i'm proposing any type of change here. just making a personal observation
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
<sharkdp>
"would be nice if it preserved the names of the generics" => this is one of the open points, yes
<sharkdp>
And it will correctly infer the `A: Dim` bound, which is needed by `sum`. And the List<A> to A type for the function
<achin>
ah, very nice
<achin>
i think i might try to experiment with compiling numbat to a wasm to do hot-reloading within the bot
<achin>
at the moment i have to restart the bot to get the latest version, which is not that suitable for the breakneck pace at which sharkdp is implementing features :)
<sharkdp>
haha
<triallax>
achin: word
sharkdp has quit [Remote host closed the connection]
sharkdp has joined #numbat
<sharkdp>
I'm currently thinking that we are very close to implementing some really powerful things in Numbat
<sharkdp>
(... there are two things that are really lacking: closures and local bindings ...)
<sharkdp>
but even without them, we could start implementing things we talked about recently
<triallax>
what's the finish line in mind here?
<sharkdp>
Things I would really like to see:
<sharkdp>
- parsing command line args in a nice way. maybe something like `let lengths: List<Length> = argv()` where `argv` would be polymorphic. Here, it would be instantiated as `argv::<Length>()` (a syntax that is not yet available). And we could then enter `1m 3inch 100cm` as arguments.
<sharkdp>
- Plotting!
<sharkdp>
- Simple I/O using CSV/JSON
<sharkdp>
Concerning plotting, imagine:
<sharkdp>
(I'm going to challenge the bot a bit)
<sharkdp>
!nb fn _linspace_helper(start, end, n_steps, i) = if i == n_steps then [] else cons(start + (end - start) * i / (n_steps - 1), _linspace_helper(start, end, n_steps, i + 1))
<Charbot9000>
fn _linspace_helper< A: Dim>(start: A, end: A, n_steps: Scalar, i: Scalar) -> List< A> = if (i == n_steps) then [] else cons(start + ((end - start) × i / (n_steps - 1)), _linspace_helper(start, end, n_steps, i + 1))
<sharkdp>
!nb fn linspace<D: Dim>(start: D, end: D, n_steps: Scalar) -> List<D> = if n_steps <= 1 then error("Number of steps must be larger than 1") else _linspace_helper(start, end, n_steps, 0)
<Charbot9000>
fn linspace< A: Dim>(start: A, end: A, n_steps: Scalar) -> List< A> = if (n_steps ≤ 1) then error("Number of steps must be larger than 1") else _linspace_helper(start, end, n_steps, 0)
<Charbot9000>
map(amplitude, linspace(0 second, 10 second, 100)) = [3 cm, -1.21842 cm, -1.63229 cm, 2.39077 cm, -0.515351 cm, -1.67091 cm, 1.80766 cm, -0.00795252 cm, -1.57343 cm, 1.28501 cm, 0.331382 cm, -1.39227 cm, 0.841281 cm, 0.53348 cm, -1.16861 cm
<triallax>
heh
<triallax>
very cool
<sharkdp>
(oh, looks like the output is capped.. anyways)
<triallax>
the output isn't capped, it's just that irc has a char limit for each message
<triallax>
classic irc limitation
<sharkdp>
I see
<triallax>
maybe i should bridge this channel to matrix...
<achin>
it's effectively a per-missage cat on line length, enforced by the irc servers
<achin>
it doesn't help that these messages are highly colorful (since the color codes add to the message length)
<sharkdp>
I think it's fine. We don't need to see the whole list :)
<sharkdp>
We could write that list of values to a file. And, crucially, also write out meta information that we have: function name, parameter names, parameter types, units..
<sharkdp>
.. and use that to generate a plot including automatically generated axis labels
<sharkdp>
But closures/lambdas are really a must-have in order for it to feel like a true functional programming language. `linspace` above would be much easier to implement with something like `map(|i| start + i * …, range(0, n))`
<sharkdp>
typechecker-wise, we're all set for introducing them. everything is there already. but the VM implementation of closures is non-trivial (https://craftinginterpreters.com/closures.html) due to captures
<sharkdp>
!nb let num_photons = P_received / E_photon -> 1/s
<Charbot9000>
let num_photons: Time⁻¹ = P_received / E_photon ➞ 1 / second
<sharkdp>
!nb "Number of photons per second: {num_photons}"
<Charbot9000>
"Number of photons per second: {num_photons}" = "Number of photons per second: 240188 s⁻¹" [ String]
<sharkdp>
Same result as the S.E. answer
<sharkdp>
Would be cool to clean this up and add as a Numbat example
<sharkdp>
And I should really fix the "readable types" feature, which I broke in the type inference PR. Otherwise, we wouldn't see things like `Mass / Time^3` above.. but `Irradiance`
<sharkdp>
And for the cleaned-up example, we use `now()` to adjust the distance-to-earth :)
sharkdp has quit [Remote host closed the connection]