Now that we finally have a documentation you can really see what procs are documented and which are not.

It could also be nice with a paragraph at the top of each module as well to describe what it provides and recommended methods.

• Differentiate
• integrate
• interpolate
• ode
• optimize
• rbf
• utils

Many thanks for this very interesting package.
I've tried the examples in your documentation.
The following CubicSpline example does not compile on my Nim (devel)

import numericalnim
var X = @[0.0, 10.0, 20.0, 30.0, 40.0, 50.0, 60.0] # seconds
var Y = @[0.0, 4.0, 6.0, 6.5, 6.4, 6.2, 0.0] # meters/second
let spline = newCubicSpline(X, Y)
echo spline.eval(5.0)
#[
NumNim_BUG.nim(6, 6) template/generic instantiation from here
$HOME/.nimble/pkgs/numericalnim-0.6.0.1/numericalnim/interpolate.nim(182, 27) Error: cannot instantiate: 'T'
]#

Nim Compiler Version 1.5.1 [Linux: amd64]
Compiled at 2021-01-07
Copyright (c) 2006-2020 by Andreas Rumpf

git hash: 89a21e4ec71e705833d2aacd069e291cf41a19c6
active boot switches: -d:release

I need to separate the internal util procs I'm using and the ones like Vector which should be exported to the user. I should possibly rename the exported module something more descriptive than utils as well :P

If there are duplicates in x-value in a dataset there are two scenarios:

1. If the y-values are the same, remove all but one of the duplicates
2. If the y-values aren't the same, raise exception.

While we are at it we should also revamp sortDataset to accept a variable number of input arrays. Should we use array[N: static int, seq[T]] as both input and output then? Nah I'll just do varargs[seq[T]] and sort along the first sequence only and keep track of the indices for the rest. That should be efficient enough. T prepare for VectorLike I should probably not use varargs though, will go with openarray.

var ptsCoord = newSeq[array[2, float]]()
# corners
ptsCoord.add([-200.0,-150.0])
ptsCoord.add([-200.0,150.0])
ptsCoord.add([200.0,-150.0])
ptsCoord.add([200.0,150.0])
var ptsValue = @[0.0,1.0,4.0,5.0]
# extra pts
ptsCoord.add([0.0,-150.0])
ptsCoord.add([0.0,150.0])
ptsCoord.add([0.0,0.0])
ptsValue.add(8.0)
ptsValue.add(1.0)
ptsValue.add(1.0)
let bary = newBarycentric2D(ptsCoord.toTensor(), ptsValue.toTensor())

Compiled with nim cpp --gc:arc with Nim 1.6.0

It seems related to Vector2 in cdt

Stage 1: Improve the error messages (specifically trimAndSortDataset)

Stage 2: improve error handling. Some universal @SciNim way of reporting back non-fatal error to the user. For example that the maximum number of iterations was reached.

• Remove recent major changes, they aren't very recent anymore.
• The Gitter channels aren't really relevant either. Link to Science channel on Matrix and Discord instead.
• Add link to scinim/getting-started

if this nim-lang/Nim#20412 get merged, it will be a problem.