The mainland

draft

slide 1.1

Why so many languages?

draft

slide 1.2

Nim's tradeoffs

draft

slide 1.3

The map

• Island 1: Accessible Performance
• Island 2: Success Stories
• Island 3: Data Stack
• Island 4: Interoperability
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

Island 1

• Island 1: Accessible Performance
• Island 2: Success Stories
• Island 3: Data Stack
• Island 4: Interoperability
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

Reads like Python

draft

slide 2.2

Runs like C

draft

slide 2.3

No ceiling

draft

slide 2.4

Island 2

• Island 1: Accessible Performance ✓
• Island 2: Success Stories
• Island 3: Data Stack
• Island 4: Interoperability
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

Success stories

How do we define success stories in scientific computing? By looking at people who used Nim to advance science.

The following are two stories:
🧬 Bioinformatics and ☀️ Physics

🧬 Brent Pedersen · @brentp

Genomics researcher and software developer: 6 papers on 6 Nim tools (2018-2022), with 1,944 citations and 1,833 stars.
Others followed (awesome-bionimformatics)

# read a file of DNA reads aligned to the genome
import hts
var b: Bam
open(b, "sample.bam", index=true)
# reads in a region of chromosome 6
for r in b.query("6", 30816675, 32816675):
  # keep reads placed with confidence
  if r.mapping_quality > 10u:
    # where each read sits on the genome
    echo r.chrom, r.start, r.stop

☀️ Sebastian Schmidt · @Vindaar

A PhD student in Physics searching for solar axions at CERN's CAST: the entire analysis code base is written in Nim.
He ended up building an important part of the data stack in Nim

# how many axions the Sun sends us at energy E_a,
# per keV, cm² and second; units checked when compiling
proc axionFluxPrimakoff(E_a: keV,
                        g_aγ: GeV⁻¹): keV⁻¹•cm⁻²•s⁻¹ =
  let g₁₀ = g_aγ / 1e-10.GeV⁻¹
  result = g₁₀^2 * 3.821e10.cm⁻²•s⁻¹•keV⁻¹ *
    (E_a / 1.keV)^3 / (exp(E_a / (1.103.keV)) - 1)

Rocks ahead

A tiny ecosystem: pioneers build what they need, or move where the libraries are

Island 3

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack
• Island 4: Interoperability
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

The stack

Meet the penguins

Dataframes: Datamancer

let df = readCsv("data/penguins.csv")
  .filter(f{`body_mass_g` > 0.0}) # drop missing

echo df.group_by("species").summarize(
  f{float: "mass (g)" << mean(`body_mass_g`)},
  f{float: "flipper (mm)" << mean(`flipper_length_mm`)})

DataFrame with 3 columns and 3 rows:
     Idx         species        mass (g)    flipper (mm)
  dtype:          string           float           float
       0          Adelie            3701             190
       1       Chinstrap            3733           195.8
       2          Gentoo            5076           217.2

Plots: ggplotnim

ggplot(df, aes("flipper_length_mm", "body_mass_g",
               color = "species")) +
  geom_point() +
  ggsave("images/penguins.png", width = 500, height = 320)

Tensors: Arraymancer

let cols = ["bill_length_mm", "bill_depth_mm",
            "flipper_length_mm", "body_mass_g"]
let X = stack(cols.mapIt(df[it, float]), axis = 1) # 342 x 4
let Z = (X -. X.mean(axis = 0)) /. X.std(axis = 0)
let C = Z.transpose * Z / float(X.shape[0] - 1) # correlations
echo C.map_inline(x.round(2))

Tensor[system.float] of shape "[4, 4]" on backend "Cpu"
|1        -0.24     0.66      0.6|
|-0.24        1    -0.58    -0.47|
|0.66     -0.58        1     0.87|
|0.6      -0.47     0.87        1|

SciNim

• flambeau: bindings to libtorch
• nim-plotly: interactive plots with plotly.js
• numericalnim: integration, ODEs, optimization, interpolation

Rocks ahead

• no scikit-learn: ML algorithms are scattered
• multiple tensor libraries
• the ecosystem depends on a few contributors (Vindaar, mratsim, Clonkk, HugoGranstrom): when life happens, development stops

Island 4

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack ✓
• Island 4: Interoperability
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

Wrapping C and C++

draft

slide 5.2

Bridges

draft

slide 5.3

Nim from Python

draft

slide 5.4

Rocks ahead

draft

slide 5.5

Island 5

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack ✓
• Island 4: Interoperability ✓
• Island 5: An Expressive Language
• Island 6: Interactivity
• Island 7: Notebook

Custom Operators: defined

type
  Line = object   # through p, along d
    p: Point
    d: Vec
  Plane = object  # through p, normal n
    p: Point
    n: Vec

func `+`(a, b: Point): Line = Line(p: a, d: b - a)
func `+`(l: Line, c: Point): Plane =
  Plane(p: l.p, n: l.d × (c - l.p))
func `*`(l: Line, s: Plane): Point =
  l.p + ((s.p - l.p) ∙ s.n) / (l.d ∙ s.n) * l.d

Custom Operators: applied

func pt(x, y, z: float): Point = Point(x: x, y: y, z: z)

let
  line = pt(1, 2, 3) + pt(2, 3, 1)
  plane = pt(0, 0, 0) + pt(1, 0, 0) + pt(0, 1, 0)
echo line * plane

(2.50, 3.50, 0.00)

Macros to define Units

declareUnits:
  BaseUnits:
    Gram:
      short: g
      prefix: siKilo
      quantity: Mass
    Meter:
      short: m
      quantity: Length
    ...

declareUnits:
  Derived:
    SolarMass:
      short: Msun
      quantity: Mass
      conversion: 1.98847e30.kg
    EarthMass:
      short: Mearth
      quantity: Mass
      conversion: 5.9722e24.kg

Units applied

func period[L: Length, M: Mass](a: L, mass: M): Second =
  (2 * PI * sqrt(a^3 / (G_Newton * mass))).to(Second)

echo "Mars: ", period(1.524.AU, 1.Msun).to(yr)
echo "Moon: ", period(384_400.km, 1.Mearth).to(day)
echo compiles(1.AU + 1.Msun) # Length + Mass

Mars: 1.88139 yr
Moon: 27.4518 day
false

Rocks ahead

draft

slide 6.6

Island 6

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack ✓
• Island 4: Interoperability ✓
• Island 5: An Expressive Language ✓
• Island 6: Interactivity
• Island 7: Notebook

The original Shiny app

draft

slide 7.2

The Nim port

draft

slide 7.3

Rocks ahead

draft

slide 7.4

Island 7

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack ✓
• Island 4: Interoperability ✓
• Island 5: An Expressive Language ✓
• Island 6: Interactivity ✓
• Island 7: Notebook

Literate computing

draft

slide 8.2

nimib

draft

slide 8.3

Rocks ahead

draft

slide 8.4

Back on land

• Island 1: Accessible Performance ✓
• Island 2: Success Stories ✓
• Island 3: Data Stack ✓
• Island 4: Interoperability ✓
• Island 5: An Expressive Language ✓
• Island 6: Interactivity ✓
• Island 7: Notebook ✓

For the beginner

draft

slide 9.2

For the builder

draft

slide 9.3

Your language, seen from Nim

draft

slide 9.4

Why still niche?

draft

slide 9.5