prographs 0.12.0 → 0.13.0

package/dist/{chunk-BEEOSNK7.cjs → chunk-PEJJ6TXJ.cjs}

@@ -1,8 +1,5 @@
 "use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _nullishCoalesce(lhs, rhsFn) { if (lhs != null) { return lhs; } else { return rhsFn(); } } function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; } var _class; var _class2; var _class3; var _class4; var _class5; var _class6;
 
-var _chunkMCPPFRLNcjs = require('./chunk-MCPPFRLN.cjs');
-
-
 var _chunkZQINOZBJcjs = require('./chunk-ZQINOZBJ.cjs');
 
 
@@ -12,16 +9,19 @@ var _chunk4VWNDCUScjs = require('./chunk-4VWNDCUS.cjs');
 var _chunkNEVHPFWRcjs = require('./chunk-NEVHPFWR.cjs');
 
 
-var _chunkUXFWDEP6cjs = require('./chunk-UXFWDEP6.cjs');
+var _chunkL6BGNHFLcjs = require('./chunk-L6BGNHFL.cjs');
+
+
+var _chunkDCGAAN3Zcjs = require('./chunk-DCGAAN3Z.cjs');
 
 
 var _chunkDKE6DIRKcjs = require('./chunk-DKE6DIRK.cjs');
 
 
-var _chunkEEF5WRRZcjs = require('./chunk-EEF5WRRZ.cjs');
+var _chunkUXFWDEP6cjs = require('./chunk-UXFWDEP6.cjs');
 
 
-var _chunkUMEKIIHJcjs = require('./chunk-UMEKIIHJ.cjs');
+var _chunkEEF5WRRZcjs = require('./chunk-EEF5WRRZ.cjs');
 
 // src/charts/modelcharts/AlePlot.ts
 var _d3 = require('d3');
@@ -109,135 +109,6 @@ var AleEffects = class _AleEffects {
   }
 };
 
-// src/charts/modelcharts/ModelSamples.ts
-
-var SHAP_FEATURES = [
-  ["tenure_months", -1, 1, "saturating", 30, 16],
-  ["monthly_charges", 1, 0.78, "linear", 68, 22],
-  ["support_tickets", 1, 0.6, "count", 2, 1.8],
-  ["contract_months", -1, 0.5, "saturating", 14, 9],
-  ["payment_delay_days", 1, 0.4, "count", 4, 5],
-  ["data_usage_gb", -1, 0.3, "linear", 42, 18],
-  ["autopay_enrolled", -1, 0.24, "binary", 0, 1],
-  ["device_count", -1, 0.18, "count", 3, 1.4],
-  ["customer_age", -1, 0.13, "linear", 44, 13],
-  ["discount_rate", -1, 0.1, "linear", 8, 5],
-  ["region_index", 1, 0.06, "linear", 5, 3],
-  ["referral_count", -1, 0.05, "count", 1, 1.2]
-];
-var SHAP_ROWS = 120;
-var ALE_ROWS = 480;
-var ALE_BINS = 12;
-var ALE_SIDE = 6;
-var GRID = 61;
-var STEPS = 60;
-var LAYER_UNITS = [[12, 24, 16, 8, 3], [16, 32, 16, 8, 4], [8, 20, 20, 10, 2]];
-var round = (value, digits = 4) => +value.toFixed(digits);
-var clamp = (value, low, high) => Math.max(low, Math.min(high, value));
-var ModelSamples = class {
-  /** `{ score, actual }` for one classifier at 30% prevalence. */
-  static scores(seed) {
-    return _chunkMCPPFRLNcjs.ClassifierCurveBase.scored(seed).filter((row) => row.model === "Logistic").map(({ score, actual }) => ({ score, actual }));
-  }
-  /**
-   * ALE of a debt-to-income feature at its quantile bin edges, `{ x, x2, effect, low, high, count }`: rows with no `x2` are the
-   * one-feature curve, rows with one are the nodes of its interaction surface with credit utilisation.
-   */
-  static ale(seed) {
-    const { n } = _chunkUMEKIIHJcjs.SampleData.rng(seed);
-    const pairs = Array.from({ length: ALE_ROWS }, () => {
-      const debt = clamp(34 + 11 * n(), 4, 68);
-      return [debt, clamp(48 + 1.2 * (debt - 34) + 15 * n(), 0, 100)];
-    });
-    const edges = (values, bins) => {
-      const sorted = [...values].sort(_d3.ascending);
-      return [...new Set(Array.from({ length: bins + 1 }, (_, k) => Math.round(_d3.quantileSorted.call(void 0, sorted, k / bins))))];
-    };
-    const within = (value, grid, k) => (k === 1 ? value >= grid[0] : value > grid[k - 1]) && value <= grid[k];
-    const debts = edges(pairs.map((pair) => pair[0]), ALE_BINS), uses = edges(pairs.map((pair) => pair[1]), ALE_SIDE);
-    const counts = debts.map((_, k) => k ? pairs.filter(([debt]) => within(debt, debts, k)).length : 0);
-    const raw = debts.map((debt) => 0.17 * Math.tanh((debt - 36) / 9) + 6e-3 * n());
-    const total = counts.reduce((sum, count2) => sum + count2, 0) || 1;
-    const centre = counts.reduce((sum, count2, k) => sum + (k ? count2 * (raw[k - 1] + raw[k]) / 2 : 0), 0) / total;
-    const curve = debts.map((x, k) => {
-      const effect = raw[k] - centre, reach = 0.012 + 16e-4 * Math.abs(x - 36);
-      return { x, x2: null, effect: round(effect), low: round(effect - reach), high: round(effect + reach), count: counts[k] };
-    });
-    const coarse = debts.filter((_, k) => k % 2 === 0);
-    const surface = coarse.flatMap((x, i) => uses.map((x2, j) => ({
-      x,
-      x2,
-      effect: round(0.05 * ((x - 34) / 11) * ((x2 - 48) / 20) + 4e-3 * n()),
-      low: null,
-      high: null,
-      count: i && j ? pairs.filter(([debt, use]) => within(debt, coarse, i) && within(use, uses, j)).length : 0
-    })));
-    return [...curve, ...surface];
-  }
-  /** `{ feature, shap, value }`, one row per customer and feature of a churn model, in log-odds. */
-  static shap(seed) {
-    const { n } = _chunkUMEKIIHJcjs.SampleData.rng(seed);
-    return Array.from({ length: SHAP_ROWS }, () => SHAP_FEATURES.map(([feature, direction, strength, shape, centre, spread]) => {
-      const z = n();
-      const value = shape === "binary" ? +(z > 0.1) : shape === "count" ? Math.max(0, Math.round(centre + spread * z)) : round(centre + spread * z, 1);
-      const level = shape === "binary" ? value ? 0.8 : -0.7 : shape === "count" ? (value - centre) / spread : z;
-      const push = shape === "saturating" ? Math.tanh(1.3 * level) * 0.75 : level * 0.5;
-      return { feature, shap: round(direction * strength * push + 0.09 * strength * n()), value };
-    })).flat();
-  }
-  /** A curved valley of loss over two parameter directions, with the paths of three optimisers from one start. */
-  static lossSurface(seed) {
-    const { n } = _chunkUMEKIIHJcjs.SampleData.rng(seed);
-    const steep = 4;
-    const loss = (x2, y2) => 0.05 + (1 - x2) ** 2 + steep * (y2 - x2 * x2) ** 2;
-    const slope = (x2, y2) => [-2 * (1 - x2) - 4 * steep * x2 * (y2 - x2 * x2), 2 * steep * (y2 - x2 * x2)];
-    const x = Array.from({ length: GRID }, (_, i) => round(-2 + 4 * i / (GRID - 1), 3)), y = x.map((value) => round(value + 1, 3));
-    const start = [-1.3 + 0.08 * n(), 2.55 + 0.08 * n()];
-    const walk = (name, step) => {
-      const points = [start];
-      for (let t = 1; t <= STEPS; t++) {
-        const [px, py] = step(points[t - 1], t);
-        points.push([clamp(px, -2, 2), clamp(py, -1, 3)]);
-      }
-      return { name, points: points.map(([px, py]) => [round(px, 3), round(py, 3)]) };
-    };
-    const speed = [0, 0], mean3 = [0, 0], square = [0, 0];
-    return {
-      x,
-      y,
-      z: y.map((row) => x.map((column) => round(loss(column, row)))),
-      paths: [
-        walk("SGD", ([px, py]) => {
-          const [gx, gy] = slope(px, py);
-          return [px - 0.03 * (gx + 0.6 * n()), py - 0.03 * (gy + 0.6 * n())];
-        }),
-        walk("Momentum", ([px, py]) => {
-          const grad = slope(px, py);
-          for (const axis of [0, 1]) speed[axis] = 0.85 * speed[axis] - 8e-3 * grad[axis];
-          return [px + speed[0], py + speed[1]];
-        }),
-        walk("Adam", ([px, py], t) => {
-          const grad = slope(px, py);
-          const move = [0, 1].map((axis) => {
-            mean3[axis] = 0.9 * mean3[axis] + 0.1 * grad[axis];
-            square[axis] = 0.999 * square[axis] + 1e-3 * grad[axis] ** 2;
-            return 0.15 * mean3[axis] / (1 - 0.9 ** t) / (Math.sqrt(square[axis] / (1 - 0.999 ** t)) + 1e-8);
-          });
-          return [px - move[0], py - move[1]];
-        })
-      ]
-    };
-  }
-  /** `{ layer, kind, units }` for a small fully connected classifier. */
-  static network(seed) {
-    const units = LAYER_UNITS[Math.abs(Math.round(seed)) % LAYER_UNITS.length];
-    return units.map((count2, i) => {
-      const last = i === units.length - 1;
-      return { layer: i === 0 ? "Input" : last ? "Output" : `Hidden ${i}`, kind: i === 0 ? "Input" : last ? "Softmax" : "Dense", units: count2 };
-    });
-  }
-};
-
 // src/charts/modelcharts/AlePlot.ts
 var AleForm = /* @__PURE__ */ ((AleForm2) => {
   AleForm2["Curve"] = "curve";
@@ -265,7 +136,7 @@ var AlePlot = (_class = class extends _chunkUXFWDEP6cjs.StatisticalChart {
     shape: "table" /* Table */,
     description: "Shows a feature's accumulated local effect on a model's prediction across its quantile bins, or the interaction of two features as a heatmap.",
     tags: ["machine learning", "explainability", "accumulated local effects", "ale", "interaction"],
-    sample: (seed) => ModelSamples.ale(seed),
+    sample: (seed) => _chunkL6BGNHFLcjs.ModelSamples.ale(seed),
     variants: {
       form: Object.values(AleForm),
       density: Object.values(AleDensity),
@@ -439,7 +310,7 @@ var TITLES = {
 var UNDER = 0.16;
 var percent = (value) => `${Math.round(value * 100)}%`;
 var figure2 = (value) => +value.toFixed(Math.abs(value) < 10 ? 3 : 1);
-var DecisionCurveChart = (_class2 = class extends _chunkMCPPFRLNcjs.ClassifierCurveBase {
+var DecisionCurveChart = (_class2 = class extends _chunkDCGAAN3Zcjs.ClassifierCurveBase {
   static __initStatic2() {this.meta = {
     id: "decision-curve",
     name: "Decision curve analysis",
@@ -448,7 +319,7 @@ var DecisionCurveChart = (_class2 = class extends _chunkMCPPFRLNcjs.ClassifierCu
     shape: "table" /* Table */,
     description: "Plots the net benefit of acting on each model across threshold probabilities, against treating everyone and treating no one.",
     tags: ["machine learning", "clinical prediction", "net benefit", "dca", "threshold"],
-    sample: (seed) => _chunkMCPPFRLNcjs.ClassifierCurveBase.scored(seed),
+    sample: (seed) => _chunkDCGAAN3Zcjs.ClassifierCurveBase.scored(seed),
     variants: {
       metric: Object.values(DecisionCurveMetric),
       shade: [false, true]
@@ -575,7 +446,7 @@ var KsChart = (_class3 = class extends _chunkUXFWDEP6cjs.StatisticalChart {
     shape: "table" /* Table */,
     description: "Draws the cumulative score distributions of the two classes and marks the widest gap between them, the Kolmogorov-Smirnov statistic of a classifier.",
     tags: ["machine learning", "classifier", "kolmogorov-smirnov", "credit scoring", "separation"],
-    sample: (seed) => ModelSamples.scores(seed),
+    sample: (seed) => _chunkL6BGNHFLcjs.ModelSamples.scores(seed),
     variants: {
       against: Object.values(KsAxis),
       resolution: Object.values(KsResolution),
@@ -768,7 +639,7 @@ var LossLandscapePlot = (_class4 = class extends _chunkUXFWDEP6cjs.StatisticalCh
     shape: "grid" /* Grid */,
     description: "Draws a loss over two parameter directions as contours with the path of each optimiser across it, or as a slice along one direction.",
     tags: ["machine learning", "optimisation", "gradient descent", "contour", "training"],
-    sample: (seed) => ModelSamples.lossSurface(seed),
+    sample: (seed) => _chunkL6BGNHFLcjs.ModelSamples.lossSurface(seed),
     variants: {
       form: Object.values(LossLandscapeForm),
       shading: Object.values(LossLandscapeShading),
@@ -954,7 +825,7 @@ var NeuralNetworkDiagram = (_class5 = class extends _chunkEEF5WRRZcjs.Chart2D {
     shape: "table" /* Table */,
     description: "Draws the layers of a neural network in order, as columns of connected units or as blocks sized by the number of units in each layer.",
     tags: ["machine learning", "deep learning", "architecture", "layers", "multilayer perceptron"],
-    sample: (seed) => ModelSamples.network(seed),
+    sample: (seed) => _chunkL6BGNHFLcjs.ModelSamples.network(seed),
     variants: {
       form: Object.values(NeuralNetworkForm),
       orientation: ["horizontal" /* Horizontal */, "vertical" /* Vertical */],
@@ -1124,7 +995,7 @@ var ShapSummaryPlot = (_class6 = class extends _chunkUXFWDEP6cjs.StatisticalChar
     shape: "table" /* Table */,
     description: "Shows the SHAP value of every observation for each feature as a beeswarm coloured by the feature value, or the mean absolute SHAP value as bars.",
     tags: ["machine learning", "explainability", "shap", "beeswarm", "feature attribution"],
-    sample: (seed) => ModelSamples.shap(seed),
+    sample: (seed) => _chunkL6BGNHFLcjs.ModelSamples.shap(seed),
     variants: {
       form: Object.values(ShapSummaryForm),
       order: Object.values(ShapSummaryOrder)
@@ -1207,8 +1078,8 @@ var ShapSummaryPlot = (_class6 = class extends _chunkUXFWDEP6cjs.StatisticalChar
   drawBars(ctx, axes, features) {
     const thick = axes.y.bandwidth(), fill = features.length ? ctx.palette.color(MEAN) : "";
     const shape = (row, value) => {
-      const top = _nullishCoalesce(axes.y(row.feature), () => ( 0)), from = axes.x(0), to = Math.max(from, axes.x(value)), round2 = Math.min(4, thick / 2, to - from);
-      return `M${from},${top}H${to - round2}Q${to},${top} ${to},${top + round2}V${top + thick - round2}Q${to},${top + thick} ${to - round2},${top + thick}H${from}Z`;
+      const top = _nullishCoalesce(axes.y(row.feature), () => ( 0)), from = axes.x(0), to = Math.max(from, axes.x(value)), round = Math.min(4, thick / 2, to - from);
+      return `M${from},${top}H${to - round}Q${to},${top} ${to},${top + round}V${top + thick - round}Q${to},${top + thick} ${to - round},${top + thick}H${from}Z`;
     };
     const marks = ctx.marks("bars", "path", features, (row) => row.feature).attr("fill", fill);
     ctx.set(marks, { d: (row) => shape(row, row.mean) }, { d: (row) => shape(row, 0) }, 12);

package/dist/{chunk-SOKOWUF5.cjs → chunk-PPGWHIU7.cjs}

@@ -9,7 +9,7 @@ var _chunkAL4XQZH6cjs = require('./chunk-AL4XQZH6.cjs');
 var _chunkPY4R3IFFcjs = require('./chunk-PY4R3IFF.cjs');
 
 
-var _chunkUMEKIIHJcjs = require('./chunk-UMEKIIHJ.cjs');
+var _chunkLNWTC72Ycjs = require('./chunk-LNWTC72Y.cjs');
 
 // src/charts/statistics/VolcanoPlot.ts
 var PValueAdjustment = /* @__PURE__ */ ((PValueAdjustment2) => {
@@ -33,7 +33,7 @@ var VolcanoPlot = (_class = class extends _chunkUXFWDEP6cjs.StatisticalChart {
     shape: "table" /* Table */,
     description: "Plots effect size against significance so strong, reliable changes stand out among thousands of tests.",
     tags: ["genomics", "differential expression", "multiple testing", "p-value"],
-    sample: (seed) => _chunkUMEKIIHJcjs.SampleData.points(seed, 360, 3).map((row, i) => {
+    sample: (seed) => _chunkLNWTC72Ycjs.SampleData.points(seed, 360, 3).map((row, i) => {
       const effect = (+row.y - +row.x * 0.8 - 10) / 9 * 0.9;
       return { gene: `G${String(i + 1).padStart(3, "0")}`, log2fc: +effect.toFixed(3), pvalue: _chunkAL4XQZH6cjs.StatMath.twoSided(effect * (+row.size / 7)), mean: row.size };
     }),

package/dist/{chunk-IELWOM2J.cjs → chunk-PPO6SDAN.cjs}

@@ -18,7 +18,7 @@ var _chunkEEF5WRRZcjs = require('./chunk-EEF5WRRZ.cjs');
 var _chunkPY4R3IFFcjs = require('./chunk-PY4R3IFF.cjs');
 
 
-var _chunkUMEKIIHJcjs = require('./chunk-UMEKIIHJ.cjs');
+var _chunkLNWTC72Ycjs = require('./chunk-LNWTC72Y.cjs');
 
 // src/charts/scientific/ClimateChart.ts
 var _d3 = require('d3');
@@ -66,15 +66,15 @@ var ClimateChart = class extends _chunkEEF5WRRZcjs.Chart2D {
 // src/charts/scientific/ScientificSample.ts
 var ScientificSample = class {
   static normal(seed, count) {
-    return _chunkUMEKIIHJcjs.SampleData.distribution(seed, count, 1).map((row) => (+row.value - 40) / 8);
+    return _chunkLNWTC72Ycjs.SampleData.distribution(seed, count, 1).map((row) => (+row.value - 40) / 8);
   }
   static uniform(seed, count) {
-    return _chunkUMEKIIHJcjs.SampleData.points(seed, count, 1).map((row) => (+row.size - 3) / 30);
+    return _chunkLNWTC72Ycjs.SampleData.points(seed, count, 1).map((row) => (+row.size - 3) / 30);
   }
   /** `{ a, b, c, size, group }`, three compositional clusters. */
   static compositions(seed, count = 90) {
     const centres = [[0.62, 0.22, 0.16], [0.18, 0.62, 0.2], [0.2, 0.2, 0.6]];
-    return _chunkUMEKIIHJcjs.SampleData.points(seed, count, 3).map((row, i) => {
+    return _chunkLNWTC72Ycjs.SampleData.points(seed, count, 3).map((row, i) => {
       const centre = centres[i % 3], noise = this.normal(seed + i, 3);
       const parts = centre.map((value, k) => value * Math.exp(noise[k] * 0.38));
       const total = parts[0] + parts[1] + parts[2];
@@ -88,7 +88,7 @@ var ScientificSample = class {
       { Ca: 0.8, Mg: 0.6, Na: 8.5, K: 0.3, HCO3: 2, CO3: 0.4, SO4: 1.2, Cl: 8, Fe: 0.2 },
       { Ca: 3.2, Mg: 4.5, Na: 2, K: 0.2, HCO3: 1.5, CO3: 0.1, SO4: 7, Cl: 1.4, Fe: 0.9 }
     ];
-    return _chunkUMEKIIHJcjs.SampleData.points(seed, count, 3).map((row, i) => {
+    return _chunkLNWTC72Ycjs.SampleData.points(seed, count, 3).map((row, i) => {
       const type = types[i % 3], noise = this.normal(seed * 7 + i, 9);
       const ions = Object.fromEntries(Object.entries(type).map(([ion, value], k) => [ion, +(value * Math.exp(noise[k] * 0.55)).toFixed(3)]));
       return { sample: `W${i + 1}`, group: row.group, ...ions, tds: +(Object.values(ions).reduce((a, b) => a + b, 0) * 34).toFixed(0) };
@@ -196,7 +196,7 @@ var ScientificSample = class {
   }
   /** Detector traces: `{ time, signal, trace }`, peak trains on a drifting baseline. */
   static chromatogram(seed) {
-    const count = 420, u = this.uniform(seed, 40), walk = _chunkUMEKIIHJcjs.SampleData.timeSeries(seed, count, 1).map((row) => +row.value);
+    const count = 420, u = this.uniform(seed, 40), walk = _chunkLNWTC72Ycjs.SampleData.timeSeries(seed, count, 1).map((row) => +row.value);
     return ["Standard", "Sample"].flatMap((trace, s) => {
       const peaks = Array.from({ length: 8 }, (_, p) => ({ at: 2.2 + p * 2.05 + u[p] * 0.8, tall: 180 + u[p + 8] ** 2 * 800 * (s ? 0.6 + u[p + 16] * 0.7 : 1), wide: 0.1 + u[p + 24] * 0.13 }));
       return Array.from({ length: count }, (_, i) => {

package/dist/{chunk-6JTZZCSL.cjs → chunk-PUMKX5F2.cjs}

@@ -9,7 +9,7 @@ var _chunkSGMB5HSQcjs = require('./chunk-SGMB5HSQ.cjs');
 var _chunkEEF5WRRZcjs = require('./chunk-EEF5WRRZ.cjs');
 
 
-var _chunkUMEKIIHJcjs = require('./chunk-UMEKIIHJ.cjs');
+var _chunkLNWTC72Ycjs = require('./chunk-LNWTC72Y.cjs');
 
 // src/charts/diagrams/architectureEnums.ts
 var ArchitectureKind = /* @__PURE__ */ ((ArchitectureKind2) => {
@@ -718,7 +718,7 @@ var ArchitectureLinkLabels = class _ArchitectureLinkLabels {
 };
 
 // src/charts/diagrams/ArchitectureSamples.ts
-var optional = (seed) => +_chunkUMEKIIHJcjs.SampleData.categories(seed, 1, 1)[0].value % 2 === 0;
+var optional = (seed) => +_chunkLNWTC72Ycjs.SampleData.categories(seed, 1, 1)[0].value % 2 === 0;
 var ArchitectureSamples = class {
   static parcelTracking(seed) {
     const nodes = [
@@ -3081,7 +3081,7 @@ var SequenceDiagram = (_class9 = class _SequenceDiagram extends _chunkEEF5WRRZcj
   }}
   /** A synthetic checkout scenario, invented for illustration; the seed decides whether a receipt is sent. */
   static checkout(seed) {
-    const receipt = +_chunkUMEKIIHJcjs.SampleData.categories(seed, 1, 1)[0].value % 2 === 0;
+    const receipt = +_chunkLNWTC72Ycjs.SampleData.categories(seed, 1, 1)[0].value % 2 === 0;
     const messages = [
       { from: "Customer", to: "Web app", label: "Place order", kind: Sync },
       { from: "Web app", to: "Orders", label: "POST /orders", kind: Sync },