gridviz 2.0.38 → 2.0.39

package/src/style/KernelSmoothingOldStyle.js

@@ -1,221 +1,221 @@
-//@ts-check
-'use strict'
-
-import { Style } from '../Style.js'
-
-/**
- * A style representing the cell as a smoothed layer, to smoothing local variations and show main trends across space.
- *
- * @author Julien Gaffuri
- */
-export class KernelSmoothingStyle extends Style {
-    //bandwidth
-    // https://observablehq.com/@uwdata/fast-kde
-    // https://observablehq.com/d/3127b6d89ada959f
-    //TODO use https://github.com/uwdata/fast-kde ?
-    //TODO https://observablehq.com/@sahilchinoy/areal-interpolation-iii ?
-
-    /** @param {object} opts */
-    constructor(opts) {
-        super(opts)
-        opts = opts || {}
-
-        /** A function specifying the value to consider for each cell. This is the value to smooth.
-         * @type {function(import("../Dataset").Cell):number} */
-        this.value = opts.value
-
-        /** The smoothing parameter, in geo unit. The larger, the more smoothed.
-         * @type {function(number,number):number}
-         */
-        this.sigma = opts.sigma // (r, zf)=>...
-
-        /** The styles to represent the smoothed cells.
-         * @type {Array.<Style>}
-         */
-        this.styles = opts.styles
-    }
-
-    /**
-     * Compute the kernel matrix, that is the matrix of the weights.
-     * One quadrant is necessary only, since it is symetrical (along both x and y axes).
-     * @param {number} s Sigma (in grid pixel !)
-     * @returns {Array.<Array<number>>}
-     * @private
-     */
-    getKernelMatrix(s) {
-        //the size of the kernel: lets limit that to ~3 times the standard deviation, as an approximation.
-        const kernelSize = Math.floor(3.25 * s)
-
-        //pre-compute coefficient for gaussian computation, to avoid computing them every time.
-        const c2 = 2 * s * s
-
-        /**
-         * The gaussian function. TODO expose that function as a parameter, to use other kernels ?
-         * @param {number} x
-         * @param {number} y
-         * @returns {number}
-         */
-        const gaussian = (x, y) => (c2 != 0 ? Math.exp(-(x * x + y * y) / c2) : x == 0 && y == 0 ? 1 : 0) //dirac case
-
-        /** @type {Array.<Array<number>>} */
-        const kw = []
-        for (let wi = 0; wi <= kernelSize; wi++) {
-            /** @type {Array<number>} */
-            const col = []
-            for (let wj = 0; wj <= kernelSize; wj++) {
-                //compute weight at wi,wj
-                col.push(gaussian(wi, wj))
-            }
-            kw.push(col)
-        }
-
-        /*
-        //debug: show values
-        for (let wj = kernelSize-1; wj >=0; wj--) {
-            let st = "";
-            for (let wi = 0; wi <= kernelSize; wi++) {
-                st += "   " + Math.floor(kw[wi][wj].w *1000) /1000
-            }
-            console.log(st)
-        }*/
-
-        return kw
-    }
-
-    /**
-     * Compute kernel smoothing of some cells.
-     *
-     * @private
-     * @param {Array.<import("../Dataset").Cell>} cells The cells to be smoothed.
-     * @param {import("../Dataset").Envelope} e Geo envelope to consider.
-     * @param {number} r Resolution, in geo unit.
-     * @param {number} s Sigma (in grid pixel !)
-     * @returns {Array.<import("../Dataset").Cell>} The list of cells, including the initial ones and the ones with smoothed values, in "ksmval" property.
-     */
-    kernelSmoothing(cells, e, r, s) {
-        //filter
-        if (this.filter) cells = cells.filter(this.filter)
-
-        const wThr = 0.01 //gaussian weights below this value will be ignored
-
-        //compute extent, in grid position
-        const xMin = Math.floor(e.xMin / r) * r
-        const xMax = Math.floor(e.xMax / r) * r
-        const yMin = Math.floor(e.yMin / r) * r
-        const yMax = Math.floor(e.yMax / r) * r
-
-        //compute matrix dimensions
-        const nbX = (xMax - xMin) / r + 1
-        const nbY = (yMax - yMin) / r + 1
-
-        //initialise smoothed value to 0
-        for (const c of cells) c['ksmval'] = 0
-
-        //make matrix. add input cells by i/j (grid position)
-        const index = {}
-        for (const c of cells) {
-            // i,j of the cell
-            const i = Math.floor((c.x - xMin) / r)
-            const j = Math.floor((c.y - yMin) / r)
-            if (!index[i]) index[i] = {}
-            index[i][j] = c
-        }
-
-        /** Check that the cell i,j is within the frame */
-        const isWithinFrame = (i, j) => i >= 0 && i < nbX && j >= 0 && j < nbY
-
-        //get kernel matrix
-        /** @type {Array.<Array.<number>>} */
-        const kernelMatrix = this.getKernelMatrix(s)
-        const kernelSize = kernelMatrix.length
-
-        //compute summ of the weights over entire kernel window
-        /** @type {number} */
-        let sumWeights = 0
-        for (let ki = 1 - kernelSize; ki < kernelSize; ki++) {
-            for (let kj = 1 - kernelSize; kj < kernelSize; kj++) {
-                let w = kernelMatrix[Math.abs(ki)][Math.abs(kj)]
-                if (w < wThr) continue
-                sumWeights += w
-            }
-        }
-
-        /** Add v as a contribution to the cell i,j */
-        const addContributionTo = (i, j, v) => {
-            //get cell at (i,j)
-            const c_ = index[i] ? index[i][j] : undefined
-
-            if (c_) {
-                //cell exists: add contribution
-                if (c_.ksmval) c_.ksmval += v
-                else c_.ksmval = v
-            } else {
-                //cell does not exist: create a new one with the smoothed value
-                if (!index[i]) index[i] = {}
-                index[i][j] = { x: xMin + i * r, y: yMin + j * r, ksmval: v }
-            }
-        }
-
-        //compute smoothing, input cell by input cell
-        for (const c of cells) {
-            /** get value of cell c
-             * @type {number} */
-            const val = this.value(c)
-            if (!val) continue
-
-            //cell matrix coordinates
-            const i = Math.floor((c.x - xMin) / r)
-            const j = Math.floor((c.y - yMin) / r)
-
-            //add contributions to smoothed values
-            for (let ki = 1 - kernelSize; ki < kernelSize; ki++) {
-                for (let kj = 1 - kernelSize; kj < kernelSize; kj++) {
-                    //check cell is within the frame
-                    if (!isWithinFrame(i + ki, j + kj)) continue
-
-                    //get contribution (ki,kj)
-                    let w = kernelMatrix[Math.abs(ki)][Math.abs(kj)]
-                    if (!w || w < wThr) continue
-                    let v = w * val
-                    if (!v) continue
-                    v /= sumWeights
-                    if (!v) continue
-
-                    //add contribution
-                    addContributionTo(i + ki, j + kj, v)
-                }
-            }
-        }
-
-        //make output list
-        const out = []
-        for (let i of Object.keys(index)) for (const j of Object.keys(index[i])) out.push(index[i][j])
-
-        return out
-    }
-
-    /**
-     * Draw the smoothed cells depending on the list of styles specified.
-     *
-     * @param {Array.<import("../Dataset").Cell>} cells
-     * @param {number} r
-     * @param {import("../GeoCanvas").GeoCanvas} cg
-     */
-    draw(cells, r, cg) {
-        if (!cells || cells.length == 0) return
-
-        //get smoothing param in geo unit
-        /** @type {number} */
-        const sG = this.sigma(r, cg.zf)
-
-        //apply kernel smoothing
-        cells = this.kernelSmoothing(cells, cg.extGeo, r, sG / r)
-
-        //draw smoothed cells from styles
-        for (let s of this.styles) s.draw(cells, r, cg)
-
-        //update legends
-        //for (let s of this.styles)
-        //    s.updateLegends({ style: s, r: r, zf: cg.getZf() });
-    }
-}
+//@ts-check
+'use strict'
+
+import { Style } from '../Style.js'
+
+/**
+ * A style representing the cell as a smoothed layer, to smoothing local variations and show main trends across space.
+ *
+ * @author Julien Gaffuri
+ */
+export class KernelSmoothingStyle extends Style {
+    //bandwidth
+    // https://observablehq.com/@uwdata/fast-kde
+    // https://observablehq.com/d/3127b6d89ada959f
+    //TODO use https://github.com/uwdata/fast-kde ?
+    //TODO https://observablehq.com/@sahilchinoy/areal-interpolation-iii ?
+
+    /** @param {object} opts */
+    constructor(opts) {
+        super(opts)
+        opts = opts || {}
+
+        /** A function specifying the value to consider for each cell. This is the value to smooth.
+         * @type {function(import("../Dataset").Cell):number} */
+        this.value = opts.value
+
+        /** The smoothing parameter, in geo unit. The larger, the more smoothed.
+         * @type {function(number,number):number}
+         */
+        this.sigma = opts.sigma // (r, zf)=>...
+
+        /** The styles to represent the smoothed cells.
+         * @type {Array.<Style>}
+         */
+        this.styles = opts.styles
+    }
+
+    /**
+     * Compute the kernel matrix, that is the matrix of the weights.
+     * One quadrant is necessary only, since it is symetrical (along both x and y axes).
+     * @param {number} s Sigma (in grid pixel !)
+     * @returns {Array.<Array<number>>}
+     * @private
+     */
+    getKernelMatrix(s) {
+        //the size of the kernel: lets limit that to ~3 times the standard deviation, as an approximation.
+        const kernelSize = Math.floor(3.25 * s)
+
+        //pre-compute coefficient for gaussian computation, to avoid computing them every time.
+        const c2 = 2 * s * s
+
+        /**
+         * The gaussian function. TODO expose that function as a parameter, to use other kernels ?
+         * @param {number} x
+         * @param {number} y
+         * @returns {number}
+         */
+        const gaussian = (x, y) => (c2 != 0 ? Math.exp(-(x * x + y * y) / c2) : x == 0 && y == 0 ? 1 : 0) //dirac case
+
+        /** @type {Array.<Array<number>>} */
+        const kw = []
+        for (let wi = 0; wi <= kernelSize; wi++) {
+            /** @type {Array<number>} */
+            const col = []
+            for (let wj = 0; wj <= kernelSize; wj++) {
+                //compute weight at wi,wj
+                col.push(gaussian(wi, wj))
+            }
+            kw.push(col)
+        }
+
+        /*
+        //debug: show values
+        for (let wj = kernelSize-1; wj >=0; wj--) {
+            let st = "";
+            for (let wi = 0; wi <= kernelSize; wi++) {
+                st += "   " + Math.floor(kw[wi][wj].w *1000) /1000
+            }
+            console.log(st)
+        }*/
+
+        return kw
+    }
+
+    /**
+     * Compute kernel smoothing of some cells.
+     *
+     * @private
+     * @param {Array.<import("../Dataset").Cell>} cells The cells to be smoothed.
+     * @param {import("../Dataset").Envelope} e Geo envelope to consider.
+     * @param {number} r Resolution, in geo unit.
+     * @param {number} s Sigma (in grid pixel !)
+     * @returns {Array.<import("../Dataset").Cell>} The list of cells, including the initial ones and the ones with smoothed values, in "ksmval" property.
+     */
+    kernelSmoothing(cells, e, r, s) {
+        //filter
+        if (this.filter) cells = cells.filter(this.filter)
+
+        const wThr = 0.01 //gaussian weights below this value will be ignored
+
+        //compute extent, in grid position
+        const xMin = Math.floor(e.xMin / r) * r
+        const xMax = Math.floor(e.xMax / r) * r
+        const yMin = Math.floor(e.yMin / r) * r
+        const yMax = Math.floor(e.yMax / r) * r
+
+        //compute matrix dimensions
+        const nbX = (xMax - xMin) / r + 1
+        const nbY = (yMax - yMin) / r + 1
+
+        //initialise smoothed value to 0
+        for (const c of cells) c['ksmval'] = 0
+
+        //make matrix. add input cells by i/j (grid position)
+        const index = {}
+        for (const c of cells) {
+            // i,j of the cell
+            const i = Math.floor((c.x - xMin) / r)
+            const j = Math.floor((c.y - yMin) / r)
+            if (!index[i]) index[i] = {}
+            index[i][j] = c
+        }
+
+        /** Check that the cell i,j is within the frame */
+        const isWithinFrame = (i, j) => i >= 0 && i < nbX && j >= 0 && j < nbY
+
+        //get kernel matrix
+        /** @type {Array.<Array.<number>>} */
+        const kernelMatrix = this.getKernelMatrix(s)
+        const kernelSize = kernelMatrix.length
+
+        //compute summ of the weights over entire kernel window
+        /** @type {number} */
+        let sumWeights = 0
+        for (let ki = 1 - kernelSize; ki < kernelSize; ki++) {
+            for (let kj = 1 - kernelSize; kj < kernelSize; kj++) {
+                let w = kernelMatrix[Math.abs(ki)][Math.abs(kj)]
+                if (w < wThr) continue
+                sumWeights += w
+            }
+        }
+
+        /** Add v as a contribution to the cell i,j */
+        const addContributionTo = (i, j, v) => {
+            //get cell at (i,j)
+            const c_ = index[i] ? index[i][j] : undefined
+
+            if (c_) {
+                //cell exists: add contribution
+                if (c_.ksmval) c_.ksmval += v
+                else c_.ksmval = v
+            } else {
+                //cell does not exist: create a new one with the smoothed value
+                if (!index[i]) index[i] = {}
+                index[i][j] = { x: xMin + i * r, y: yMin + j * r, ksmval: v }
+            }
+        }
+
+        //compute smoothing, input cell by input cell
+        for (const c of cells) {
+            /** get value of cell c
+             * @type {number} */
+            const val = this.value(c)
+            if (!val) continue
+
+            //cell matrix coordinates
+            const i = Math.floor((c.x - xMin) / r)
+            const j = Math.floor((c.y - yMin) / r)
+
+            //add contributions to smoothed values
+            for (let ki = 1 - kernelSize; ki < kernelSize; ki++) {
+                for (let kj = 1 - kernelSize; kj < kernelSize; kj++) {
+                    //check cell is within the frame
+                    if (!isWithinFrame(i + ki, j + kj)) continue
+
+                    //get contribution (ki,kj)
+                    let w = kernelMatrix[Math.abs(ki)][Math.abs(kj)]
+                    if (!w || w < wThr) continue
+                    let v = w * val
+                    if (!v) continue
+                    v /= sumWeights
+                    if (!v) continue
+
+                    //add contribution
+                    addContributionTo(i + ki, j + kj, v)
+                }
+            }
+        }
+
+        //make output list
+        const out = []
+        for (let i of Object.keys(index)) for (const j of Object.keys(index[i])) out.push(index[i][j])
+
+        return out
+    }
+
+    /**
+     * Draw the smoothed cells depending on the list of styles specified.
+     *
+     * @param {Array.<import("../Dataset").Cell>} cells
+     * @param {number} r
+     * @param {import("../GeoCanvas").GeoCanvas} cg
+     */
+    draw(cells, r, cg) {
+        if (!cells || cells.length == 0) return
+
+        //get smoothing param in geo unit
+        /** @type {number} */
+        const sG = this.sigma(r, cg.zf)
+
+        //apply kernel smoothing
+        cells = this.kernelSmoothing(cells, cg.extGeo, r, sG / r)
+
+        //draw smoothed cells from styles
+        for (let s of this.styles) s.draw(cells, r, cg)
+
+        //update legends
+        //for (let s of this.styles)
+        //    s.updateLegends({ style: s, r: r, zf: cg.getZf() });
+    }
+}