/**
 * Recursive function to find new candidate medians.
 *
 * @private
 * @param {Position} candidateMedian current candidate median
 * @param {Position} previousCandidate the previous candidate median
 * @param {FeatureCollection<Point>} centroids the collection of centroids whose median we are determining
 * @param {number} counter how many attempts to try before quitting.
 * @returns {Feature<Point>} the median center of the dataset.
 */
function findMedian(candidateMedian, previousCandidate, centroids, counter) {
    var tolerance = centroids.properties.tolerance || 0.001;
    var candidateXsum = 0;
    var candidateYsum = 0;
    var kSum = 0;
    var centroidCount = 0;
    featureEach(centroids, function (theCentroid: any) {
        var weightValue = theCentroid.properties.weight;
        var weight = (weightValue === undefined || weightValue === null) ? 1 : weightValue;
        weight = Number(weight);
        if (!isNumber(weight)) throw new Error('weight value must be a number');
        if (weight > 0) {
            centroidCount += 1;
            var distanceFromCandidate = weight * distance(theCentroid, candidateMedian);
            if (distanceFromCandidate === 0) distanceFromCandidate = 1;
            var k = weight / distanceFromCandidate;
            candidateXsum += theCentroid.geometry.coordinates[0] * k;
            candidateYsum += theCentroid.geometry.coordinates[1] * k;
            kSum += k;
        }
    });
    if (centroidCount < 1) throw new Error('no features to measure');
    var candidateX = candidateXsum / kSum;
    var candidateY = candidateYsum / kSum;
    if (centroidCount === 1 || counter === 0 || (Math.abs(candidateX - previousCandidate[0]) < tolerance && Math.abs(candidateY - previousCandidate[1]) < tolerance)) {
        return point([candidateX, candidateY], {medianCandidates: centroids.properties.medianCandidates});
    } else {
        centroids.properties.medianCandidates.push([candidateX, candidateY]);
        return findMedian([candidateX, candidateY], candidateMedian, centroids, counter - 1);
    }
}

/**
 * Takes a {@link FeatureCollection} of points and calculates the median center,
 * algorithimically. The median center is understood as the point that is
 * requires the least total travel from all other points.
 *
 * Turfjs has four different functions for calculating the center of a set of
 * data. Each is useful depending on circumstance.
 *
 * `@turf/center` finds the simple center of a dataset, by finding the
 * midpoint between the extents of the data. That is, it divides in half the
 * farthest east and farthest west point as well as the farthest north and
 * farthest south.
 *
 * `@turf/center-of-mass` imagines that the dataset is a sheet of paper.
 * The center of mass is where the sheet would balance on a fingertip.
 *
 * `@turf/center-mean` takes the averages of all the coordinates and
 * produces a value that respects that. Unlike `@turf/center`, it is
 * sensitive to clusters and outliers. It lands in the statistical middle of a
 * dataset, not the geographical. It can also be weighted, meaning certain
 * points are more important than others.
 *
 * `@turf/center-median` takes the mean center and tries to find, iteratively,
 * a new point that requires the least amount of travel from all the points in
 * the dataset. It is not as sensitive to outliers as `@turf/center-mean`, but it is
 * attracted to clustered data. It, too, can be weighted.
 *
 * **Bibliography**
 *
 * Harold W. Kuhn and Robert E. Kuenne, “An Efficient Algorithm for the
 * Numerical Solution of the Generalized Weber Problem in Spatial
 * Economics,” _Journal of Regional Science_ 4, no. 2 (1962): 21–33,
 * doi:{@link https://doi.org/10.1111/j.1467-9787.1962.tb00902.x}.
 *
 * James E. Burt, Gerald M. Barber, and David L. Rigby, _Elementary
 * Statistics for Geographers_, 3rd ed., New York: The Guilford
 * Press, 2009, 150–151.
 *
 * @name centerMedian
 * @param {FeatureCollection<any>} features Any GeoJSON Feature Collection
 * @param {Object} [options={}] Optional parameters
 * @param {string} [options.weight] the property name used to weight the center
 * @param {number} [options.tolerance=0.001] the difference in distance between candidate medians at which point the algorighim stops iterating.
 * @param {number} [options.counter=10] how many attempts to find the median, should the tolerance be insufficient.
 * @returns {Feature<Point>} The median center of the collection
 * @example
 * var points = turf.points([[0, 0], [1, 0], [0, 1], [5, 8]]);
 * var medianCenter = turf.centerMedian(points);
 *
 * //addToMap
 * var addToMap = [points, medianCenter]
 */
function centerMedian(
    features: FeatureCollection<any>,
    options: { weight?: string, tolerance?: number, counter?: number} = {}
): Feature<Point, {
    medianCandidates: Array<Position>,
    [key: string]: any
}> {
    // Optional params
    options = options || {};
    if (!isObject(options)) throw new Error('options is invalid');
    var counter = options.counter || 10;
    if (!isNumber(counter)) throw new Error('counter must be a number');
    var weightTerm = options.weight;

    // Calculate mean center:
    var meanCenter = centerMean(features, {weight: options.weight});

    // Calculate center of every feature:
    var centroids: any = featureCollection([]);
    featureEach(features, function (feature) {
        centroids.features.push(centroid(feature, {properties: {weight: feature.properties[weightTerm]}}));
    });

    centroids.properties = {
        tolerance: options.tolerance,
        medianCandidates: []
    };
    return findMedian(meanCenter.geometry.coordinates, [0, 0], centroids, counter);
}

export function createBins(geojson: FeatureCollection<any>, property: string | number) {
    var bins = {};

    featureEach(geojson, function (feature, i) {
        var properties = feature.properties || {};
        if (properties.hasOwnProperty(String(property))) {
            var value = properties[property];
            if (bins.hasOwnProperty(value)) bins[value].push(i);
            else bins[value] = [i];
        }
    });
    return bins;
}

 featureEach(lineSegment(line1), (segment) => {
     featureEach(tree.search(segment), (match) => {
         const intersect = intersects(segment, match);
         if (intersect) {
             // prevent duplicate points https://github.com/Turfjs/turf/issues/688
             const key = getCoords(intersect).join(",");
             if (!unique[key]) {
                 unique[key] = true;
                 results.push(intersect);
             }
         }
     });
 });

 featureEach(lineSegment(line1), function (segment) {
     featureEach(tree.search(segment), function (match) {
         var intersect = intersects(segment, match);
         if (intersect) {
             // prevent duplicate points https://github.com/Turfjs/turf/issues/688
             var key = getCoords(intersect).join(',');
             if (!unique[key]) {
                 unique[key] = true;
                 results.push(intersect);
             }
         }
     });
 });

/**
 * Removes duplicated points in a collection returning a new collection
 *
 * @private
 * @param {FeatureCollection<Point>} points to be cleaned
 * @returns {FeatureCollection<Point>} cleaned set of points
 */
function removeDuplicates(points: FeatureCollection<Point>): FeatureCollection<Point> {
    const cleaned: Array<Feature<Point>> = [];
    const existing: {[key: string]: boolean} = {};

    featureEach(points, (pt) => {
        if (!pt.geometry) { return; }
        const key = pt.geometry.coordinates.join("-");
        if (!existing.hasOwnProperty(key)) {
            cleaned.push(pt);
            existing[key] = true;
        }
    });
    return featureCollection(cleaned);
}

export function getCluster<G extends any, P = any>(
    geojson: FeatureCollection<G, P>,
    filter: any
): FeatureCollection<G, P> {
    // Validation
    if (!geojson) throw new Error('geojson is required');
    if (geojson.type !== 'FeatureCollection') throw new Error('geojson must be a FeatureCollection');
    if (filter === undefined || filter === null) throw new Error('filter is required');

    // Filter Features
    var features: Feature<G, P>[] = [];
    featureEach<G, P>(geojson, function (feature) {
        if (applyFilter(feature.properties, filter)) features.push(feature);
    });
    return featureCollection(features);
}

/**
 * Takes any LineString or Polygon GeoJSON and returns the intersecting point(s).
 *
 * @name lineIntersect
 * @param {GeoJSON} line1 any LineString or Polygon
 * @param {GeoJSON} line2 any LineString or Polygon
 * @returns {FeatureCollection<Point>} point(s) that intersect both
 * @example
 * var line1 = turf.lineString([[126, -11], [129, -21]]);
 * var line2 = turf.lineString([[123, -18], [131, -14]]);
 * var intersects = turf.lineIntersect(line1, line2);
 *
 * //addToMap
 * var addToMap = [line1, line2, intersects]
 */
function lineIntersect<
    G1 extends LineString|MultiLineString|Polygon|MultiPolygon,
    G2 extends LineString|MultiLineString|Polygon|MultiPolygon
>(
    line1: FeatureCollection<G1> | Feature<G1> | G1,
    line2: FeatureCollection<G2> | Feature<G2> | G2,
): FeatureCollection<Point> {
    const unique: any = {};
    const results: any[] = [];

    // First, normalize geometries to features
    // Then, handle simple 2-vertex segments
    if (line1.type === "LineString") { line1 = feature(line1); }
    if (line2.type === "LineString") { line2 = feature(line2); }
    if (line1.type === "Feature" &&
        line2.type === "Feature" &&
        line1.geometry !== null &&
        line2.geometry !== null &&
        line1.geometry.type === "LineString" &&
        line2.geometry.type === "LineString" &&
        line1.geometry.coordinates.length === 2 &&
        line2.geometry.coordinates.length === 2) {
        const intersect = intersects(line1, line2);
        if (intersect) { results.push(intersect); }
        return featureCollection(results);
    }

    // Handles complex GeoJSON Geometries
    const tree = rbush();
    tree.load(lineSegment(line2));
    featureEach(lineSegment(line1), (segment) => {
        featureEach(tree.search(segment), (match) => {
            const intersect = intersects(segment, match);
            if (intersect) {
                // prevent duplicate points https://github.com/Turfjs/turf/issues/688
                const key = getCoords(intersect).join(",");
                if (!unique[key]) {
                    unique[key] = true;
                    results.push(intersect);
                }
            }
        });
    });
    return featureCollection(results);
}

    segmentEach(line2, function (segment) {
        var doesOverlaps = false;

        // Iterate over each segments which falls within the same bounds
        featureEach(tree.search(segment), function (match) {
            if (doesOverlaps === false) {
                var coordsSegment = getCoords(segment).sort();
                var coordsMatch: any = getCoords(match).sort();

                // Segment overlaps feature
                if (equal(coordsSegment, coordsMatch)) {
                    doesOverlaps = true;
                    // Overlaps already exists - only append last coordinate of segment
                    if (overlapSegment) overlapSegment = concatSegment(overlapSegment, segment);
                    else overlapSegment = segment;
                // Match segments which don't share nodes (Issue #901)
                } else if (
                    (tolerance === 0) ?
                        booleanPointOnLine(coordsSegment[0], match) && booleanPointOnLine(coordsSegment[1], match) :
                        nearestPointOnLine(match, coordsSegment[0]).properties.dist <= tolerance &&
                        nearestPointOnLine(match, coordsSegment[1]).properties.dist <= tolerance) {
                    doesOverlaps = true;
                    if (overlapSegment) overlapSegment = concatSegment(overlapSegment, segment);
                    else overlapSegment = segment;
                } else if (
                    (tolerance === 0) ?
                        booleanPointOnLine(coordsMatch[0], segment) && booleanPointOnLine(coordsMatch[1], segment) :
                        nearestPointOnLine(segment, coordsMatch[0]).properties.dist <= tolerance &&
                        nearestPointOnLine(segment, coordsMatch[1]).properties.dist <= tolerance) {
                    // Do not define (doesOverlap = true) since more matches can occur within the same segment
                    // doesOverlaps = true;
                    if (overlapSegment) overlapSegment = concatSegment(overlapSegment, match);
                    else overlapSegment = match;
                }
            }
        });

        // Segment doesn't overlap - add overlaps to results & reset
        if (doesOverlaps === false && overlapSegment) {
            features.push(overlapSegment);
            overlapSegment = undefined;
        }
    });

/**
 * Takes a reference {@link Point|point} and a FeatureCollection of Features
 * with Point geometries and returns the
 * point from the FeatureCollection closest to the reference. This calculation
 * is geodesic.
 *
 * @name nearestPoint
 * @param {Coord} targetPoint the reference point
 * @param {FeatureCollection<Point>} points against input point set
 * @returns {Feature<Point>} the closest point in the set to the reference point
 * @example
 * var targetPoint = turf.point([28.965797, 41.010086], {"marker-color": "#0F0"});
 * var points = turf.featureCollection([
 *     turf.point([28.973865, 41.011122]),
 *     turf.point([28.948459, 41.024204]),
 *     turf.point([28.938674, 41.013324])
 * ]);
 *
 * var nearest = turf.nearestPoint(targetPoint, points);
 *
 * //addToMap
 * var addToMap = [targetPoint, points, nearest];
 * nearest.properties['marker-color'] = '#F00';
 */
function nearestPoint(targetPoint: Coord, points: FeatureCollection<Point>): NearestPoint {
    // Input validation
    if (!targetPoint) throw new Error('targetPoint is required');
    if (!points) throw new Error('points is required');

    let nearest: NearestPoint;
    let minDist: number = Infinity;
    let bestFeatureIndex: number = 0;
    featureEach(points,  (pt, featureIndex) => {
        const distanceToPoint = distance(targetPoint, pt);
        if (distanceToPoint < minDist) {
            bestFeatureIndex = featureIndex;
            minDist = distanceToPoint;
        }
    });
    nearest = clone(points.features[bestFeatureIndex]);
    nearest.properties.featureIndex = bestFeatureIndex;
    nearest.properties.distanceToPoint = minDist;
    return nearest;
}

/**
 * Combines a {@link FeatureCollection} of {@link Point}, {@link LineString}, or {@link Polygon} features
 * into {@link MultiPoint}, {@link MultiLineString}, or {@link MultiPolygon} features.
 *
 * @name combine
 * @param {FeatureCollection<Point|LineString|Polygon>} fc a FeatureCollection of any type
 * @returns {FeatureCollection<MultiPoint|MultiLineString|MultiPolygon>} a FeatureCollection of corresponding type to input
 * @example
 * var fc = turf.featureCollection([
 *   turf.point([19.026432, 47.49134]),
 *   turf.point([19.074497, 47.509548])
 * ]);
 *
 * var combined = turf.combine(fc);
 *
 * //addToMap
 * var addToMap = [combined]
 */
function combine(fc: FeatureCollection<Point|LineString|Polygon>) {
    var groups = {
        MultiPoint: {coordinates: [], properties: []},
        MultiLineString: {coordinates: [], properties: []},
        MultiPolygon: {coordinates: [], properties: []}
    };

    var multiMapping = Object.keys(groups).reduce(function (memo, item) {
        memo[item.replace('Multi', '')] = item;
        return memo;
    }, {});

    function addToGroup(feature, key, multi) {
        if (!multi) {
            groups[key].coordinates.push(feature.geometry.coordinates);
        } else {
            groups[key].coordinates = groups[key].coordinates.concat(feature.geometry.coordinates);
        }
        groups[key].properties.push(feature.properties);
    }

    featureEach(fc, function (feature) {
        if (!feature.geometry) return;
        if (groups[feature.geometry.type]) {
            addToGroup(feature, feature.geometry.type, true);
        } else if (multiMapping[feature.geometry.type]) {
            addToGroup(feature, multiMapping[feature.geometry.type], false);
        }
    });

    return featureCollection(Object.keys(groups)
        .filter(function (key) {
            return groups[key].coordinates.length;
        })
        .sort()
        .map(function (key) {
            var geometry = { type: key, coordinates: groups[key].coordinates };
            var properties = { collectedProperties: groups[key].properties };
            return feature(geometry, properties);
        }));
}

/**
 * Nearest Neighbor Analysis calculates an index based the average distances
 * between points in the dataset, thereby providing inference as to whether the
 * data is clustered, dispersed, or randomly distributed within the study area.
 *
 * It returns a {@link Feature<Polygon>} of the study area, with the results of
 * the analysis attached as part of of the `nearestNeighborAnalysis` property
 * of the study area's `properties`. The attached
 * [_z_-score](https://en.wikipedia.org/wiki/Standard_score) indicates how many
 * standard deviations above or below the expected mean distance the data's
 * observed mean distance is. The more negative, the more clustered. The more
 * positive, the more evenly dispersed. A _z_-score between -2 and 2 indicates
 * a seemingly random distribution. That is, within _p_ of less than 0.05, the
 * distribution appears statistically significantly neither clustered nor
 * dispersed.
 *
 * **Remarks**
 *
 * - Though the analysis will work on any {@link FeatureCollection} type, it
 * works best with {@link Point} collections.
 *
 * - This analysis is _very_ sensitive to the study area provided.
 * If no {@link Feature<Polygon>} is passed as the study area, the function draws a box
 * around the data, which may distort the findings. This analysis works best
 * with a bounded area of interest within with the data is either clustered,
 * dispersed, or randomly distributed. For example, a city's subway stops may
 * look extremely clustered if the study area is an entire state. On the other
 * hand, they may look rather evenly dispersed if the study area is limited to
 * the city's downtown.
 *
 * **Bibliography**
 *
 * Philip J. Clark and Francis C. Evans, “Distance to Nearest Neighbor as a
 * Measure of Spatial Relationships in Populations,” _Ecology_ 35, no. 4
 * (1954): 445–453, doi:[10.2307/1931034](http://doi.org/10.2307/1931034).
 *
 * @name nearestNeighborAnalysis
 * @param {FeatureCollection<any>} dataset FeatureCollection (pref. of points) to study
 * @param {Object} [options={}] Optional parameters
 * @param {Feature<Polygon>} [options.studyArea] polygon representing the study area
 * @param {string} [options.units='kilometers'] unit of measurement for distances and, squared, area.
 * @param {Object} [options.properties={}] properties
 * @returns {Feature<Polygon>} A polygon of the study area or an approximation of one.
 * @example
 * var bbox = [-65, 40, -63, 42];
 * var dataset = turf.randomPoint(100, { bbox: bbox });
 * var nearestNeighborStudyArea = turf.nearestNeighborAnalysis(dataset);
 *
 * //addToMap
 * var addToMap = [dataset, nearestNeighborStudyArea];
 */
function nearestNeighborAnalysis(dataset: FeatureCollection<any>, options?: {
    studyArea?: Feature<Polygon>;
    units?: Units;
    properties?: Properties;
}): NearestNeighborStudyArea {
    // Optional params
    options = options || {};
    const studyArea = options.studyArea || bboxPolygon(bbox(dataset));
    const properties = options.properties || {};
    const units = options.units || 'kilometers';

    const features: Array<Feature<Point>> = [];
    featureEach(dataset, (feature) => {
        features.push(centroid(feature));
    });
    const n = features.length;
    const observedMeanDistance = features.map((feature, index) => {
        const otherFeatures = featureCollection<Point>(features.filter((f, i) => {
            return i !== index;
        }));
        // Have to add the ! to make typescript validation pass
        // see https://stackoverflow.com/a/40350534/1979085
        return distance(feature, nearestPoint(feature, otherFeatures).geometry!.coordinates, {units});
    }).reduce((sum, value) => { return sum + value; }, 0) / n;

    const populationDensity = n / convertArea(area(studyArea), 'meters', units);
    const expectedMeanDistance = 1 / (2 * Math.sqrt(populationDensity));
    const variance = 0.26136 / (Math.sqrt(n * populationDensity));
    properties.nearestNeighborAnalysis = {
        units: units,
        arealUnits: units + '²',
        observedMeanDistance: observedMeanDistance,
        expectedMeanDistance: expectedMeanDistance,
        nearestNeighborIndex: observedMeanDistance / expectedMeanDistance,
        numberOfPoints: n,
        zScore: (observedMeanDistance - expectedMeanDistance) / variance,
    };
    studyArea.properties = properties;

    return studyArea as NearestNeighborStudyArea;
}

/**
 * Takes any LineString or Polygon GeoJSON and returns the intersecting point(s).
 *
 * @name lineIntersect
 * @param {Geometry|FeatureCollection|Feature<LineString|MultiLineString|Polygon|MultiPolygon>} line1 any LineString or Polygon
 * @param {Geometry|FeatureCollection|Feature<LineString|MultiLineString|Polygon|MultiPolygon>} line2 any LineString or Polygon
 * @returns {FeatureCollection<Point>} point(s) that intersect both
 * @example
 * var line1 = turf.lineString([[126, -11], [129, -21]]);
 * var line2 = turf.lineString([[123, -18], [131, -14]]);
 * var intersects = turf.lineIntersect(line1, line2);
 *
 * //addToMap
 * var addToMap = [line1, line2, intersects]
 */
function lineIntersect<G1 extends LineString|MultiLineString|Polygon|MultiPolygon, G2 extends LineString|MultiLineString|Polygon|MultiPolygon>(
    line1: FeatureCollection<G1> | Feature<G1> | G1,
    line2: FeatureCollection<G2> | Feature<G2> | G2,
): FeatureCollection<Point> {
    var unique = {};
    var results = [];

    // First, normalize geometries to features
    // Then, handle simple 2-vertex segments
    if (line1.type === 'LineString') line1 = feature(line1);
    if (line2.type === 'LineString') line2 = feature(line2);
    if (line1.type === 'Feature' &&
        line2.type === 'Feature' &&
        line1.geometry.type === 'LineString' &&
        line2.geometry.type === 'LineString' &&
        line1.geometry.coordinates.length === 2 &&
        line2.geometry.coordinates.length === 2) {
        var intersect = intersects(line1, line2);
        if (intersect) results.push(intersect);
        return featureCollection(results);
    }

    // Handles complex GeoJSON Geometries
    var tree = rbush();
    tree.load(lineSegment(line2));
    featureEach(lineSegment(line1), function (segment) {
        featureEach(tree.search(segment), function (match) {
            var intersect = intersects(segment, match);
            if (intersect) {
                // prevent duplicate points https://github.com/Turfjs/turf/issues/688
                var key = getCoords(intersect).join(',');
                if (!unique[key]) {
                    unique[key] = true;
                    results.push(intersect);
                }
            }
        });
    });
    return featureCollection(results);
}

/**
 * Takes a set of {@link Point|points} and partition them into clusters using the k-mean .
 * It uses the [k-means algorithm](https://en.wikipedia.org/wiki/K-means_clustering)
 *
 * @name clustersKmeans
 * @param {FeatureCollection<Point>} points to be clustered
 * @param {Object} [options={}] Optional parameters
 * @param {number} [options.numberOfClusters=Math.sqrt(numberOfPoints/2)] numberOfClusters that will be generated
 * @param {boolean} [options.mutate=false] allows GeoJSON input to be mutated (significant performance increase if true)
 * @returns {FeatureCollection<Point>} Clustered Points with an additional two properties associated to each Feature:
 * - {number} cluster - the associated clusterId
 * - {[number, number]} centroid - Centroid of the cluster [Longitude, Latitude]
 * @example
 * // create random points with random z-values in their properties
 * var points = turf.randomPoint(100, {bbox: [0, 30, 20, 50]});
 * var options = {numberOfClusters: 7};
 * var clustered = turf.clustersKmeans(points, options);
 *
 * //addToMap
 * var addToMap = [clustered];
 */
function clustersKmeans(points: FeatureCollection<Point>, options: {
    numberOfClusters?: number,
    mutate?: boolean
} = {}): FeatureCollection<Point, KmeansProps> {
    // Default Params
    var count = points.features.length;
    options.numberOfClusters = options.numberOfClusters || Math.round(Math.sqrt(count / 2));

    // numberOfClusters can't be greater than the number of points
    // fallbacks to count
    if (options.numberOfClusters > count) options.numberOfClusters = count;

    // Clone points to prevent any mutations (enabled by default)
    if (options.mutate !== true) points = clone(points);

    // collect points coordinates
    var data = coordAll(points);

    // create seed to avoid skmeans to drift
    var initialCentroids = data.slice(0, options.numberOfClusters);

    // create skmeans clusters
    var skmeansResult = skmeans(data, options.numberOfClusters, initialCentroids);

    // store centroids {clusterId: [number, number]}
    var centroids = {};
    skmeansResult.centroids.forEach(function (coord, idx) {
        centroids[idx] = coord;
    });

    // add associated cluster number
    featureEach(points, function (point, index) {
        var clusterId = skmeansResult.idxs[index];
        point.properties.cluster = clusterId;
        point.properties.centroid = centroids[clusterId];
    });

    return points;
}

/**
 * Returns the closest {@link Point|point}, of a {@link FeatureCollection|collection} of points,
 * to a {@link LineString|line}. The returned point has a `dist` property indicating its distance to the line.
 *
 * @name nearestPointToLine
 * @param {FeatureCollection|GeometryCollection<Point>} points Point Collection
 * @param {Feature|Geometry<LineString>} line Line Feature
 * @param {Object} [options] Optional parameters
 * @param {string} [options.units='kilometers'] unit of the output distance property
 * (eg: degrees, radians, miles, or kilometers)
 * @param {Object} [options.properties={}] Translate Properties to Point
 * @returns {Feature<Point>} the closest point
 * @example
 * var pt1 = turf.point([0, 0]);
 * var pt2 = turf.point([0.5, 0.5]);
 * var points = turf.featureCollection([pt1, pt2]);
 * var line = turf.lineString([[1,1], [-1,1]]);
 *
 * var nearest = turf.nearestPointToLine(points, line);
 *
 * //addToMap
 * var addToMap = [nearest, line];
 */
function nearestPointToLine<P = {dist: number, [key: string]: any}>(
    points: FeatureCollection<Point> | Feature<GeometryCollection> | GeometryCollection,
    line: Feature<LineString> | LineString,
    options: {
        units?: Units,
        properties?: Properties,
    } = {},
): Feature<Point, P> {
    const units = options.units;
    const properties = options.properties || {};

    // validation
    const pts = normalize(points);
    if (!pts.features.length) { throw new Error("points must contain features"); }

    if (!line) { throw new Error("line is required"); }
    if (getType(line) !== "LineString") { throw new Error("line must be a LineString"); }

    let dist = Infinity;
    let pt: any = null;

    featureEach(pts, (point) => {
        const d = pointToLineDistance(point, line, { units });
        if (d < dist) {
            dist = d;
            pt = point;
        }
    });
    /**
     * Translate Properties to final Point, priorities:
     * 1. options.properties
     * 2. inherent Point properties
     * 3. dist custom properties created by NearestPointToLine
     */
    if (pt) { pt.properties = objectAssign({dist}, pt.properties, properties); }
    // if (pt) { pt.properties = objectAssign({dist}, pt.properties, properties); }
    return pt;
}

let arrayReduce: number[] = coordReduce(pointFeatureCollection, (prev, curr) => {
  prev.push(curr);
  return prev;
}, [], false);
let stringReduce: string = coordReduce(pointFeatureCollection, (prev, curr) => prev + curr.toString(), "", false);

propEach(pointFeatureCollection, prop => console.log(prop));

let propCombine: any = propReduce(pointFeatureCollection, (prev, curr) => {
  for (let key in curr) {
    prev[key] = curr[key];
  }
  return prev;
}, {});

featureEach(pointFeatureCollection, feature => console.log(feature));

coordAll(pointFeatureCollection);

turf.midpoint(pointFeature, pointFeature2);

turf.nearest(pointFeature, pointFeatureCollection);

turf.planepoint(pointFeature, triangleFeature);

turf.pointGrid(bbox, 20, "miles");
turf.pointGrid(bbox, 20);

turf.pointOnLine(lineFeature, pointFeature, "nauticalmiles");
turf.pointOnLine(lineFeature, pointFeature);

/**
 * @typedef {Object} DirectionalMeanLine
 * @property {number} cartesianAngle the mean angle of all lines. (measure from due earth counterclockwise).
 * @property {number} bearingAngle the mean angle of all lines. (bearing).
 * @property {number} circularVariance the extent to which features all point in the same direction.
 *  the value ranges 0-1, the bigger the value, the more variation in directions between lines.
 * @property {number} averageX the centroid of all lines.
 * @property {number} averageY the centroid of all line.
 * @property {number} averageLength the average length of line.
 * @property {number} countOfLines the count of features.
 */

/**
 * This module calculate the average angle of a set of lines, measuring the trend of it.
 * It can be used in both project coordinate system and geography coordinate system.
 * It can handle segments of line or the whole line.
 * @name directionalMean
 * @param {FeatureCollection<LineString>} lines
 * @param {object} [options={}]
 * @param {boolean} [options.planar=true] whether the spatial reference system is projected or geographical.
 * @param {boolean} [options.segment=false] whether treat a LineString as a whole or a set of segments.
 * @returns {DirectionalMeanLine} Directional Mean Line
 * @example
 *
 * var lines = turf.lineStrings([
 *   [[110, 45], [120, 50]],
 *   [[100, 50], [115, 55]],
 * ])
 * var directionalMeanLine = turf.directionalMean(lines);
 * // => directionalMeanLine
 */
export default function directionalMean(lines: FeatureCollection<LineString>, options: {
    planar?: boolean;
    segment?: boolean;
} = {}): DirectionalMeanLine {

    const isPlanar: boolean = !!options.planar; // you can't use options.planar || true here.
    const isSegment: boolean = options.segment || false;
    let sigmaSin: number = 0;
    let sigmaCos: number = 0;
    let countOfLines: number = 0;
    let sumOfLen: number = 0;
    const centroidList: Array<Feature<Point>> = [];

    if (isSegment) {
        segmentEach(lines, (currentSegment: any) => { // todo fix turf-meta's declaration file
            const [sin1, cos1]: [number, number] = getCosAndSin(currentSegment.geometry.coordinates, isPlanar);
            const lenOfLine = getLengthOfLineString(currentSegment, isPlanar);
            if (isNaN(sin1) || isNaN(cos1)) {
                return;
            } else {
                sigmaSin += sin1;
                sigmaCos += cos1;
                countOfLines += 1;
                sumOfLen += lenOfLine;
                centroidList.push(centroid(currentSegment));
            }
        });
        // planar and segment
    } else {
        // planar and non-segment
        featureEach(lines, (currentFeature: Feature<LineString>, featureIndex: number) => {
            if (currentFeature.geometry.type !== "LineString") {
                throw new Error("shold to support MultiLineString?");
            }
            const [sin1, cos1]: [number, number] = getCosAndSin(currentFeature.geometry.coordinates, isPlanar);
            const lenOfLine = getLengthOfLineString(currentFeature, isPlanar);
            if (isNaN(sin1) || isNaN(cos1)) {
                return;
            } else {
                sigmaSin += sin1;
                sigmaCos += cos1;
                countOfLines += 1;
                sumOfLen += lenOfLine;
                centroidList.push(centroid(currentFeature));
            }
        });
    }

    const cartesianAngle: number = getAngleBySinAndCos(sigmaSin, sigmaCos);
    const bearingAngle: number = bearingToCartesian(cartesianAngle);
    const circularVariance = getCircularVariance(sigmaSin, sigmaCos, countOfLines);
    const averageLength = sumOfLen / countOfLines;
    const centroidOfLines = centroid(featureCollection(centroidList));
    const [averageX, averageY]: number[] = getCoord(centroidOfLines);
    let meanLinestring;
    if (isPlanar) {
        meanLinestring = getMeanLineString([averageX, averageY], cartesianAngle, averageLength, isPlanar);
    } else {
        meanLinestring = getMeanLineString([averageX, averageY], bearingAngle, averageLength, isPlanar);
    }

    return lineString(meanLinestring, {
        averageLength,
        averageX,
        averageY,
        bearingAngle,
        cartesianAngle,
        circularVariance,
        countOfLines,
//.........这里部分代码省略.........

/**
 *
 *
 * @name distanceWeight
 * @param {FeatureCollection<any>} fc FeatureCollection.
 * @param {Object} [options] option object.
 * @param {number} [options.threshold=10000] If the distance between neighbor and
 * target features is greater than threshold, the weight of that neighbor is 0.
 * @param {number} [options.p=2] Minkowski p-norm distance parameter.
 * 1: Manhattan distance. 2: Euclidean distance. 1=<p<=infinity.
 * @param {boolean} [options.binary=false] If true, weight=1 if d <= threshold otherwise weight=0.
 *  If false, weight=Math.pow(d, alpha).
 * @param {number} [options.alpha=-1] distance decay parameter.
 * A big value means the weight decay quickly as distance increases.
 * @param {boolean} [options.standardization=false] row standardization.
 * @returns {Array<Array<number>>} distance weight matrix.
 * @example
 *
 * var bbox = [-65, 40, -63, 42];
 * var dataset = turf.randomPoint(100, { bbox: bbox });
 * var result = turf.distanceWeight(dataset);
 */
export default function distanceWeight(fc: FeatureCollection<any>, options?: {
  threshold?: number;
  p?: number;
  binary?: boolean;
  alpha?: number;
  standardization?: boolean;
}): number[][] {

  options = options || {};
  const threshold = options.threshold || 10000;
  const p = options.p || 2;
  const binary = options.binary || false;
  const alpha = options.alpha || -1;
  const rowTransform = options.standardization || false;

  const features: Array<Feature<Point>> = [];
  featureEach(fc, (feature) => {
    features.push(centroid(feature));
  });

  // computing the distance between the features
  const weights: number[][] = [];
  for (let i = 0; i < features.length; i++) {
    weights[i] = [];
  }

  for (let i = 0; i < features.length; i++) {
    for (let j = i; j < features.length; j++) {
      if (i === j) {
        weights[i][j] = 0;
      }
      const dis = pNormDistance(features[i], features[j], p);
      weights[i][j] = dis;
      weights[j][i] = dis;
    }
  }

  // binary or distance decay
  for (let i = 0; i < features.length; i++) {
    for (let j = 0; j < features.length; j++) {
      const dis: number = weights[i][j];
      if (dis === 0) {
        continue;
      }
      if (binary) {
        if (dis <= threshold) {
          weights[i][j] = 1.0;
        } else {
          weights[i][j] = 0.0;
        }
      } else {
        if (dis <= threshold) {
          weights[i][j] = Math.pow(dis, alpha);
        } else {
          weights[i][j] = 0.0;
        }
      }
    }
  }

  if (rowTransform) {
    for (let i = 0; i < features.length; i++) {
      const rowSum = weights[i].reduce((sum: number, currentVal: number) => {
        return sum + currentVal;
      }, 0);
      for (let j = 0; j < features.length; j++) {
        weights[i][j] = weights[i][j] / rowSum;
      }
    }
  }

  return weights;

}