001-generate-points.py

import numpy as np
import math
import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = [5, 7]
import os
from subprocess import call, check_call, check_output
from scipy.spatial import distance_matrix
import json

class Drawer:
    def __init__(self, n):
        self.totalPoints = n
    
    def drawSomething(self, n, xBounds, yBounds):
        methodnr = np.random.randint(1,4)
        if methodnr == 1:
            return self.matrixArrange(n, 4, xBounds, yBounds)
        elif methodnr == 2:
            return self.matrixAlternateShift(n, 4, xBounds, yBounds)
        elif methodnr == 3:
            return self.trianglePoints(xBounds, yBounds)
        else:
            return self.diagonalPoints(n, xBounds, yBounds)

    def matrixArrange(self, n, nRows, xBounds, yBounds):
        nCols = math.floor(n/nRows)
        #print(nCols)
        x = np.tile(np.linspace(xBounds[0],xBounds[1],nCols), nRows)
        y = np.linspace(yBounds[0],yBounds[1],nRows).repeat(nCols)
        seq = np.stack((x,y), axis=-1)
        return seq.T, seq.shape[0]

    def matrixAlternateShift(self, n, nRows, xBounds, yBounds):
        if nRows < 2:
            return self.matrixArrange(n, nRows, xBounds, yBounds)
        # get default matrix
        seq, n = self.matrixArrange(n, nRows, xBounds, yBounds)
        nCols =int(n/nRows)
        shift = 1
        s = np.tile(np.array([0,0]), (nCols,1))
        # shift along x
        p = np.tile(np.array([shift,0]), (nCols,1))
        # shift alternate rows or cols
        fin = np.tile(np.vstack((s,p)), (math.ceil(nRows/2),1))[:n]
        #print(fin.shape, seq.shape)
        return (seq+fin.T), n

    def randomPoints(self, n, xBounds, yBounds):
        x = np.random.uniform(xBounds[0], xBounds[1], size=n)
        y = np.random.uniform(yBounds[0], yBounds[1], size=n)
        self.totalPoints -= x.shape[0]
        return np.array([x, y]), n

    def trianglePoints(self, xBounds, yBounds):
        bottomLeft = [xBounds[0], yBounds[0]]
        bottomRight = [xBounds[1], yBounds[0]]
        top = [(xBounds[1]-xBounds[0])/2, yBounds[1]]
        return np.array([bottomLeft, bottomRight, top]).T, 3

    def diagonalPoints(self, n, xBounds, yBounds):
        x = np.linspace(xBounds[0], xBounds[1], n)
        y = np.linspace(yBounds[0], yBounds[1], n)
        m = np.array([x, y])
        return m, n

pointsToGenerate = [10,20,30,40,50,60,70,80]
samplesNr = 10
for points in pointsToGenerate:
    finished = False
    it = 1
    while not finished:
        correctPoints = points
        totalPoints = correctPoints + round(points*0.15)
        draw = Drawer(totalPoints)

        xLimits = [0.5,14.5]
        yLimits = [0.5,9.5]
        initialRandomNr = 5
        [x, y], n = draw.randomPoints(initialRandomNr, xLimits, yLimits)

        # sums to totalPoints
        nrInterval = np.random.dirichlet(np.ones(4))
        nrSequence = np.ceil(nrInterval*(totalPoints))

        nrInterval = np.random.dirichlet(np.ones(2))
        xInterval = np.floor(nrInterval*xLimits[1])
        yInterval = np.floor(nrInterval*yLimits[1])

        xTot = x.tolist()
        yTot = y.tolist()
        conta = 0
        currentY = yLimits[0]
        for j in yInterval:
            currentX = xLimits[0]
            for i in xInterval:
                xBounds = [currentX, i]
                yBounds = [currentY, j]
                seq, n = draw.drawSomething(nrSequence[conta], xBounds, yBounds)
                x, y = seq
                print(x.shape)
                xTot += x.tolist()
                yTot += y.tolist()
                #plt.scatter(x, y, c='white', edgecolors='black')
                currentX += i
                conta += 1
            currentY += j

        # remove elements in eccess
        if len(xTot)>correctPoints:
            xTot = xTot[:correctPoints]
            yTot = yTot[:correctPoints]
        else:
            print('Error, not enough values', len(xTot), correctPoints)
            continue
        #print(len(xTot), correctPoints)

        c = [[i,yTot[n]] for n,i in enumerate(xTot)]
        uniquenr = np.unique(np.array(c), axis=0)
        xTot, yTot = uniquenr.T
        if (xTot.size != correctPoints or yTot.size != correctPoints):
            print('Error, there were duplicate values', xTot.size, yTot.size, correctPoints)
            continue

        # plot points
        fig, ax = plt.subplots(2,1)
        ax[0].scatter(xTot, yTot, c='white', edgecolors='black')
        ax[0].set_facecolor('xkcd:green')

        ax[1].scatter(xTot, yTot, c='white', edgecolors='black')
        ax[1].set_facecolor('xkcd:green')
        for i, x in enumerate(xTot):
            ax[1].annotate(i, (xTot[i], yTot[i]))
        plt.show()

        keepPoints = input("Should I keep this point distribution? [y/n]")
        if str(keepPoints) == 'n':
            print('passo')
            continue

        if not os.path.exists('./points/{}'.format(correctPoints)):
            os.mkdir('./points/{}'.format(correctPoints))

        # compute distance matrix
        dist = distance_matrix(uniquenr, uniquenr)

        # save OPL data file
        if not os.path.exists('./points/{}'.format(correctPoints)):
            os.mkdir('./points/{}'.format(correctPoints))
        with open('./points/{}/{}.dat'.format(correctPoints,it), 'w') as datFile:
            datFile.write('totalHoles = {};\n'.format(len(xTot)))
            datFile.write('zeroHoleID = {};\n'.format(np.random.randint(0, len(xTot))+1))
            datFile.write('C = {};'.format(json.dumps(dist.tolist())))
            # store numpy format files
            np.savez('./points/{}/{}'.format(correctPoints,it), dist=dist, xTot=xTot, yTot=yTot)

            if it == samplesNr:
                finished = True
            it += 1