How to use examples_by_label method in hypothesis

Best Python code snippet using hypothesis

helpers.py

Source:helpers.py

1from constants import *2import numpy as np, os, pickle, struct, socket3from copy import deepcopy4from bisect import bisect_left5import shutil67def get_difference(set1, set2):8	"""Gets set1 - set2."""9	set1 = set(set1); set2 = set(set2)10	return list(set1.difference(set2))11def get_asn_dist(asns):12	# ASN_LIST -> %13	asn_dist = np.zeros(len(ASN_LIST))14	for asn in asns:15		try:16			asn_dist[ASN_LIST[asn]] += 117		except KeyError:18			asn_dist[ASN_LIST["OTHER"]] += 119	assert np.sum(asn_dist) > 02021	return asn_dist / np.sum(asn_dist) 2223def make_mask(n):24	"""return a mask of n bits as a long integer"""25	return (2<<n-1) - 12627def dotted_quad_to_num(ip):28	"""convert decimal dotted quad string to long integer"""29	return struct.unpack('<L',socket.inet_aton(ip))[0]3031def network_mask(ip,bits):32	"""Convert a network address to a long integer"""33	return dotted_quad_to_num(ip) & make_mask(bits)3435def address_in_network(ip, net, netmask):36	"""Is an address in a network"""37	return ip & netmask == net3839def is_internal(ip):40	ip_num = dotted_quad_to_num(ip)41	for net in INTERNAL_NETWORKS:42		net = net.split("/")43		if address_in_network(ip_num,dotted_quad_to_num(net[0]),make_mask(int(net[1]))):44			return True45	return False4647def get_ip_likelihood(ip_list, _type, modify=False):4849	ip_likelihood = 050	if os.path.exists(os.path.join(METADATA_DIR, KNOWN_IP_LIST_FN)):51		known_ips = pickle.load(open(os.path.join(METADATA_DIR, KNOWN_IP_LIST_FN),'rb'))52	else:53		known_ips = {"_types": ["twitch","youtube","netflix"], "data": {"twitch": [], "youtube": [], "netflix": []}}54	_types = known_ips["_types"]55	# TODO update this as model gets more complicated56	ip_likelihood = [len(set(ip_list) & set(known_ips["data"][_type])) / len(set(ip_list)) for _type in _types]5758	if _type == "no_video":59		# no_video ips are too numerous to track60		return ip_likelihood6162	63	if modify:64		# These IPs were communicated with when accessing this service, 65		# add them to the data structure66		# TODO - incorporate some sort of frequency, staleness, likelihood thing here67		for ip in ip_list: # maybe change this to /24, although these could be v6 IP's, and there's no /24 analogy there68			known_ips["data"][_type].append(ip)69		known_ips["data"][_type] = list(set(known_ips["data"][self.type]))70		pickle.dump(known_ips, open(os.path.join(self.metadata_dir, self.known_ip_list_fn),'wb'))7172	return ip_likelihood7374class discrete_cdf:75	# Got from https://tinyurl.com/y6dlvbsb76	def __init__(self, data,weighted=False):77		self.weighted=weighted78		if weighted:79			# assume data is tuple (value, count of value)80			self._data = [el[0] for el in data]81			self._counts = [el[1] for el in data]82			self._data_len = float(np.sum(self._counts)) # "length" is number of everything83		else:84			self._data = data85			self._data_len = float(len(data))8687	def __call__(self, point):88		if self.weighted:89			return np.sum(self._counts[:bisect_left(self._data, point)]) / self._data_len90		else:91			return (len(self._data[:bisect_left(self._data, point)]) /92				self._data_len)9394def get_cdf_xy(data, logx=False, logy=False, n_points = 500, weighted=False):95	"""Returns x, cdf for your data on either log-lin or lin-lin plot."""9697	# sort it98	if weighted:99		data.sort(key = lambda val : val[0]) # sort by the value, not the weight of the value100	else:101		data.sort()102103	if logx:104		if weighted:105			if data[0][0] <= 0:106				log_low = -1; 107			else:108				log_low = np.floor(np.log10(data[0][0]))109			log_high = np.ceil(np.log10(data[-1][0]))110		else:111			if data[0] <= 0: # check for bad things before you pass them to log112				log_low = -1113			else:114				log_low = np.floor(np.log10(data[0]))115			log_high = np.ceil(np.log10(data[-1]))116		x = np.logspace(log_low,log_high,num=n_points)117	elif logy:118		# Do an inverted log scale on the y axis to get an effect like119		# .9, .99, .999, etcc120		log_low = -5121		log_high = 0122		x = np.linspace(data[0], data[-1], num=n_points)123	else:124		if weighted:125			x = np.linspace(data[0][0], data[-1][0], num=n_points)126		else:127			x = np.linspace(data[0], data[-1],num=n_points)128129	# Generate the CDF130	cdf_data_obj = discrete_cdf(data, weighted=weighted)131	cdf_data = [cdf_data_obj(point) for point in x]132133	return [x, cdf_data]134135def get_even_train_split(all_x, all_y, all_metadata, train_proportion, 136	verbose=True, is_dist=False):137	# forms train and validation sets138	# y is an array of labels, for various problem types139	# each problem type is limited by a sub-class (the one with the least examples)140	# form a training set for each problem type that maximizes # of limiting examples we train on141	# while retaining an even number of examples from each sub-class142143	# returns x_train, y_train, x_val, y_val144	# each x,y -> problem_type -> examples, labels145146	# if is_dist is true, each element of y is a list of distributions, where each 147	# distrbution represents the label for that example148149	n_problem_types = len(all_y[0])150	X = {151		"train": {i: [] for i in range(n_problem_types)},152		"val": {i: [] for i in range(n_problem_types)},153	}154	Y = {155		"train": {i: [] for i in range(n_problem_types)},156		"val": {i: [] for i in range(n_problem_types)},157	}158	metadata = {159		"train": {i: [] for i in range(n_problem_types)},160		"val": {i: [] for i in range(n_problem_types)},161	}162163	for problem_type in range(n_problem_types):164		these_labels = [_y[problem_type] for _y in all_y]165		# Number of classes for this problem type166		if is_dist:167			num_sub_classes = len(these_labels[0])168		else:169			num_sub_classes = len(set(these_labels))170		# Get the limiting sub-class for this problem type171		if is_dist:172			# Count the distribution as belonging to the class associated with the 173			# most common label174			these_labels_int = [np.argmax(el) for el in these_labels]175		else:176			these_labels_int = these_labels177		u, c = np.unique(these_labels_int, return_counts = True)178		print("Problem type: {} U: {} C: {}".format(problem_type, u,c))179		if len(c) - 1 != np.max(these_labels_int):180			raise ValueError("You need at least two examples for each sub-class -- {} -- {}.".format(u,c))181		limiting_factor = np.min(c)182		limiting_subclass = np.argmin(c)183		if verbose:184			print("Limiting number for problem type {} is {} examples, subclass {}.".format(185				problem_type, limiting_factor, limiting_subclass))186		if is_dist:187			examples_by_label = [[(x,_y, md) for x,_y, md in zip(all_x, these_labels, all_metadata) if np.argmax(_y) == y] for y in range(num_sub_classes)]188		else:189			examples_by_label = [[(x,_y, md) for x,_y, md in zip(all_x, these_labels, all_metadata) if _y == y] for y in range(num_sub_classes)]190		# Number of examples of each sub-class to use in the training set191		n_to_pull = int(limiting_factor*train_proportion)192		example_indices_by_class = {}193		# Get indices of examples of train and val to use, for each sub-class194		example_indices_by_class["train"] = [np.random.choice(range(len(this_class)), 195			size=n_to_pull, replace=False) for this_class in examples_by_label]196		example_indices_by_class["val"] = [get_difference(range(len(this_class)), train_examples_this_class) for 197			train_examples_this_class, this_class in zip(example_indices_by_class["train"], examples_by_label)]198		# Fill in the examples & labels, given the indices for each sub-class199		for k in example_indices_by_class:200			for i,example_indices in enumerate(example_indices_by_class[k]):201				[X[k][problem_type].append(examples_by_label[i][j][0]) for j in example_indices]202				[Y[k][problem_type].append(examples_by_label[i][j][1]) for j in example_indices]203				[metadata[k][problem_type].append(examples_by_label[i][j][2]) for j in example_indices]204	return X["train"], Y["train"], X["val"], Y["val"], metadata['train'], metadata['val']205206def service_tls_hostnames(hostname):207	if "googlevideo.com" in hostname:208		return "youtube"209	elif "nflxvideo.net" in hostname:210		return "netflix"211	elif "ttvnw.net" in hostname:212		return "twitch"213	else:214		raise KeyError215216def clear_path(path):217    """218    Clears the entire directory represented by path if it exists.219    """220    if not os.path.exists(path):221        return222223    for file in os.listdir(path):224        file_path = os.path.join(path, file)225        try:226            if os.path.isfile(file_path):227                os.unlink(file_path)228            elif os.path.isdir(file_path):229                shutil.rmtree(file_path)230        except Exception as e:231            print(e)232233def make_path(path):234    """235    Makes all directories represented by path236    """237    if not os.path.exists(path):
...

table_structure_analysis.py

Source:table_structure_analysis.py

1# !/usr/bin/env python2# -*- coding: utf-8 -*-3######################################################################4#5# (c) Copyright University of Southampton, 20206#7# Copyright in this software belongs to University of Southampton,8# Highfield, University Road, Southampton SO17 1BJ9#10# Created By : Juliusz Ziomek11# Created Date : 2020/09/0912# Project : GloSAT13#14######################################################################15import numpy as np16import warnings17from sklearn.cluster import DBSCAN18area = lambda box: (box[2]-box[0]) * (box[3] - box[1]) if box[2]>=box[0] and box[3]>=box[1] else 019def run_dbs_1D(cells:list, eps:int,include_outliers=True,min_samples=2) -> list: 20    '''21    Runs DBScan in 1d and returns the average values for each label.22    If outliers are detected (label = -1), each of them is appended to the average values.23    '''24    centers = np.array([cells]).reshape(-1,1)25    labels = DBSCAN(eps=eps, min_samples=min_samples).fit_predict(centers)26    examples_by_label = {label:[] for label in labels}27    mean_by_label = dict()28    for no, center in enumerate(centers):29        examples_by_label[labels[no]].append(center)30    for label in examples_by_label:31        if label!= -1 :32            mean_by_label[label] = sum(examples_by_label[label])/len(examples_by_label[label])33    return list(mean_by_label.values()) + (examples_by_label[-1] if -1 in examples_by_label.keys() and include_outliers else [])34def reconstruct_table(cells:list,table:list,eps:int) -> (list,list):35    '''36    Reconstructs the cells, given the table region using DBScan with hyperparmeter eps.37    '''38    table_width = table[2] - table[0]39    table_height = table[3] - table[1]40    #Normalise cells41    cells = [[(cell[0]-table[0])/table_width,(cell[1]-table[1])/table_height,(cell[2]-table[0])/table_width,(cell[3]-table[1])/table_height] for cell in cells]42    cells_x = [0,1]43    cells_y = [0,1]44    for cell in cells:45        cells_x += [cell[0], cell[2]]46        cells_y += [cell[1], cell[3]]   47    eps_x, eps_y = check_hyperparams(cells,eps)    48    rows = run_dbs_1D(cells_y,eps_y)49    cols = run_dbs_1D(cells_x,eps_x)50    rows = [(int)(row * table_height) + table[1] for row in rows]51    cols = [(int)(col * table_width) + table[0] for col in cols]52    return rows,cols53def reconstruct_table_coarse_and_fine(coarse_cells:list,fine_cells:list,table:list,eps:int) -> (list,list):54    '''55    Reconstructs the cells, given the table region using DBScan with hyperparmeter eps.56    '''57    table_width = table[2] - table[0]58    table_height = table[3] - table[1]59    rows = []60    cols = []61    if fine_cells!=[]:62        #Normalise cells63        fine_cells = [[(cell[0]-table[0])/table_width,(cell[1]-table[1])/table_height,(cell[2]-table[0])/table_width,(cell[3]-table[1])/table_height] for cell in fine_cells]64        cells_x = [0,1]65        cells_y = [0,1]66        for cell in fine_cells:67            cells_x += [cell[0], cell[2]]68            cells_y += [cell[1], cell[3]]69        fine_eps_x, fine_eps_y = check_hyperparams(fine_cells,eps)70        rows += run_dbs_1D(cells_y,fine_eps_y)71        cols += run_dbs_1D(cells_x,fine_eps_x)72    if coarse_cells!=[]:73        coarse_cells = [[(cell[0]-table[0])/table_width,(cell[1]-table[1])/table_height,(cell[2]-table[0])/table_width,(cell[3]-table[1])/table_height] for cell in coarse_cells]74        cells_x = [0,1]75        cells_y = [0,1]76        for cell in coarse_cells:77            cells_x += [cell[0], cell[2]]78            cells_y += [cell[1], cell[3]]  79        eps_x, eps_y = check_hyperparams(coarse_cells,eps)  80        rows += run_dbs_1D(cells_y,eps_y)81        cols += run_dbs_1D(cells_x,eps_x)82    if fine_cells!=[]:83        rows = run_dbs_1D(rows,fine_eps_y)84        cols = run_dbs_1D(cols,fine_eps_x)85    86    elif coarse_cells!=[]:87        rows = run_dbs_1D(rows,eps_y)88        cols = run_dbs_1D(cols,eps_x)89    rows = [(int)(row * table_height) + table[1] for row in rows]90    cols = [(int)(col * table_width) + table[0] for col in cols]91    return rows,cols92def check_hyperparams(cells:list,eps:int) -> (int,int):93    '''94    Check whether the eps paramter is smaller than avarega width and height of cell.95    If one of those conditions is violated, prints a warning.96    Returns adjusted hyperparameters for x and y.97    '''98    diff_x, diff_y = [], []99    for cell in cells:100        diff_x.append(cell[2] - cell[0])101        diff_y.append(cell[3] - cell[1])102    avg_diff_x = sum(diff_x)/len(diff_x)103    avg_diff_y = sum(diff_y)/len(diff_y)104    105    if avg_diff_x/2<eps:106        warnings.warn("Hyperparameter eps = {} larger than half of average cell size in x. Changing to {}".format(eps,avg_diff_x/2),RuntimeWarning)107        eps_x = avg_diff_x/2108    else:109        eps_x = eps110    if avg_diff_y/2<eps:111        warnings.warn("Hyperparameter eps = {} larger than half of average cell size in y. Changing to {}".format(eps,avg_diff_y/2),RuntimeWarning)112        eps_y =  avg_diff_y/2113    else:114        eps_y = eps115    return eps_x,eps_y116def how_much_contained(box1:list,box2:list) -> int:117    '''118    Checks how much of the first box lies inside the second one.119    '''120    area1 = area(box1)121    intersection_box = [max(box1[0],box2[0]),122                        max(box1[1],box2[1]),123                        min(box1[2],box2[2]),124                        min(box1[3],box2[3])]125    intersection_area = area(intersection_box)...

compute_centroids_plus.py

Source:compute_centroids_plus.py

1import numpy as np2import glob3from matplotlib import pyplot as plt4import os5import sys6from collections import defaultdict7from util import *8from sklearn.cluster import KMeans9def compute_centroids():10    """11    Computes centroid for each data file given.12    """13    centroids = {}14    cnts = defaultdict(int)15    idx_to_category, _ = get_category_mappings()16    K = len(idx_to_category)17    #train_examples = np.load("data/split/train_examples.npy")18    #train_labels = np.load("data/split/train_labels.npy")19    examples_by_label = [train_examples[np.where(train_labels[:] == j)] for j in range(K)]20    #for i in range(train_examples.shape[0]):21    #    idx = train_labels[i]22    #    if train_labels[i] not in examples_by_label:23    #        examples_by_label[idx] = np.array(train_examples[i], dtype=np.float32)24    #    else:25    #        examples_by_label[idx] = np.append(examples_by_label[idx],\26    #                np.array(train_examples[i], dtype=np.float32))27    for idx, category in enumerate(idx_to_category):28        if idx%10 == 0:29            print("Done with category", idx)30        kmeans = KMeans(n_clusters=5, random_state=0).fit(examples_by_label[idx])31        clusters = kmeans.cluster_centers_32        category = idx_to_category[idx]33        for a in range(5):34            name = category + "_" + str(a)35            centroids[name] = clusters[a]36    37    return centroids38def create_centroids_dir():39    """40    Create centroids directory to save results.41    """42    try:43        os.makedirs("centroids_plus_normalized/")44        os.makedirs("centroids_plus_normalized/npy")45        os.makedirs("centroids_plus_normalized/png")46    except OSError:47        pass # already exists48def save_centroids(centroids):49    """50    Save all images of centroids to centroids/png.51    Save all numpy arrays of centroids to centroids/npy.52    """53    for category, centroid in centroids.items():54        plt.imshow(np.reshape(centroid, (28, 28)), cmap='gray')55        plt.title(category)56        save_path = os.path.join("centroids_plus_normalized", category)57        plt.savefig("centroids_plus_normalized/png/"+category)58        np.save("centroids_plus_normalized/npy/"+category, centroid)59        # plt.show()60if __name__ == "__main__":61    if not os.path.isdir("data/split"):62        sys.exit("Need data directory.")63    centroids = compute_centroids()64    create_centroids_dir() 65    save_centroids(centroids)...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.