Coverage for application/utils.py : 15%

import pickle 

import time 

import regex as re 

import pathlib 

import bokeh.plotting 

import bokeh.models 

import bokeh.layouts 

import pandas as pd 

import threading 

import lxml 

import queue 

import socket 

import random 

TOOLS = "hover, pan, reset, wheel_zoom, zoom_in, zoom_out" 

JAVASCRIPT = """ 

var f = cb_obj.value; 

var options = %s; 

f = f.replace(/[!\"#$&\'()*+,-.:;<=>?@^_`{}~]/g, ""); 

f = f.replace(/\s/g, ""); 

for (var i in options) { 

if (f == options[i]) { 

console.log("Visible: " + options[i]) 

eval(options[i]).visible = true; 

} 

else { 

console.log("Unvisible: " + options[i]) 

eval(options[i]).visible = false; 

} 

} 

""" 

def compress(data, filepath): 

""" 

Dumps generated data. 

""" 

with open(filepath, "wb") as file: 

pickle.dump(data, file) 

def decompress(filepath): 

""" 

Loads dumped data. 

""" 

with open(filepath, "rb") as file: 

return pickle.load(file) 

def load_data(tempdir): 

""" 

Loads the generated data. 

""" 

data_path = str(pathlib.Path(tempdir, "data.pickle")) 

parameter_path = str(pathlib.Path(tempdir, "parameter.csv")) 

topics_path = str(pathlib.Path(tempdir, "topics.csv")) 

data = decompress(data_path) 

parameter = pd.read_csv(parameter_path, index_col=0, encoding="utf-8", sep=";") 

parameter.columns = [""] # remove column names 

topics = pd.read_csv(topics_path, index_col=0, encoding="utf-8", sep=";") 

data["parameter"] = [parameter.to_html(classes="parameter", border=0)] 

data["topics"] = [topics.to_html(classes="topics")] 

return data 

def remove_markup(content): 

""" 

Removes markup from text. If lxml fails, a simple regex is used. 

""" 

try: 

parser = lxml.etree.XMLParser(recover=True) 

tree = lxml.etree.parse(content, parser=parser) 

ns = dict(tei="http://www.tei-c.org/ns/1.0") 

lxml.etree.strip_elements(tree, "speaker", with_tail=False) 

lxml.etree.strip_elements(tree, "note", with_tail=False) 

lxml.etree.strip_elements(tree, "stage", with_tail=False) 

lxml.etree.strip_elements(tree, "head", with_tail=False) 

text = tree.xpath("//text()") 

text = "\n".join(text) 

text = re.sub(" ", "", text) 

text = re.sub(" ", "", text) 

text = re.sub("\n{1,6}", "\n", text) 

text = re.sub("\n \n", "\n", text) 

text = re.sub("\t\n", "", text) 

return text 

except: 

text = [] 

for line in content: 

line = re.sub("<.*?>", "", line) 

line = re.sub("(<.[^(><.)]+>)|<.?>", "", line) 

line = re.sub("\\n", "", line) 

line = re.sub("[ ]{2,}", " ", line) 

line = re.sub("<?(.*?)?>", "", line) 

text.append(line) 

return "".join(text) 

def boxplot(stats): 

""" 

Creates a boxplot for corpus statistics. 

""" 

x_labels = ["Document size (clean)", "Document size (raw)"] 

groups = stats.groupby("group") 

q1 = groups.quantile(q=0.25) 

q2 = groups.quantile(q=0.5) 

q3 = groups.quantile(q=0.75) 

iqr = q3 - q1 

upper = q3 + 1.5 * iqr 

lower = q1 - 1.5 * iqr 

def outliers(group): 

cat = group.name 

return group[(group.score > upper.loc[cat]["score"]) | 

(group.score < lower.loc[cat]["score"])]["score"] 

out = groups.apply(outliers).dropna() 

fig = bokeh.plotting.figure(tools="", background_fill_color="#EFE8E2", 

title="", x_range=x_labels, logo=None, 

sizing_mode="fixed", plot_width=500, 

plot_height=350) 

qmin = groups.quantile(q=0.00) 

qmax = groups.quantile(q=1.00) 

upper.score = [min([x, y]) for (x, y) in zip(list(qmax.loc[:, "score"]), upper.score)] 

lower.score = [max([x, y]) for (x, y) in zip(list(qmin.loc[:, "score"]), lower.score)] 

fig.segment(x_labels, upper.score, x_labels, q3.score, line_color="black") 

fig.segment(x_labels, lower.score, x_labels, q1.score, line_color="black") 

fig.vbar(x_labels, 0.7, q2.score, q3.score, fill_color="#729fcf", line_color="black") 

fig.vbar(x_labels, 0.7, q1.score, q2.score, fill_color="#729fcf", line_color="black") 

fig.rect(x_labels, lower.score, 0.2, 0.01, line_color="black") 

fig.rect(x_labels, upper.score, 0.2, 0.01, line_color="black") 

fig.yaxis.axis_label = "Tokens" 

fig.xgrid.grid_line_color = None 

fig.ygrid.grid_line_color = "white" 

fig.grid.grid_line_width = 2 

fig.xaxis.major_label_text_font_size = "11pt" 

fig.yaxis.major_label_text_font_size = "9pt" 

return fig 

def barchart(document_topics, height, topics=None, script=JAVASCRIPT, tools=TOOLS): 

""" 

Creates an interactive barchart for document-topics proportions. 

""" 

y_range = document_topics.columns.tolist() 

fig = bokeh.plotting.figure(y_range=y_range, plot_height=height, tools=tools, 

toolbar_location="right", sizing_mode="scale_width", 

logo=None) 

plots = {} 

options = document_topics.index.tolist() 

for i, option in enumerate(options): 

x_axis = document_topics.loc[option] 

source = bokeh.models.ColumnDataSource(dict(Describer=y_range, Proportion=x_axis)) 

option = re.sub(" ", "_", option) 

bar = fig.hbar(y="Describer", right="Proportion", source=source, 

height=0.5, color="#053967") 

if i == 0: 

bar.visible = True 

else: 

bar.visible = False 

plots[exclude_punctuations(option)] = bar 

fig.xgrid.grid_line_color = None 

fig.x_range.start = 0 

fig.select_one(bokeh.models.HoverTool).tooltips = [("Proportion", "@Proportion")] 

fig.xaxis.axis_label = "Proportion" 

fig.xaxis.major_label_text_font_size = "9pt" 

fig.yaxis.major_label_text_font_size = "9pt" 

options = list(plots.keys()) 

callback = bokeh.models.CustomJS(args=plots, code=script % options) 

if len(options) < 11: 

if topics is not None: 

selection = [" ".join(topics.iloc[i].tolist()) + " ..." for i in range(topics.shape[0])] 

menu = [(select, option) for select, option in zip(selection, options)] 

label = "Select topic to display proportions" 

else: 

menu = [(select, option) for select, option in zip(document_topics.index, options)] 

label = "Select document to display proportions" 

dropdown = bokeh.models.widgets.Dropdown(label=label, menu=menu, callback=callback) 

return bokeh.layouts.column(dropdown, fig, sizing_mode="scale_width") 

else: 

if topics is not None: 

what = "topic" 

else: 

what = "document" 

textfield = bokeh.models.widgets.AutocompleteInput(completions=document_topics.index.tolist(), 

placeholder="Type a {} name".format(what), 

css_classes=["customTextInput"], 

callback=callback) 

return bokeh.layouts.row(fig, textfield, sizing_mode="scale_width") 

def read_logfile(logfile, total_iterations): 

""" 

Reads a logfile and returns the current number of iterations. 

""" 

time.sleep(3) 

pattern = re.compile("-?\d+") 

with open(logfile, "r", encoding="utf-8") as file: 

text = file.readlines() 

line = text[-1][:-1] 

wait = "Still initializing LDA topic model ..." 

info = "Iteration {0} of {1} ..." 

i = pattern.findall(line)[0] 

if "likelihood" in line: 

return i, info.format(i, total_iterations) 

elif "n_documents" in line: 

return 0, wait 

elif "vocab_size" in line: 

return 0, wait 

elif "n_words" in line: 

return 0, wait 

elif "n_topics" in line: 

return 0, wait 

elif "n_iter" in line: 

return 0, wait 

def enthread(target, args): 

""" 

Threads a process. 

""" 

q = queue.Queue() 

def wrapper(): 

q.put(target(*args)) 

t = threading.Thread(target=wrapper) 

t.start() 

return q 

def is_connected(host="8.8.8.8", port=53, timeout=3): 

""" 

Checks if your machine is connected to the internet. 

Host: 8.8.8.8 (google-public-dns-a.google.com) 

OpenPort: 53/tcp 

Service: domain (DNS/TCP) 

""" 

try: 

socket.setdefaulttimeout(timeout) 

socket.socket(socket.AF_INET, socket.SOCK_STREAM).connect((host, port)) 

return "include" 

except: 

return "exclude" 

def exclude_punctuations(s): 

""" 

Excludes punctuations from a string. 

""" 

exclude = set("!\"#$&'()*+,-.:;<=>?@^_`{}~") 

s = "".join(c for c in s if c not in exclude) 

return re.sub(" ", "", s) 

def unlink_content(directory, pattern="*"): 

""" 

Deletes the content of a directory. 

""" 

for p in pathlib.Path(directory).rglob(pattern): 

if p.is_file(): 

p.unlink()