5
\$\begingroup\$

This code inputs a weighted edgelist, positive and negative sentiment seed words, and a target word. The program computes the sum of the weights along shortest paths from the seed words to the target and from the target to the seed words, as it generates 9 output values.

The program is very slow. Running large edgelist files takes days, rather than minutes or seconds.

How can this program be sped up?

from tkinter import Tk, X, Y, TOP, BOTTOM, LEFT, RIGHT, BOTH, END
from tkinter import filedialog, messagebox
from tkinter.ttk import Frame, Button, Entry, Label, Progressbar
import os, glob, time
import pandas as pd

root = Tk()
root.geometry("600x400+300+300")

def read_edge_list(filename):
    edges = {}
    words = set()

    with open(filename) as fp:
        lines = fp.readlines()
    
        for line in lines:
            token = line.split()
        
            if len(token) != 3:
                continue
        
            word1 = token[0]
            word2 = token[1]
            freq = token[2]
        
            words = words | {word1, word2}
        
            if not word1 in edges.keys():
                edges[word1] = {}
            
            if not word2 in edges[word1]:
                edges[word1][word2] = {}
            
            edges[word1][word2] = freq
        
    return edges, words

def read_sentiment(filename):
    with open(filename, encoding='utf-8-sig') as fp:
        lines = fp.readlines()
    
    words = {line.strip() for line in lines}
        
    return words

def read_target_word():
    word = input("Please input target word: ")
    return word

def run_shortest_path_algorithm(edges, positive, negative, target):
    positivedict = {}
    negativedict = {}

    for source in positive:
        dist1 = dijkstra(edges, source, target)
        dist2 = dijkstra(edges, target, source)
        if dist1 and dist2:
            positivedict[source] = dist1 + dist2
    
    for source in negative:
        dist1 = dijkstra(edges, source, target)
        dist2 = dijkstra(edges, target, source)
        if dist1 and dist2:
            negativedict[source] = dist1 + dist2
    
    return positivedict, negativedict

def calculate_statistics_summary(positivedict, negativedict, 
positivewords, negativewords):
    numpositive = len(positivedict)
    numnegative = len(negativedict)

    actualnumpositive = len(positivewords)
    actualnumnegative = len(negativewords)

    sumpositive = sum(positivedict.values())
    sumnegative = sum(negativedict.values())

    if actualnumpositive == 0:
        s1 = 0
    else:
        s1 = sumpositive / actualnumpositive
    
    if actualnumnegative == 0:
        s2 = 0
    else:
        s2 = sumnegative / actualnumnegative
    
    if numnegative == 0:
        s3 = 0
    else:
        s3 = s1 * numpositive / numnegative
    
    if s2 == 0:
        s4 = 0
    else:
        s4 = s3 / s2
    
    if numpositive == 0:
        s5 = 0
    else:
        s5 = sumpositive / numpositive
    
    if numnegative == 0:
        s6 = 0
    else:
        s6 = sumnegative / numnegative
    
    if numnegative == 0:
        s7 = 0
    else:
        s7 = s5 * numpositive / numnegative
    
    if s6 == 0:
        s8 = 0
    else:
        s8 = s7 / s6
    
    s9 = s3 - s2

    return [s1, s2, s3, s4, s5, s6, s7, s8, s9]

def write_output_file():
    pass

def dijkstra(graph, start, end):
    shortest_paths = {start: (None, 0)}
    current_node = start
    visited = set()

    while current_node != end:
        visited.add(current_node)
    
        if current_node not in graph:
            destinations = []
        else:
            destinations = graph[current_node].keys()
    
        weight_to_current_node = shortest_paths[current_node][1]
    
        for next_node in destinations:
            weight = int(graph[current_node][next_node]) + 
weight_to_current_node
            if next_node not in shortest_paths:
                shortest_paths[next_node] = (current_node, weight)
            else:
                current_shortest_weight = shortest_paths[next_node][1]
                if current_shortest_weight > weight:
                    shortest_paths[next_node] = (current_node, weight)
                
        next_destinations = {node: shortest_paths[node] for node in 
shortest_paths if node not in visited}
        if not next_destinations:
            return None
    
        current_node = min(next_destinations, key=lambda k: 
next_destinations[k][1])
    
    #path = []
    #while current_node is not None:
        #path.append(current_node)
        #next_node = shortest_paths[current_node][0]
        #current_node = next_node
    
    #path = path[::-1]
    #return path

    return shortest_paths[end][1]

class SentimentWindow(Frame):
    def __init__(self):
        super().__init__()
    
        self.initUI()
    
        self.initPositiveDir = None
        self.initNegativeDir = None
        self.initSaveDir = None
    
        self.summary = pd.DataFrame(columns=['S1', 'S2', 'S3', 'S4', 
'S5', 'S6', 'S7', 'S8', 'S9'])
    
    def initUI(self):
        self.master.title("Sentiment")
        self.pack(fill=BOTH, expand=True, padx=15, pady=15)
    
        frmEdges = Frame(self)
        frmEdges.pack(fill=X, expand=True)
    
        lblEdges = Label(frmEdges, text="Select the directory of edge 
list.")
        lblEdges.pack(expand=True, fill=X, side=TOP, pady=2)
    
        frmEdgesPath = Frame(frmEdges)
        frmEdgesPath.pack(expand=True, fill=X, side=BOTTOM, pady=2)
    
        self.entEdgesPath = Entry(frmEdgesPath, width=60)
        self.entEdgesPath.pack(expand=True, fill=X, side=LEFT)
    
        btnEdgesPath = Button(frmEdgesPath, width=20, text="Load 
Edges", command=self.loadEdges)
        btnEdgesPath.pack(expand=True, side=RIGHT)
    
        frmPositive = Frame(self)
        frmPositive.pack(fill=X, expand=True)
    
        lblPositive = Label(frmPositive, text="Select the positive 
file.")
        lblPositive.pack(expand=True, fill=X, side=TOP, pady=2)
    
        frmPositivePath = Frame(frmPositive)
        frmPositivePath.pack(expand=True, fill=X, side=BOTTOM, pady=2)
    
        self.entPositivePath = Entry(frmPositivePath, width=60)
        self.entPositivePath.pack(expand=True, fill=X, side=LEFT)
    
        btnPositivePath = Button(frmPositivePath, width=20, text="Load 
Positive", command=self.loadPositive)
        btnPositivePath.pack(expand=True, side=RIGHT)
    
        frmNegative = Frame(self)
        frmNegative.pack(fill=X, expand=True)
    
        lblNegative = Label(frmNegative, text="Select the negative 
file.")
        lblNegative.pack(expand=True, fill=X, side=TOP, pady=2)
    
        frmNegativePath = Frame(frmNegative)
        frmNegativePath.pack(expand=True, fill=X, side=BOTTOM, pady=2)
    
        self.entNegativePath = Entry(frmNegativePath, width=60)
        self.entNegativePath.pack(expand=True, fill=X, side=LEFT)
     
        btnNegativePath = Button(frmNegativePath, width=20, text="Load 
Negative", command=self.loadNegative)
        btnNegativePath.pack(expand=True, side=RIGHT)
    
        frmTarget = Frame(self)
        frmTarget.pack(fill=X, expand=True)
    
        lblTarget = Label(frmTarget, text="Input the target word.")
        lblTarget.pack(expand=True, fill=X, side=TOP, pady=2)
    
        self.entTarget = Entry(frmTarget)
        self.entTarget.pack(fill=X, expand=True, pady=2)
    
        frmRun = Frame(self)
        frmRun.pack(fill=X, expand=True, pady=20)
    
        self.proRun = Progressbar(frmRun, value=0)
        self.proRun.pack(fill=X, expand=True, side=LEFT)
    
        btnRun = Button(frmRun, text = "Run", width=20, 
command=self.run)
        btnRun.pack(side=RIGHT, padx=20)
    
    
    def loadEdges(self):
        edgesFolderName = filedialog.askdirectory()
    
        if edgesFolderName:
            self.entEdgesPath.delete(0, END)
            self.entEdgesPath.insert(0, edgesFolderName)
        
    def loadPositive(self):
        if self.initPositiveDir is None:
            self.initPositiveDir = "/"
        
        positiveFileName = 
filedialog.askopenfilename(initialdir=self.initPositiveDir, 
title="Open Positive File", filetypes=(("Text file", "*.txt"),))
    
        if positiveFileName:
            self.initPositiveDir = positiveFileName
            self.entPositivePath.delete(0, END)
            self.entPositivePath.insert(0, positiveFileName)

    def loadNegative(self):
        if self.initNegativeDir is None:
            self.initNegativeDir = "/"
        
        negativeFileName = 
filedialog.askopenfilename(initialdir=self.initNegativeDir, 
title="Open Positive File", filetypes=(("Text file", "*.txt"),))
    
        if negativeFileName:
            self.initNegativeDir = negativeFileName
            self.entNegativePath.delete(0, END)
            self.entNegativePath.insert(0, negativeFileName)

    def run(self):
        edgesFolderName = self.entEdgesPath.get()
        if not os.path.isdir(edgesFolderName):
            messagebox.showerror("Invalid Path", "The directory of 
edge list is invalid.")
            return
    
        positiveFileName = self.entPositivePath.get()
        if not os.path.isfile(positiveFileName):
            messagebox.showerror("Invalid Path", "The positive 
filename is invalid.")
            return
    
        negativeFileName = self.entNegativePath.get()
        if not os.path.isfile(negativeFileName):
            messagebox.showerror("Invalid Path", "The negative 
filename is invalid.")
            return
    
        targetWord = self.entTarget.get()
        if targetWord is None or len(targetWord) <= 0:
            messagebox.showerror("No Target", "Please input the target 
word.")
    
        os.chdir(edgesFolderName)
        edgefiles = glob.glob("*.pr")
    
        if len(edgefiles) <= 0:
            messagebox.showerror("No Edge File", "Cannot find the edge 
files.")

        positivewords = read_sentiment(positiveFileName)
    
        negativewords = read_sentiment(negativeFileName)
    
        self.summary.drop(self.summary.index, inplace=True)
        self.proRun["value"] = 0.0
        self.proRun.update()
        root.config(cursor="wait")
        root.update()
        time.sleep(0.300)
    
        for index, edgefile in enumerate(edgefiles):
            edges, words = read_edge_list(edgefile)
        
            if targetWord not in words:
                messagebox.showerror("Invalid Target", "Target does 
not exist in " + edgefile)
            else:
                possiblepositive = positivewords & words
                possiblenegative = negativewords & words
            
                positivedict, negativedict = \
                    run_shortest_path_algorithm(edges, 
possiblepositive, possiblenegative, targetWord)
            
                statistics_summary = 
calculate_statistics_summary(positivedict, negativedict, 
positivewords, negativewords)
                self.summary.loc[edgefile] = statistics_summary
            self.proRun["value"] = 100 * (index + 1) / len(edgefiles)
            self.proRun.update()
    
        root.config(cursor="")    
    
        if self.summary.shape[0] > 0:
            self.summary.loc['mean'] = self.summary.mean()
            self.summary.loc['std'] = self.summary.std()
        
            if self.initSaveDir is None:
                self.initSaveDir = "/"
            
            outputFile = 
filedialog.asksaveasfilename(initialdir=self.initSaveDir, title="Save 
Summary File", filetypes=(("Text file", "*.txt"),))
            self.initSaveDir = outputFile
        
            if outputFile:
                with open(outputFile, 'w') as outfp:
                    self.summary.to_string(outfp)
    
app = SentimentWindow()
root.mainloop()

Some data files for this program:

Here's the code in a file.

\$\endgroup\$
5
  • \$\begingroup\$ How are these files supposed to be opened? \$\endgroup\$
    – T145
    Nov 10, 2019 at 20:56
  • 1
    \$\begingroup\$ For all the comments in the code, I can't tell whether it tries to implement a standard algorithm for shortest path or something home-grown. (I can take a guess given the function names that it uses one as a partial solution.) \$\endgroup\$
    – greybeard
    Nov 10, 2019 at 23:48
  • \$\begingroup\$ It uses Dijkstra's algorithm. Wondering if using igraph's implementation would be better. \$\endgroup\$ Nov 11, 2019 at 3:39
  • \$\begingroup\$ It appears to me that your Dijkstra's algorithm implementation is not using a priority queue? \$\endgroup\$
    – Andrew Au
    Nov 11, 2019 at 6:47
  • \$\begingroup\$ How would a priority queue work? \$\endgroup\$ Nov 12, 2019 at 19:13

1 Answer 1

1
\$\begingroup\$

Just going to comment on style

Method Naming

Method names should be in snake_case. You got it right with the class names, as they should be PascalCase, but method names are in snake_case.

Ternary Operations

Your huge block of if statements in calculate_statistics_summary can be reduced to one line each, utilizing the ternary operator:

s1 = 0 if actualnumpositive == 0 else sumpositive / actualnumpositive
s2 = 0 if actualnumnegative == 0 else sumnegative / actualnumnegative
s3 = 0 if numnegative == 0 else s1 * numpositive / numnegative
s4 = 0 if s2 == 0 else s3 / s2
s5 = 0 if numpositive == 0 else sumpositive / numpositive
s6 = 0 if numnegative == 0 else sumnegative / numnegative
s7 = 0 if numnegative == 0 else s5 * numpositive / numnegative
s8 = 0 if s6 == 0 else s7 / s6
s9 = s3 - s2

Type Hints

You should use type hints to make it clear what types of parameters are acceptable to methods, and what type are returned, if any. For example:

def calculate_statistics_summary(positivedict, negativedict, positivewords, negativewords):

can be this (added spacing to make it readable)

def calculate_statistics_summary(
    positivedict: dict,
    negativedict: dict,
    positivewords: set,
    negativewords: set
) -> List[float]:

Now it's much clearer what types are passed to the method, and that a list containing floats (from my interpretation) is being returned.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.