Python Efficient way of parsing string to list of floats from a file

Question

This document has a word and tens of thousands of floats per line, I want to transform it to a dictionary with the word as key and a vector with all the floats. That is how I am doing, but due to the size of the file (about 20k lines each one with about 10k values) the process is taking a bit too long. I could not find a more efficient way of doing the parsing. Just some alternative ways that were not guaranteed to decrease run time.

with open("googlenews.word2vec.300d.txt") as g_file:
  i = 0;
  #dict of words: [lots of floats]
  google_words = {}

  for line in g_file:
    google_words[line.split()[0]] = [float(line.split()[i]) for i in range(1, len(line.split()))]

Answer 1

In your solution you preform slow line.split() for every word, twice. Consider following modification:

with open("googlenews.word2vec.300d.txt") as g_file:
    i = 0;
    #dict of words: [lots of floats]
    google_words = {}

    for line in g_file:
        word, *numbers = line.split()
        google_words[word] = [float(number) for number in numbers]

One advanced concept I used here is "unpacking": word, *numbers = line.split()

Python allows to unpack iterable values into multilple variables:

a, b, c = [1, 2, 3]
# This is practically equivalent to
a = 1
b = 2
c = 3

The * is a shortcut for "take the leftovers, put them in the list and assign the list to the name":

a, *rest = [1, 2, 3, 4]
# results in
a == 1
rest == [2, 3, 4]

Answer 2

Just don't call line.split() more than once.

with open("googlenews.word2vec.300d.txt") as g_file:
    i = 0;
    #dict of words: [lots of floats]
    google_words = {}

    for line in g_file:
        temp = line.split()
        google_words[temp[0]] = [float(temp[i]) for i in range(1, len(temp))]

Here's a simple generator of such file:

s = "x"
for i in range (10000):
    s += " 1.2345"
print (s)

The former version takes some time. The version with only one split call is instant.

Answer 3

You could also use the csv module, which should be more efficient that what you are doing.

It would be something like:

import csv

d = {}
with (open("huge_file_so_huge.txt", "r")) as g_file:
    for row in csv.reader(g_file, delimiter=" "):
        d[row[0]] = list(map(float, row[1:]))