@time_this(lambda x: len(x))
def numba_fast_cusum(values: np.ndarray) -> np.ndarray:
return _numba_fast_cusum(values)
@time_this(lambda x: len(x))
def np_fast_cusum(values: np.ndarray) -> np.ndarray:
"""
This is O(n) and optimized with C code
"""
return values.cumsum()
if __name__ == '__main__':
exp_range = ExponentialRange(0, 8, 1 / 4)
values = random_numeric_list(exp_range.max)
with timed_report():
for i in exp_range.iterator(4):
slow_cusum(values[:i])
for i in exp_range.iterator(4):
slow_cusum_expanded(values[:i])
for i in exp_range.iterator():
python_fast_cusum(values[:i])
for i in exp_range.iterator():
pandas_fast_cusum(pd.Series(values[:i]))
return result_count
if __name__ == '__main__':
# the_words = ['A', 'A', 'A', 'B', 'B', 'B', 'C'] * 100
# print(slow_count_occurrences(the_words))
# print(fast_count_occurrences(the_words))
# print(defaultdict_fast_count(the_words))
# print(counter_fast_count(the_words))
# print(np_fast_count(np.array(the_words)))
# print(pd_fast_count(pd.Series(the_words)))
# print(parallel_fast_count(the_words))
exp_range = ExponentialRange(0, 7, 1 / 4)
the_words = random_words(exp_range.max)
the_array = np.array(the_words)
the_series = pd.Series(the_words)
with timed_report():
for i in exp_range.iterator(4):
slow_count_occurrences(the_words[:i])
for i in exp_range.iterator():
fast_count_occurrences(the_words[:i])
for i in exp_range.iterator():
defaultdict_fast_count(the_words[:i])
for i in exp_range.iterator():
"""
Sizing information for dictionaries
"""
import sys
from utils.profiler import ExponentialRange
exp_range = ExponentialRange(0, 7, 1 / 8)
for i in exp_range.iterator():
_dict = {j: j**2 for j in range(i)}
_dict_size = sys.getsizeof(_dict)
print(f'{len(_dict):<8} keys {_dict_size:>12} bytes')
# Returns ...
# 1 keys 248 bytes
# 2 keys 248 bytes
# 3 keys 248 bytes
# 4 keys 248 bytes
# 5 keys 248 bytes
# 7 keys 376 bytes
# 10 keys 376 bytes
# 13 keys 656 bytes
# 17 keys 656 bytes
# 23 keys 1192 bytes
# 31 keys 1192 bytes
# 42 keys 1192 bytes
# 56 keys 2288 bytes
# 74 keys 2288 bytes
# 100 keys 4712 bytes
# 133 keys 4712 bytes
# 177 keys 9328 bytes
n = len(values)
is_sorted = all(values[i] >= values[i + 1] for i in range(n - 1))
assert is_sorted, 'values are not sorted.'
def assert_top_k(top_k_values, values):
assert_sorted(top_k_values)
kth_value = top_k_values[-1]
k = len(top_k_values)
assert sum(v >= kth_value for v in values) == k, \
'Something went wrong'
if __name__ == '__main__':
exp_range = ExponentialRange(2, 7, 1 / 4)
values = random_numeric_list(exp_range.max)
with timed_report():
for i in exp_range.iterator():
_values = values[:i].copy()
_top_k = naive_find_top_k(_values)
assert_top_k(_top_k, _values)
for i in exp_range.iterator():
_values = values[:i].copy()
_top_k = heap_find_top_k(_values)
assert_top_k(_top_k, _values)
for i in exp_range.iterator():
_values = values[:i].copy()
Generate sample CSV files for file-reading tests
"""
import numpy as np
import pandas as pd
import os
import string
import itertools
from utils.profiler import ExponentialRange
src_dir = os.path.dirname(os.path.abspath(__file__))
data_dir = os.path.join(src_dir, '..', 'data')
target_dir = os.path.join(data_dir, 'big_numeric_csv_files')
# Max size in rows as a power of ten
exp_range = ExponentialRange(0, 7, 1 / 4)
num_cols = 10
col_names = list(string.ascii_uppercase[:num_cols])
_data = np.random.random((exp_range.max, num_cols))
data = pd.DataFrame(_data, columns=col_names)
_letters = string.ascii_uppercase
_file_codes = itertools.product(_letters, repeat=2)
_file_codes = list(_file_codes)[:exp_range.max]
file_codes = [''.join(code) for code in _file_codes]
for j, i in enumerate(exp_range.iterator()):
code = file_codes[j]
filename = f'file_{code}_rows_{i}.csv'
filepath = os.path.join(target_dir, filename)