p_tqdm makes parallel processing with progress bars easy.

p_tqdm is a wrapper around pathos.multiprocessing and tqdm. Unlike Python's default multiprocessing library, pathos provides a more flexible parallel map which can apply almost any type of function, including lambda functions, nested functions, and class methods, and can easily handle functions with multiple arguments. tqdm is applied on top of pathos's parallel map and displays a progress bar including an estimated time to completion.

pip install p_tqdm

Let's say you want to add two lists element by element. Without any parallelism, this can be done easily with a Python map.

l1 = ['1', '2', '3']
l2 = ['a', 'b', 'c']

def add(a, b):
    return a + b
    
added = map(add, l1, l2)
# added == ['1a', '2b', '3c']

But if the lists are much larger or the computation is more intense, parallelism becomes a necessity. However, the syntax is often cumbersome. p_tqdm makes it easy and adds a progress bar too.

from p_tqdm import p_map

added = p_map(add, l1, l2)
# added == ['1a', '2b', '3c']

  0%|                                    | 0/3 [00:00<?, ?it/s]
 33%|████████████                        | 1/3 [00:01<00:02, 1.00s/it]
 66%|████████████████████████            | 2/3 [00:02<00:01, 1.00s/it]
100%|████████████████████████████████████| 3/3 [00:03<00:00, 1.00s/it]

• p_map - parallel ordered map
• p_imap - iterator for parallel ordered map
• p_umap - parallel unordered map
• p_uimap - iterator for parallel unordered map

• t_map - sequential ordered map
• t_imap - iterator for sequential ordered map

Performs an ordered map in parallel.

from p_tqdm import p_map

def add(a, b):
    return a + b

added = p_map(add, ['1', '2', '3'], ['a', 'b', 'c'])
# added = ['1a', '2b', '3c']

Returns an iterator for an ordered map in parallel.

from p_tqdm import p_imap

def add(a, b):
    return a + b

iterator = p_imap(add, ['1', '2', '3'], ['a', 'b', 'c'])

for result in iterator:
    print(result) # prints '1a', '2b', '3c'

Performs an unordered map in parallel.

from p_tqdm import p_umap

def add(a, b):
    return a + b

added = p_umap(add, ['1', '2', '3'], ['a', 'b', 'c'])
# added is an array with '1a', '2b', '3c' in any order

Returns an iterator for an unordered map in parallel.

from p_tqdm import p_uimap

def add(a, b):
    return a + b

iterator = p_uimap(add, ['1', '2', '3'], ['a', 'b', 'c'])

for result in iterator:
    print(result) # prints '1a', '2b', '3c' in any order

Performs an ordered map sequentially.

from p_tqdm import t_map

def add(a, b):
    return a + b

added = t_map(add, ['1', '2', '3'], ['a', 'b', 'c'])
# added == ['1a', '2b', '3c']

Returns an iterator for an ordered map to be performed sequentially.

from p_tqdm import p_imap

def add(a, b):
    return a + b

iterator = t_imap(add, ['1', '2', '3'], ['a', 'b', 'c'])

for result in iterator:
    print(result) # prints '1a', '2b', '3c'

All p_tqdm functions accept any number of iterables as input, as long as the number of iterables matches the number of arguments of the function.

To repeat a non-iterable argument along with the iterables, use Python's partial from the functools library. See the example below.

from functools import partial

l1 = ['1', '2', '3']
l2 = ['a', 'b', 'c']

def add(a, b, c=''):
    return a + b + c

added = p_map(partial(add, c='!'), l1, l2)
# added == ['1a!', '2b!', '3c!']

All the parallel p_tqdm functions can be passed the keyword num_cpus to indicate how many CPUs to use. The default is all CPUs. num_cpus can either be an integer to indicate the exact number of CPUs to use or a float to indicate the proportion of CPUs to use.

Note that the parallel Pool objects used by p_tqdm are automatically closed when the map finishes processing.

All the parallel p_tqdm functions can be passed the keyword tqdm to choose a specific flavor of tqdm. By default, this value is taken from tqdm.auto. The tqdm parameter can be used pass p_tqdm output to tqdm.gui, tqdm.tk or any customized subclass of tqdm.