ProgressBars.jl (formerly Tqdm.jl)

A fast, extensible progress bar for Julia. This is a Julia clone of the great Python package tqdm.

Installation

Run the following in a julia prompt:

using Pkg
Pkg.add("ProgressBars")

Usage

julia> using ProgressBars

julia> for i in ProgressBar(1:100000) #wrap any iterator
          #code
       end
100.00%┣█████████████████████████████████████████████████┫ 100000/100000 [00:12<00:00 , 8616.43 it/s]

There is a tqdm alias, so that people coming from python will feel right at home :)

julia> using ProgressBars

julia> for i in tqdm(1:100000) #wrap any iterator
          #code
       end
100.00%┣█████████████████████████████████████████████████┫ 100000/100000 [00:12<00:00 , 8616.43 it/s]

Iterating with a custom unit (e.g. bytes)

julia> iter = ProgressBar(1:2_000_000, unit="B", unit_scale=true)
       for i in iter
           # Do stuff
       end
100.0%┣██████████████████████████████████████████████████████████┫ 2.0MB/2.0MB [00:00<00:00, 4.8MB/s]

Or with a set description (e.g. for loss values when training neural networks)

julia> iter = ProgressBar(1:100)
       for i in iter
          # ... Neural Network Training Code
          loss = exp(-i)
          set_description(iter, string(@sprintf("Loss: %.2f", loss)))
       end
Loss: 0.02 3.00%┣█▌                                                  ┫ 3/100 00:00<00:02, 64.27 it/s]

Printing persistent messages while using a ProgressBar:

julia> iter = ProgressBar(1:5)
       for i in iter
         println(iter, "Printing from iteration $i")
         sleep(0.2)
       end
Printing from iteration 1
Printing from iteration 2
Printing from iteration 3
Printing from iteration 4
Printing from iteration 5
100.0%┣█████████████████████████████████████████████████████████████████┫ 5/5 [00:03<00:00, 1.5 it/s]

Postfixes are also possible, if that's your kind of thing:

julia> iter = ProgressBar(1:100)
       for i in iter
          # ... Neural Network Training Code
          loss = exp(-i)
          set_postfix(iter, Loss=@sprintf("%.2f", loss))
       end
100.0%┣█████████████████████████████████████████████┫ 1000/1000 [00:02<00:00, 420.4 it/s, Loss: 0.37]

You can also use multi-line postfixes, like so:

julia> iter = ProgressBar(1:100)
       for i in iter
          # ... Neural Network Training Code
          loss = exp(-i)
          set_multiline_postfix(iter, "Test 1: $(rand())\nTest 2: $(rand())\nTest 3: $loss)")
       end
100.0%┣█████████████████████████████████████████████████████████┫ 1000/1000 [00:02<00:00, 420.4 it/s]
Test1: 0.6740503146383823
Test2: 0.23694728303439727
Test3: 0.06787944117144233

No wrapped iterable

Sometimes you just want to display a progress bar without wrapping an iterable. This is possible through the combination of the ProgressBar(total=<N>) constructor and the update method:

julia> pbar = ProgressBar(total=100)
       update(pbar)     # Advances pbar count by 1
       update(pbar, 99) # Can give an explicit count for advancing the progress bar

Parallel for-loops

Now with added support for Threads.@threads for:

julia> a = []
       Threads.@threads for i in ProgressBar(1:1000)
         push!(a, i * 2)
       end
100.00%┣██████████████████████████████████████████████████████┫ 1000/1000 00:00<00:00, 28753.50 it/s]

Printing Delay

By default, the progress bar will update at most every 50ms in order to prevent the string IO from slowing down very fast iterations. This can be adjusted by passing the desired printing delay (in seconds) to the printing_delay-parameter when constructing the progress bar:

julia> for i in ProgressBar(1:1000, printing_delay=0.001)
         # do stuff
       end
100.0%┣████████████████████████████████████████████████████████┫ 1000/1000 [00:00<00:00, 3006.8 it/s]

cloud-oak/ProgressBars.jl

cloud-oak

Reviews