• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    迪恩网络公众号

JAAdrian/MatlabProgressBar: This MATLAB class provides a smart progress bar like ...

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称(OpenSource Name):

JAAdrian/MatlabProgressBar

开源软件地址(OpenSource Url):

https://github.com/JAAdrian/MatlabProgressBar

开源编程语言(OpenSource Language):

MATLAB 100.0%

开源软件介绍(OpenSource Introduction):

View MatlabProgressBar on File Exchange

MatlabProgressBar

This project hosts the source code to the original MATLAB FileExchange project and is place of active development.

A drawback in MATLAB's own waitbar() function is the lack of some functionalities and the loss of speed due to the rather laggy GUI updating process. Therefore, this MATLAB class aims to provide a smart progress bar in the command window and is optimized for progress information in simple iterations or large frameworks with full support of parallel parfor loops and asynchronous processing via parfeval() provided by the MATLAB Parallel Computing Toolbox.

A design target was to mimic the best features of the progress bar tqdm for Python. Thus, this project features a Unicode-based bar and some numeric information about the current progress and the average iterations per second.

Several projects exist on MATLAB's File Exchange but none incorporates the feature set shown below. That's why I decided to start this project.

Easy progress bar example

Supported features include (or are planned):

  • proper unit testing
  • display the bar name as a ticker. That way, a fixed bar width could be used
  • inherit from MATLAB System Object to gain benefits from the setup method
  • TQDM Unicode blocks
  • optional constructor switch for ASCII number signs (hashes)
    • those will be used if ProgressBar() is used in deploy mode (MATLAB Compiler)
  • optional bar title
  • optional visual update interval in Hz [defaults to 10 Hz]
  • when no total number of iterations is passed the bar shows the elapsed time, the number of (elapsed) iterations and iterations/s
  • nested bars (at the moment only one nested bar is supported [one parent, one child])
  • printMessage() method for debug printing (or the like)
  • print an info when a run was not successful
  • support another meaningful 'total of something' measure where the number of items is less meaningful (for example non-uniform processing time) such as total file size (processing multiple files with different file size). At the moment, the only alternative supported unit is Bytes
  • when the internal updating process is faster than the actual updates via update(), the internal counter and printing of the process bar stops until the next update to save processing time
  • linear ETA estimate over all last iterations
  • support parfor loops provided by the Parallel Computing Toolbox
  • show s/it if it/sec < 1
  • override MATLAB's default non-UTF font if OverrideDefaultFont is true. This will switch the font for the command line to Courier New for the lifetime of the bar. Default is false
  • disable the progress bar if IsActive is false. This will disable the functionality completely and can be used in situations in which the bar is not beneficial (e.g. if the bar is used in a sub-application of a processing cluster). Default is true.

Note:
Be sure to have a look at the Known Issues section for current known bugs and possible work-arounds.

Dependencies

No dependencies to toolboxes.

The code has been tested with MATLAB R2016a, R2016b and R2020b on Windows 10, Xubuntu 16.04.2 LTS and Linux Mint 20.

Installation

Put the files ProgressBar.m, progress.m and updateParallel.m into your MATLAB path or the directory of your MATLAB project.

Usage

Detailed information and examples about all features of ProgressBar are stated in the demo scripts in the ./demos/ directory.

Proposed Usage for Simple Loops

The simplest use in for-loops is to use the progress() function. It wraps the main ProgressBar class and is intended to only support the usual progress bar. Be aware that functionalities like printMessage(), printing success information or a step size different to 1 are not supported with progress.m. Also, this only works for non-parallel loops.

See the example below:

numIterations = 10e3;

% create the loop using the progress() class
for iIteration = progress(1:numIterations)
    % do some processing
end

Example 2

Extended Usage with all Features

The basic work flow is to instantiate a ProgressBar object and use either the step() method to update the progress state (MATLAB <= R2015b) or use the instantiated object directly as seen below. Refer to the method's help for information about input parameters. The shown call is the default call and sufficient. If you want to pass information about the step size, the iteration's success or if a new bar should be printed immediately (e.g. when iterations take long time) you can pass these information instead of empty matrices.

All settings are done using name-value pairs in the constructor. It is strongly encouraged to call the object's release() method after the loop is finished to clean up the internal state and avoid possibly unrobust behavior of following progress bars.

Usage
obj = ProgressBar(totalIterations, varargin)

A simple but quite common example looks like this:

numIterations = 10e3;

% instantiate an object with an optional title and an update
% rate of 5 Hz, i.e. 5 bar updates per seconds, to save
% printing load.
progBar = ProgressBar(...
    numIterations, ...
    'Title', 'Awesome Computation' ...
    );
    
% begin the actual loop and update the object's progress
% state
for iIteration = 1:numIterations
    %%% do some processing here
    % ...
    
    progBar([], [], []);
    % or in releases <= R2015b
    % progBar.step([], [], []);
end
% call the 'release()' method to clean up
progBar.release();

Extended usage with the object method calls

Parallel Toolbox Support

If you use MATLAB's Parallel Computing Toolbox, refer to the following example or the demo file k_parallelSetup.m. Tested parallel functionalities are parfor and parfeval() for asynchronous processing.

numIterations = 10e3;

% Instantiate the object with the 'IsParallel' switch set to true
progBar = ProgressBar(numIterations, ...
    'IsParallel', true, ...
    'Title', 'Parallel Processing' ...
    );
    
% ALWAYS CALL THE SETUP() METHOD FIRST!!!
progBar.setup([], [], []);
parfor iIteration = 1:numIterations
    pause(0.1);
    
    % USE THIS FUNCTION AND NOT THE STEP() METHOD OF THE OBJECT!!!
    updateParallel([], pwd);
end
progBar.release();

Known Issues

Flickering Bar or Flooding of the Command Window

MATLAB's speed to print to the command window is actually pretty low. If the update rate of the progress bar is high the mentioned effects can occur. Try to reduce the update rate from the default 5 Hz to something lower (say 3 Hz) with the 'UpdateRate', 3 name-value pair.

The Bar Gets Longer With Each Iteration

There seems to be a problem with the default font Monospaced in Windows. If this behavior is problematic, change the font for the command window to a different monospaced font, preferably with proper Unicode support.

If you do not want to or cannot change the font in the setting, you can set the class's OverrideDefaultFont to true while you are in the configuration phase. This will change MATLAB's coding font to Courier New for the duration in which the bar is alive (until the release() method is executed). After the object's lifetime, your previous font will be restored automatically.

For convenience, this property can also be set in the progress() wrapper to always trigger for the wrapper if desired.

Thanks @GenosseFlosse for the fix!

Strange Symbols in the Progress Bar

The display of the updating progress bar is highly dependent on the font you use in the command window. Be sure to use a proper font that can handle Unicode characters. Otherwise be sure to always use the 'Unicode', false switch in the constructor.

Remaining Timer Objects in MATLAB's Background

Sometimes, if the user cancels a loop in which a progress bar was used, the destructor is not called properly and the timer object remains in memory. This can lead to strange behavior of the next progress bar instantiated because it thinks it is nested. If you encounter strange behavior like wrong line breaks or disappearing progress bars after the bar has finished, just call the following static method to delete all remaining timer objects in memory which belong(ed) to progress bars and start over.

ProgressBar.deleteAllTimers();

Issues Concerning Parallel Processing

The work-flow when using the progress bar in a parallel setup is to instantiate the object with the IsParallel switch set to true and using the updateParallel() function to update the progress state instead of the step() method of the object. If this results in strange behavior check the following list. Generally, it is advisable to first be sure that the executed code or functions in the parallel setup run without errors or warnings. If not the execution may prevent the class destructor to properly clean up all files and timer objects.

  • are there remaining timer objects that haven't been deleted from canceled for/parfor loops or parfeval() calls when checking with timerfindall('Tag', 'ProgressBar')?
    • use delete(timerfindall('Tag', 'ProgressBar'))
  • does the progress exceed 100%?
    • try to call clear all or, specifically, clear updateParallel to clear the internal state (persistent variables) in the mentioned function. This should have been done by the class destructor but sometimes gets unrobust if there have been errors in parallel executed functions.
    • Also try to look into your temp directory (returned by tempdir()) if remaining progbarworker_* files exist. Delete those if necessary.

TL/DR:
clear all and delete(timerfindall('Tag', 'ProgressBar')) are your friend! Be sure that no files following the pattern progbarworker_* remain in the directory returned by tempdir().

Unit Tests

You can run all available tests in the project directory by navigating into the tests folder and executing runtests in MATLAB. However, if you want to omit the parallel tests (e.g. you don't have the Parallel Toolbox installed), just execute

runtests Tag NonParallel

License

The code is licensed under BSD 3-Clause as stated in the LICENSE file




鲜花

握手

雷人

路过

鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
热门推荐
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服(服务时间 9:00~18:00)

在线QQ客服
地址:深圳市南山区西丽大学城创智工业园
电邮:jeky_zhao#qq.com
移动电话:139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap