Diary does not print backspace ('\b') properly - alternative/solution?

I have a bunch of scripts + functions that output their progress to the command window. At the same time I would like to write the same information to some kind of log file. diary normally does the job well in such a situation. However, I am printing a number of counters that make use of backspace characters. When I check my diary log afterwards, it turns out that every single counter output, as well as some weird characters are printed to my log file...

Is there any way to fix this behavior? And if not, is there any alternative? Using fprintf to print a file is not really an option, as I also want to print to the command window. Additionally, I would have to pass FID along many functions.

Answer 1

[Walter Roberson]

fprintf() with no fid writes to the command window, as does fprintf(1). fprintf(2) also writes to the command window but might be formatted specially as it is reserved for error messages.

Could you say more about the "weird characters" ? Have you checked to see which ASCII or Unicode positions they occupy?

[Niels]

Right now I'm using fprintf without fid to print to the command window. Diary is used along with fprintf to also save this output to a text file.

Difference occurs when I print e.g. '01\b\b02\b\b03'. The command window performs these operations nicely, but my logfile contains 01[BS][BS]02[BS][BS]03 where I guess [BS] (Notepad++ prints it that way) represents the ascii character for backspace.

As a result, a long counter (e.g. 000, 001, ..., 100) is printed as a line containing every single of these numbers as well as the backspace characters.

[Walter Roberson]

Log files show everything written to them. If you emit a backspace to the log then it is going to record it. Log files don't know what is important. If you had a "cls" would you want the log file to go back and erase everything it had logged so far, since that is what the screen would look like?

You have 3 options:

1. Don't count to output like that, switch to waitbar()
2. Create your own log file and don't write the intermediate counter values to it; or
3. post-process the log file.

For example if you were using OS-X or Linux, you could just run the output through col -b

In any editor that permits search and replace with meta characters, you can tell it to replace "anything" followed by backspace, with nothingness, and do that repeatedly. For example in age old vi this would be:

1,$s/.^H//g

where the ^H would be entered by pressing control-V and the backspace. Though you might have to repeat this command up to the maximum number of consecutive backspaces you have.

[Niels]

You've got a very valid point there. I think a combination of option 1 and 2 will be probably be the best solution for me. I was hoping for a less time consuming option, but I guess I'll simply have to go for a different approach if I don't want to pollute my log file with intermediate counter values :)

Answer 2

[Cedric]

I suspect that it does save the backslash character correctly, but that the editor that you are using to open the diary doesn't interpret it correctly or that you don't understand the code that the editor displays.

\b is a special character ( ASCII code 8 ); it is not meant to be displayed (same as others like \t\r\n). Instead, it tells whatever processes/displays the content to do something (e.g. go to a new line for \n, etc). If you open the diary file with Notepad++, for example, it will display a special BS symbol to indicate that there is a \b. If you open it with MATLAB editor, MATLAB displays a square that is not very explicit.

You can experiment with this. Open a diary in a new folder (so you have a fresh diary), print something with a backspace, and close the diary:

 >> diary on
 >> fprintf( 'AB\bC\n' ) ;
 AC
 >> diary off

Here you see that we inserted a \b after the B, so it is not display, and we get AC and a new line as output.

Now we can load the diary as characters, and DISP it:

 >> content = fileread( 'diary' ) ;
 >> content
 content =

 fprintf( 'AB\bC\n' ) ;

 AC

 diary off

So is it just the text that you saw, or does it contain the special characters (interpreted/displayed correctly here)? If you opened the diary in an editor, you know that it contains the special characters, but let's observe them. One way to display ASCII codes is to typecast to double by adding 0 to the char variable content:

 >> content + 0
 ans =

 Columns 1 through 19
   102   112   114   105   110   116   102    40    32    39    65    66    92    98    67    92   110    39    32

 Columns 20 through 38
    41    32    59    13    10    65    66     8    67    13    10   100   105    97   114   121    32   111   102

 Columns 39 through 41
   102    13    10

If you look at an ASCII table, you see that characters 'A','B','C' correspond to ASCII codes 65, 66, 67, that you see in columns 25, 26, and ... 28, because the 27th ASCII code is the ASCII code of the backspace: 8.

Why are you using a backslash by the way? This is not something that is done very often nowadays. Did you do something like:

 fprintf( '000') ; for k = 1 : 10, fprintf( '\b\b\b%03d', k ) ;  pause(1) ; end,  fprintf('\n') ;

Depending what you want to achieve, you can load the diary and process it's content to account for backspaces, but I predict that it is not going to be a very clean and stable approach.

[Niels]

Thanks for your response. I just commented Walter's response before seeing yours, but it's exactly as you stated.

Indeed, notepad++ shows a BS sign; loading it back into matlab makes it function as it should. But text files / dat files both do not like the backspace character.

You're also right on how I use it. I intend to use it as a simple counter, but the log file should only show that the counter has completed successfully, which it does not do correctly at this moment.

Answer 3

[Alejandro Estay]

the proper behavior is to make text cursor go back one space, without writing nothing. Matlab behavior ins not correct because it writes the symbol, but not perfom any action.