Using sprintf to match the results of format

If I read the help for format, it tells me:

format may be used to switch between different output display formats
of all float variables as follows:
  format SHORT     Scaled fixed point format with 5 digits.
  format LONG      Scaled fixed point format with 15 digits for double
                   and 7 digits for single.
  format SHORTE    Floating point format with 5 digits.
  format LONGE     Floating point format with 15 digits for double and
                   7 digits for single.
  format SHORTG    Best of fixed or floating point format with 5 
                   digits.
  format LONGG     Best of fixed or floating point format with 15 
                   digits for double and 7 digits for single.
  format SHORTENG  Engineering format that has at least 5 digits
                   and a power that is a multiple of three
  format LONGENG   Engineering format that has exactly 16 significant
                   digits and a power that is a multiple of three.

I'd like to replicate the behavior of each of these formats, using an appropriate format spec in sprintf. The purpose is for a custom numeric class I was thinking of writing. I'd happily ignore longeng and shorteng, if they were problematic. While I can probably play around with format specs and get something reasonable, someone else is surely much more knowledgable in this matter than me.

Answer 1

[Steven Lord]

The easiest way to replicate the format specification is to just let MATLAB do it.

oldFormat = get(0, 'Format');

% may want to set up an onCleanup object here to restore the format
% to guard against one of the next few commands throwing an error
% But for demonstration purposes, I assume everything will work fine

% 'replicate' the short format
format short
dataToDisplay = pi;
s = evalc('disp(dataToDisplay)');

% Everything worked fine up to this point, so reset the format
format(oldFormat)

If you're displaying an array, you may also want to get(0, 'FormatSpacing') so you can handle 'loose' and 'compact'.

There are some tools in MATLAB for customizing object display, though I believe those are targeted for helping display groups of properties of an object.

[Ulises Nunez Garzon]

This is a great answer. To top it off, I would wrap your code to get the desired result in the following way:

function c_str = c_format_short(a)
oldFormat = get(0, 'Format');
dataToDisplay = a;
c_str = evalc('disp(dataToDisplay)');
c_str = strtrim(c_str(2:end-1));
format(oldFormat)
end

Answer 2

[dpb]

As a starting point

>> fnprnt=@(x) fprintf('%.5g\n',x)
fnprnt = 
  @(x)fprintf('%.5g\n',x)
>> fnprnt(pi)
3.1416
>> fnprnt(pi*1E6)
3.1416e+06
>> format short
>> pi
ans =
  3.1416
>> pi*1E6
ans =
 3.1416e+06
>>

[John D'Errico]

Reasonably close to format short, with some minor differences. Ones that I'd probably be willing to tolerate.

format short
x = 2255.334
x =
   2.2553e+03

fprintf('%.5g\n',x)
2255.3

x = 225533.4
x =
   2.2553e+05

fprintf('%.5g\n',x)
2.2553e+05

[dpb]

Yeah, I don't think there's any single format string that reproduces any of the FORMAT choices exactly excepting perhaps for 'bank' as a '|.2f'|. But I'm not certain there on the rounding...does Matlab use "bankers' rounding"? Don't think it's documented.

[Walter Roberson]

There is no one format that can replicate the format short output. %13.5g helps (the 13 part deals with the spacing), but format short treats integral values differently than non-integral value, always displaying 0 after the decimal point for non-integral values that %g would not display. 1+eps for example is displayed as 1.0000 but 1 exactly is displayed with no decimal places. %g would display 1+eps with no trailing 0, and the %e and %f formats would always display trailing 0 even for integral values.

Answer 3

[Stephen23]

Introduced in R2021a: https://www.mathworks.com/help/matlab/ref/formatteddisplaytext.html

format short
str = strtrim(formattedDisplayText(pi))

str = "3.1416"

format long E
str = strtrim(formattedDisplayText(pi))

str = "3.141592653589793e+00"

[Captain Karnage]

This is a welcome addition. However, in your example, you're still setting the system format first, then calling formattedDisplayText. The beauty of formattedDisplayText is you can, as a parameter, specify the exact format you want using the 'NumericFormat' name/value pair without calling the format function e.g.

strtrim( formattedDisplayText( pi, 'NumericFormat', 'longEng' ) )

ans = "3.14159265358979e+000"