6,101 views (last 30 days)

How to make MATLAB output the full number in digits, and not using scientific notation?

Oleg Komarov
on 6 Aug 2011

format long

or

sprintf('%16.f',2332456943534324)

Walter Roberson
on 16 Jul 2019

James Upton
on 16 Jul 2019

Walter Roberson
on 21 Apr 2020

Image Analyst
on 7 Aug 2011

Edited: MathWorks Support Team
on 8 Nov 2018

To display the maximum number of digits in a variable without using scientific notation, set the output display format to "longG":

format longG

After you set the display format, variables display in decimal notation:

m = rand(1,3)/1000

m =

0.000546881519204984 0.000957506835434298 0.00096488853519927

To avoid displaying scientific notation for variables that exceed 2^50 use "sprintf". For example, this code displays the number 2332456943534324 in decimal notation:

sprintf('%16.f',2332456943534324)

ans =

'2332456943534324'

For more information, see the "format" documentation:

Walter Roberson
on 7 Aug 2011

format long g

helps. However, integers that exceed 2^53 will be represented in scientific notation with "format long g". To get the full digits of those, you need to use sprintf() or fprintf()

Image Analyst
on 7 Aug 2011

Yes it can help. Sometimes some sneak through even with that (if there would be more than three 0's to the right of the decimal point), like this which I tried:

m =

Columns 1 through 4

0.000538342435260057 0.000996134716626886 7.81755287531837e-005 0.000442678269775446

Columns 5 through 8

0.000106652770180584 0.000961898080855054 4.63422413406744e-006 0.000774910464711502

Walter Roberson
on 19 Feb 2018

For MS Windows and Linux, to get full number of digits and not in exponential form, you need to either use the Symbolic toolbox or you need to use a tool such as https://www.mathworks.com/matlabcentral/fileexchange/22239-num2strexact--exact-version-of-num2str- from the File Exchange. This is crucial for MS Windows, which does a rather poor job of converting exact values; Linux does a better job but still has inaccuracies after a while.

On Mac (OS-X, MacOS), the built in conversion is exact, and you can choose to sprintf() with a '%.1074f' format. For example,

>> sprintf('%.1074f', eps(realmin))

ans =

'0.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000004940656458412465441765687928682213723650598026143247644255856825006755072702087518652998363616359923797965646954457177309266567103559397963987747960107818781263007131903114045278458171678489821036887186360569987307230500063874091535649843873124733972731696151400317153853980741262385655911710266585566867681870395603106249319452715914924553293054565444011274801297099995419319894090804165633245247571478690147267801593552386115501348035264934720193790268107107491703332226844753335720832431936092382893458368060106011506169809753078342277318329247904982524730776375927247874656084778203734469699533647017972677717585125660551199131504891101451037862738167250955837389733598993664809941164205702637090279242767544565229087538682506419718265533447265625'

For larger values you might want to trim out trailing zeros from the converted string

val = pi*1E-200;

regexprep( sprintf('%.1074f', val), '0+$', '', 'lineanchors')

ans =

'0.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000003141592653589793111936498419027683964072757959391149845317813416927695644722162706379483043156554579881967829022575831926635177847590589777088086173081089243142930507159490615800591052996089483276727788901006686618108987452642387169053033459820326372299902201815389727889699071056417123601253516892437642498120285079407325647552658339885701180059456745257476645670329996938769926310811984167666114826593537757304481509915842491117931968666219637406979734598283259283758102504979792257699955371208488941192626953125'

Mark Bower
on 20 Oct 2017

Edited: Mark Bower
on 20 Oct 2017

A nice, consistent solution is to use "num2str()". The same call works for both display from the command line:

> val = 1234567890

val =

1.234567890000000e+09

> num2str(val)

ans =

1234567890

and also within print statements:

> sprintf(num2str(val))

ans =

1234567890

It also works for floating point numbers:

> val = 123456.789

val =

1.234567890000000e+05

> sprintf(num2str(val))

ans =

123456.789

>

Walter Roberson
on 19 Feb 2018

>> num2str(pi*10^5)

ans =

'314159.2654'

This is not "full decimal places"

Using num2str() inside sprintf() is redundant.

Stephen Cobeldick
on 20 Feb 2018

sprintf(num2str(val))

The sprintf is totally superfluous, it does nothing useful at all here, just slows down the code. In any case, using a proper sprintf format string would be quicker than calling num2str, and provide more control over the number of digits, so why not do that?

Huw S
on 31 Jan 2017

If you don't need to know all the decimal points, then do your equation inside round.

saves all the other bother of exponentials.

Walter Roberson
on 31 Jan 2017

Unfortunately not the case:

>> format short

>> round(2^54)

ans =

1.8014e+16

>> format long g

>> round(2^54)

ans =

1.8014398509482e+16

>> uint64(2^54)

ans =

uint64

18014398509481984

Kaveh Vejdani
on 19 Feb 2018

I don't understand why you have accepted the wrong answers. What you're looking for is: format short g

Cheers, Kaveh

Walter Roberson
on 19 Feb 2018

>> format short g

>> pi*10^5

ans =

3.1416e+05

That does not appear to satisfy either part of the requirement,

"how to make matlab output the full number in digits, and not in exponential form?"

Kaveh Vejdani
on 19 Feb 2018

Walter Roberson
on 19 Feb 2018

>> format short g

>> pi

ans =

3.1416

This is not "full number in digits"

>> 1000000

ans =

1e+06

this is not even close to being an "absurdly huge number"

format short g gives you at most 5 significant figures.

format long g gives you at most 15 significant figures. It turns out that is not enough in practice to be unique. There are 24 distinct representable values in unique(pi-37*eps:eps:pi+9*eps), all of which display as 3.14159265358979 under format long g. If the goal is to output enough digits to be able to transfer the values exactly in text form, then format long g is not sufficient.

People get caught by this all the time!

format long g

T = 0.3 - 0.2

T == 0.1

T - 0.1

T =

0.1

ans =

logical

0

ans =

-2.77555756156289e-17

People have difficulty understanding why a value that shows up as 0.1 does not compare as equal to 0.1: the limits of format long g have real effects.

Christos Boutsikas
on 21 Apr 2020

You can also use Variable-precision arithmetic via command vpa.

vpa(x) %if x is the output number you are interesting in

Walter Roberson
on 21 Apr 2020

This is what Steven and I were referring to when we discussed Symbolic Toolbox.

Opportunities for recent engineering grads.

Apply TodayFind the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!
## 0 Comments

Sign in to comment.