Converting string to a double

Hello everyone.

i'm programming a software that gets different values from a txt file, and plotting a graph based on these values.

In order to create the graph I had to convert the strings in the txt file to a double, but the str2double function rounds the values so the graph isn't accurate.

for example, the string to convert is: 1.7356922617E-3, and the doublue value after applying the method is 1.73569E-3.

is there any method that convert without rounding?

thank you.

[Stephen23]

" but the str2double function rounds the values so the graph isn't accurate."

I very much doubt that str2double "rounds the values", what is much more likely is that you are confusing how numeric data are displayed with how numeric data are stored, which are two completely different things.

You can easily change how the data are displayed:

format short
str = '1.7356922617E-3';
val = str2double(str)

val = 0.0017

format long
val

val = 
   0.001735692261700

So str2double has not "rounded" the value at all, all of the digits are converted and stored as expected.

Answer 1

[Fangjun Jiang]

Are you sure?

>> format long
>> a=str2double('1.7356922617E-3')

a =

   0.001735692261700

[Jan]

Exactly. The values are not rounded, but just the display in the command window.

[Fangjun Jiang]

Note, there is a limit on the precision. I believe it is 15 digits of precision. Need to find it somewhere in the doc.

>> b=str2double('100.12345678901234567')

b =

     1.001234567890124e+02

[Stephen23]

"I believe it is 15 digits of precision"

The "precision" is more complex than that, it really depends what you are doing with the values:

https://www.exploringbinary.com/decimal-precision-of-binary-floating-point-numbers/

https://www.exploringbinary.com/number-of-digits-required-for-round-trip-conversions/

[Fangjun Jiang]

In MATLAB, it seems to be 15 digits after the decimal point.

format long
c=str2double('10.12345678901234567890')
c=str2double('1.12345678901234567890')
c=str2double('0.12345678901234567890')
c =
  10.123456789012346
c =
   1.123456789012346
c =
   0.123456789012346

[Stephen23]

"In MATLAB, it seems to be 15 digits after the decimal point."

Your examples only demonstrate the display format you selected (which has 15 digits after the decimal point)

Your examples do not demonstrate any "limit" on the conversion precision.

[Walter Roberson]

c=str2double('10.12345678901234567890')

c = 10.1235

fprintf('%.999g\n', c)

10.1234567890123461353368838899768888950347900390625

This has apparently remembered 16 total digits correctly, 10.12345678901234. Is that really the case?

fprintf('%.999g\n', 10.12345678901234567890)  %do we get the same?

10.1234567890123461353368838899768888950347900390625

fprintf('%.999g\n', 10.123456789012340)

10.123456789012340806266365689225494861602783203125

fprintf('%.999g\n', 10.123456789012344)

10.12345678901234435898004448972642421722412109375

fprintf('%.999g\n', 10.123456789012345)

10.12345678901234435898004448972642421722412109375

fprintf('%.999g\n', 10.123456789012346)

10.1234567890123461353368838899768888950347900390625

fprintf('%.999g\n', 10.123456789012350)

10.1234567890123496880505626904778182506561279296875

fprintf('%.999g\n', 10.12345678901234567890*(1-eps))

10.12345678901234435898004448972642421722412109375

fprintf('%.999g\n', 10.12345678901234567890*(1+eps))

10.123456789012347911693723290227353572845458984375

So after 10.12345678901234 it is taking into account that what follows is something on the high side of 5x in order to decide whether to round down to 10.123456789012344 or up to 10.123456789012346

[Fangjun Jiang]

good points. Now I understand better. It certainly can read in more than 15 digits after decimal point.

>> p1=sprintf('%40.30f',pi)

p1 =

' 3.141592653589793115997963468544'

>> d=str2double(p1)

d =

3.141592653589793

>> p2=sprintf('%40.30f',d)

p2 =

' 3.141592653589793115997963468544'