Reading from a JSON still isn’t as fas as it could be. My json_read mex function is over 10x faster than fromJSON(fileread('filename.json')).

Is there anything that would work in C# parsing JSON to MatLab native types? (MWCharArray, MWNumericalArray etc)

@oro77
The number of digits used depends on the JSon library and how it does the formatting. Sun/Oracle/openJDK provide a full spec for the java.lang.Double.toString() method. Remembering that MATLAB calls that method to get the text to display in the MATLAB Command Window for Java objects including Double, the following may be informative:

>> a = java.lang.Double('0.1')
a =
0.1
 
>> b = java.lang.Double('0.10000000000000001')
b =
0.1 
 
>> a==b
ans =
     1

>> java.lang.Double(0.1)==java.lang.Double('0.1')
ans =
     1

@Yair Altman – Thank you for your advice. I was thinking about rounding off it but the user may want to edit the json file in a text editor and he may find awkward about the insignificant digits. It is hard to decide what is the best solution. Is there something wrong about storing string representations in the JSON payload ?

@oro77 – I would think that it would make more sense to simply round off the insignificant digits upon loading/displaying (in those rare cases where this difference in LSB makes any difference), rather than store string representations in the JSON payload.

Thank you for your reply.
So the best way to keep 0.1 would be to store in my structure a string instead of a double value.

For a long answer:
What Every Computer Scientist Should Know About Floating-Point Arithmetic, by David Goldberg, published in the March, 1991 issue of Computing Surveys. Copyright 1991, Association for Computing Machinery, Inc.

Short answer:
Just as 1/3 can not be exactly represented in decimal, 1/10 has no exact representation in binary.

The java.lang.Double class is probably underused by MATLABers: its toString() method returns a string representation of a double that uses the minimum number of digits required to exactly reproduce that double during deserialisation (i.e to get back to that member of the subset of numbers represented in IEEE double).

java.lang.Double uses the FloatingDecimal class

jsonString = matlab.internal.webservices.toJSON(s);

It works but if my structure is using a real like below.

s = struct('this', {'a', 'b'}, 'that', {0.1, 2})

jsonString will return

[{"this":"a","that":0.10000000000000001},{"this":"b","that":2}]

Why is it 0.10000000000000001 instead of 0.1 ?
How can I correct this ?

ET, I tried your code without success in R2015a :

 
>> matlab.internal.webservices.toJSON('abc')
Undefined variable "matlab" or class "matlab.internal.webservices.toJSON".