Some strings returned as "string" class instead of "char" in recent matlab versions

[jobh]

Arrays of strings are parsed by matlab's eval, and since they use double quotes they end up as string instead of char starting from 2017a.
A simple fix is to convert them back at the end, something like the following in loadjson.m (after parsing):

% The eval in parse_array creates strings, not char arrays, in R2017a+. Convert
% back (in try block, because the method probably doesn't exist in earlier versions)
try
    data = controllib.internal.util.hString2Char(data);
catch
end

[fangq]

hi @jobh, thanks for the comment. can you give me a short example showing this issue? I am also curious if adverse-effect if we leave it as "string"?

[pch-envision]

Since char and string have different behaviors in Matlab, you could encounter adverse effects in a downstream program which uses the loaded JSON object. A simple example would be a function which checks the input type using ischar - this would return zero if the variable in question is a string.

Since JSON does not permit single quote strings, the only way to resolve string vs. char would be to encode the type with the variable. Note that the neither built-in jsondecode nor mps.json.decode differentiate between char/string; everything is converted to char on deserialization.

[fangq]

right now, I don't see a problem storing strings as char array in JSON since there is no difference in JSON's specification. I do see a problem when one tries to store a string array, jsonlab can raise an error, but I think it should be fixed now after this commit by treating string arrays as a cell.

ed2645e

[fangq]

fixed. A user now can use ParseStringArray=1 or 0 to choose parsing string array or convert to char array. default is covert to char array (ParseStringArray=0). Try

jdata=loadjson('{"a": ["a1", "a10", 2, "a100"]}')
jdata = 

  struct with fields:

    a: {'a1'  'a10'  [2]  'a100'}

and

jdata=loadjson('{"a": ["a1", "a10", 2, "a100"]}','ParseStringArray',1)

jdata = 

  struct with fields:

    a: {["a1"]  ["a10"]  [2]  ["a100"]}

an alternative is to disable the fastarrayparser using

jdata=loadjson('{"a": ["a1", "a10", 2, "a100"]}','FastArrayParser',0)