Converting strings to other types can change values

[yaccob]

Hi Douglas

The attempt to guess the type of a text node or an attribute value may change the type which may be unintended. If the intended type is a string, "01" is not equal with "1". So operations change values which can cause unwanted effects.
Furthermore the behavior is not consistent: "00" remains unchanged (remains a String) while "01" is converted to the number 1 as you can see in the sample below.
Also automatic conversion to boolean might have unwanted effects. Imagine the title of a book's chapter is "True" - intentionally with upper case "T". As far as I could see currently I am not able to convert this title to Json without implicitly changing it to a boolean and therefore losing information.

Sample:

xml:
<root>
  <id>01</id>
  <id>1</id>
  <id>00</id>
  <id>0</id>
  <title>True</title>
</root>

code:
println org.json.JSONML.toJSONArray(xml)

output:
["root",["id",1],["id",1],["id","00"],["id",0],["title",true]]

I would suggest to either change this or at least to add an option to control the behavior in a static way. (e.g. adding a non-final static boolean field KEEP_STRINGS which can be set to true)

Best regards,
yaccob

[douglascrockford]

You have the source.

[uklimaschewski]

@yaccob adding a non-final static boolean field KEEP_STRINGS which can be set to true

This will make toJSONArray() not thread-safe. Better add another boolean parameter to the parse() method and introduce new toXXX() methods with this parameter.

[yaccob]

@uklimaschewski introduce new toXXX() methods
Agreed, introducing new toXXX() methods is anyway much cleaner. I'll simply overload the parse method and add a new toJsonML(String) and a new toJsonML(XMLTokener) method using the overloaded parse method. Thanks for your input.