I have come across an issue converting a JSON string in R (with either the
package 'rjson' or 'RJSONIO') when big integers are included in
the JSON
string.
A simple example of the problem is:
> options(scipen=999) # To prevent representing the answer in scientific
notation> json.str <- '{"bigInt":123456789123456789}' # Sample
JSON string with
big integer
Using the RJSONIO package:> fromJSON(json.str) # From package 'RJSONIO'
bigInt
123456789123456784
Using the rjson package:> fromJSON(json.str) # From package 'rjson'
$bigInt
[1] 123456789123456784
Notice the difference in the last digit (4 vs. 9).
In my research, I have come across various ways to handle big integers in R
(the 'int64' package, for example), but I don't believe this helps
with the
numeric representation/interpretation that occurs inside of the rjson and
RJSONIO packages.
It is not important in my application for numeric nodes in the JSON to be
preserved as a numeric class - if everything came back as a string, that
would be acceptable. Therefore, I was hoping the 'handler' or
'stringFun'
arguments within the RJSONIO package would allow me to specify a character
representation for all nodes inside the JSON string, such
that:> fromJSON(json.str)
would result in:
$bigInt
[1] "123456789123456789"
However, it looks like the 'stringFun' is only applied to the character
string nodes within the JSON:
>
fromJSON('{"bigInt1":123456789123456789,"bigInt2":"123456789123456789"}',stringFun=function(x)
class(x))
$bigInt1
[1] 123456789123456784
$bigInt2
[1] "character"
I think there might be some promise to the 'handler' object, but I'm
having
trouble interpreting how to implement it effectively.
Thank you in advance for any solutions/suggestions/references that you may
be able to provide!
Regards,
Trannon
______________________________
Trannon Mosher
Data Modeling Analyst
Clear Capital
[[alternative HTML version deleted]]
