Add support for parsing nan/inf values

[jschueller]

Similar to numpy or pandas Python libraries, I would like to add support for reading nan or inf values
Here is a snippet of how it would look like for c++ with inf values:

std::map<std::string, double> infMap;
infMap["inf"] = std::numeric_limits<double>::infinity();
infMap["+inf"] = std::numeric_limits<double>::infinity();
infMap["-inf"] = -std::numeric_limits<double>::infinity();
infMap["Inf"] = std::numeric_limits<double>::infinity();
infMap["+Inf"] = std::numeric_limits<double>::infinity();
infMap["-Inf"] = -std::numeric_limits<double>::infinity();
infMap["INF"] = std::numeric_limits<double>::infinity();
infMap["+INF"] = std::numeric_limits<double>::infinity();
infMap["-INF"] = -std::numeric_limits<double>::infinity();
...
// handle inf values
if (infMap.count(pStr))
  pVal = infMap[pStr];
else
  // invalid values

what do you think ?

[d99kris]

Hi @jschueller - thanks for suggesting a feature. I am a little hesitant though, as I like to keep this library small and simple.

The approach taken by this library is to offload "special data conversions" to consumers of the library. Specifically it exposes APIs for "Custom Data Type Conversion", see examples here and here.

Would that technique work for your use case?

[jschueller]

fair enough, but I would have to get over the limitations mentioned in
#181

[d99kris]

I added some suggestion in #181 - let's see if that can work. Otherwise I may need more specifics about the data format to propose some way forward.