Most efficient way to check if object is a String

Question

I get an object ( msg.data ) that is either a number or a string. If it is a string I don't want to use it, if it is a number I want to. I currently solved this problem with an ostringstream , though I think there are way better solutions:

void scan_cb(const sensor_msgs::LaserScan::ConstPtr& scan)
{
float dist = 0.0f;

std::ostringstream s;
s << scan->ranges[0]; // can be string (always "inf") or a float
if(s.str() != "inf"){
    dist += scan->ranges[0];
}
...

I care about efficiency, because this is part of a for loop running many times each second.

The basic structure is a ROS message, coming from a certain topic, and can have basically any data type. In this case I use a LaserScan message , the documentation does not mention that range[x] can return "inf" .

Answer 1

According to ROS documentation, scan->ranges is an array of float. That makes sense, because you add it to a float ( dist ) when the string representation of ranges[0] is not inf .

That means that (as MSalters guessed in its comment), you have a true float value, and you just want to make sure it is a real number and neither an infinite value nor a NaN (Not a Number) value.

So provided you include cmath (or math.h ) you can use the C classifications macros to determine whether the number is finite (but subnormal values are allowed) or normal (even subnormal values are rejected):

void scan_cb(const sensor_msgs::LaserScan::ConstPtr& scan)
{
float dist = 0.0f;

if(isfinite(scan->ranges[0])){
    dist += scan->ranges[0];
}
...

(more references on IEEE-754 representation of floating point numbers on wikipedia )