Could we make this more clear or simplified? If we accept any representations, the situation may be a little bit complicated: 1. "1.5e4F" -> Some characters like ".eF" must be grabbed when tokenizing a real number. 2. "- 4" -> A space-separated minus may not be an operator, but the sign of a real number. 3. ".123" -> Not all real numbers start with or end with digits. Although APIs like atof or strtof can check and parse potential real number tokens very well, knowing characters in real number token is necessary when splitting words. Another solution may be simply the syntax of real number, that will make life easier. For example, all real numbers will be like '[+-]([0-9]*)[.]([0-9]*)'.