Representing parsed numbers

[D. Ogranos]

If you read "numbers" from a text source, how would you represent them?

I'm writing a little parser, and know I can find "number" tokens: for example "4711", "12.75" or "1.5E23". So there can be integer values and float values. Would you always represent the parsed value as a float (or double), and leave it to the user to convert the number to integer when needed? Or would you provide something like "floatValue" and "intValue" and perhaps a method isInteger() so a user can decide which value he wants? Or is there a better way to represent numbers that I haven't thought of?

Just looking for ideas

[vicky chauhan]

I dont know if I am right:

But you can write an IF ELSE block which checks for presence of '.' in the read string and accordingly parses it into an integer or float..


am i correct ??

[D. Ogranos]

Sure you can, but the parsing of the number is not the problem. My question is more like, WOULD you make two cases of it, or simply use a double to save the parsed value regardless if it is an int or float?

I think the best way might be to create a datatype "Number" which holds just a double value, but also offers methods isInteger() and intValue() in case the number found was not a float.

[Campbell Ritchie]

How are you writing the parser? Are you using regular expressions? Google for JFlex or a similar lexer, find its documentation and look through the handbook. You will probably find a simple example for numbers there (JFlex has one), or there might be something useful in the Java™ Language Specification (try §3.10).

[Pedantic mode]You are not parsing yet, only scanning (or lexing)[/pedantic mode]

[D. Ogranos]

There is no problem with scanning numbers, only with how you store them...how you represent the scanned value in a token. Do you simply always store it as a double, or do you have two value fields in a number token, one for int, one for double? And a flag indicating isInteger() or the like?

I had a similar case before, and then made a class with two value fields, and a flag. Also with two different methods to access the value of the number. It just seemed awkward, so I was looking for a better solution.

Something I'm just now thinking of, perhaps this would be better:

double value;   // only one value field

public Number getValue() {
  return isInteger() ? Integer.valueOf((int) value) : Double.valueOf(value);
}

Although the user of the token still has to check what he gets when calling getValue(), so perhaps the whole topic is just moot..

[Jesper de Jong]

How about returning a java.math.BigDecimal? It has methods such as intValue(), longValue(), doubleValue() etc. to get the value as an int, long, double etc.