Re: [cinjug-users] Re: More results from cluster run on V matrix.

[James Hurt <jhurt@xxxxxxxxxxxxxxxx>]

James Hurt wrote:

> The description for StreamTokenizer includes:
...
> Translation: Numbers in exponential format (e.g. 2.9e-17) are not
> correctly parsed.  The exponential component is missed.
>
> What is the easiest way to modify StreamTokenizer so it will recognize 
> numbers in scientific notation?

Solution is outlined here:

            // Create the StreamTokenizer
            f = new FileReader(fileName);
            st = new StreamTokenizer(f);

            // Modify it to turn off recognition of numbers
            st.ordinaryChars('0', '9');
            st.ordinaryChars('-', '-');
            st.ordinaryChars('+', '+');

            // Modify again so these chars are recognized as word
            st.wordChars('0','9');
            st.wordChars('-','-');
            st.wordChars('+','+');

            // Inside the nextToken loop
            int tokenType = st.nextToken();
            if (tokenType == StreamTokenizer.TT_WORD) {
                // Is this a number or not?
                double nval = 0.0;
                try {
                    nval = Double.parseDouble(st.sval);
                    // Is a number, treat as such
                    ...
                } catch (NumberFormatException e) {
                    // Not a number - treat as a word
                    ...
                }
            } else {
                // Neither a number nor a word
                ...
            }

Comments:

1.  The method parseNumber() is a do-nothing method.  This is the 
default and there is no easy way (see above code) to turn this option off.

2.  The parsing of numbers in StreamTokenizer is defective according to 
the Java Language specification and needs to be repaired.  How do I go 
about seeing that this repair gets performed?
--
James Hurt
Vice President Research & Development
Internet & Computer Institute
Phone: 513-271-4249