Re: Prize and Milestone as mutable state idioms

There are at least 2 errors in the code I posted yesterday. If you
can't find them, the updated code is at:

http://waterken.svn.sourceforge.net/viewvc/waterken/server/trunk/waterken/remote/src/org/waterken/syntax/json/ValueWriter.java?view=markup

--Tyler

On Mon, Jun 30, 2008 at 4:20 PM, Tyler Close <[email protected]> wrote:
> Hi all,
>
> I was recently refactoring the JSON serialization code in the Waterken
> Server to make auditing its correctness easier. In doing so, I wrote
> two small utility classes, each one to enforce the update pattern the
> code was using for one of its variables. For example, I was using a
> boolean variable to keep track of whether or not the most recent JSON
> value had been fully and successfully output. The update pattern here
> was that the variable started out false and then transitioned to
> forever being true. Verifying that this property was correctly
> implemented previously required examining the full file for mentions
> of the boolean variable to check that it was only ever assigned the
> value true. The Milestone class encapsulates a boolean variable that
> can only be updated in this way. There's a similar story for the Prize
> class. I'm thinking both of these classes may be generally useful and
> help communicate the programmer's intention for a mutable variable. If
> they are, I think I'll upgrade them to being reusable parts of the
> ref_send library. So, I'm looking for answers to a few questions:
>
> 1. Do the Prize and Milestone classes make the JSON serialization code
> more reviewable?
> 2. Do you suspect these classes would make other algorithms more reviewable?
> 3. Is the ref_send library the right place to host a collection of
> these utility classes for capability programming idioms?
>
> See the attached Joe-E source file for the current JSON serialization code.
>
> --Tyler
>