Re: SRFI-19 reference implementation problem

[Will Fitzgerald <wfitzg@xxxxxxxx>]

A new fixed version is up at:

http://max.cs.kzoo.edu/~wfitzg/srfi-19/

These are based on Martin Gasbichler's comments, which I address below.


Martin Gasbichler asks, and I answer:

> Some small remarks I came across while porting this to scsh:
>
> Why not using SRFI-9 for the definition of the record types? This
> would it also make possible omit the definition of mutators completely
> instead of redefining them to throw an error.

I tried to make as few assumptions about the implementation as  
possible, especially because some parts of the code are OS-specific.

> Why is the definition of COPY-TIME not simply:
>
> (define (copy-time time)
>    (make-time (time-type time)
>               (time-second time)
>               (time-nanosecond time)))

Sometimes, I'm an idiot. I fixed this.

> / with more than two arguments is an optional procedure in R5RS, same
> applies to -. IMHO using the binary variants makes the code much more
> readable...

Hmm, we could quibble on the readability, but for max. portability of  
this
less than portable code, I've changed calls to / and - to 1 or 2  
arguments. I think there was only one arity three call for each.

> Scsh's reader is case-sensitive, it therefore fails to understand
> #\Space. Also #\Tab or #\tab is not in R5RS, maybe you could use a
> string containing a tabulator character instead (for the scsh port I
> used scsh's ASCII package).

No #\tab character in R5RS! Argh! I changed this to (integer->char 9)  
and
hope for ascii or equivalents.

> The scsh version of your code with these things fixed is available
> from scsh's CVS at
>
> http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/*checkout*/scsh/scsh- 
> 0.6/scheme/srfi/srfi-19.scm?rev=HEAD&content-type=text/plain
>
>
> Before successfully running the test suite I had to fix two minor
> tweaks:
>
> In the test suite you use the non-standard |...| notation for
> symbols.

OK, I changed this in my code, too.

> Also in the test suite you use tm:nano which is obviously an internal
> constant of the library.

OK, I changed this in my code, too.


--
Will Fitzgerald