Re: duplication of SRFIs

[Sebastian Egner <sebastian.egner@xxxxxxxxxxx>]

Alex Shinn wrote:
> It claims to be lightweight yet
adds a
> CHECK-EC form, which is unnecessary,
pulls in a requirement for
> SRFI-42, and doesn't seem to offer
much more than running tests
> inside an existing comprehension
available once you are using
> SRFI-42.

Did I miss something in SRFI 42? Where
does it define CHECK-EC?

I can tell you from experience: CHECK-EC
is worth every penny.

It also does not pull in a dependency
on SRFI 42, because CHECK-EC
can be implemented as a hygienic macro
which is independent of
anything SRFI 42 defines. In other words,
if you do not intend to use
CHECK-EC you do not have to load SRFI
42 at all. If you intend to use
CHECK-EC you need to load SRFI 42 or
reimplement the parts you need.

> In exchange it removes any handling
of exceptions, which
> seems to be rather crippling for
a test SRFI.

This is indeed the biggest limitation
of this SRFI: You cannot run checks
that are expected to raise an error
conditions, e.g. (check (/ 1 0) => division-by-zero)

However, I decided that the fact R5RS
leaves error handling undefined
means that it is not the right time
to impose some particular representation
of error conditions on Scheme just for
the purpose of this simple SRFI.

In practice, it is not too much of a
problem, exactly because portable
working programs do not rely on error
conditions---and if they use
exceptions explicitly (e.g. through
SRFI 34) so can the test cases:
(check (with-exception-handler ... )
=> caught-it).

Sebastian.