When programming with the new array interface I've noticed an
inconvenience that should be easy to fix:
When checking the arguments of a procedure there's a few things one
normally checks apart from the type. It's common that a procedure
takes an array of a specific rank, and it's not unusual that only
arrays of a specific dimension qualifies as an argument.
Since both rank and dimensionality are extracted from the array handle
this handle needs to be allocated *before* argument checking is
complete. This means that we have to free the handle on an error exit.
So, the code for argument checking only becomes:
SCM_ASSERT (scm_is_typed_array (a, <array type>), a, <pos>, <function name>);
scm_array_get_handle (a, &h);
dims = scm_array_handle_dims (&h);
if (scm_array_handle_rank (&h) != <rank> || dims[0].ubnd -
dims[0].lbnd + 1 != <dim1> || ...)
{
scm_array_handle_release (&h);
scm_wrong_type_arg (<function nam>, <pos>, a);
}
Since it seems unreasonable to allow future code to modify the
dimension of an array, I don't see any reason why we couldn't ask for
the rank and dimensions fo an array without allocating the handle. The
above code could then instead look like this:
SCM_ASSERT (scm_is_typed_array (a, <array type>), a, <pos>, <function name>);
dims = scm_array_dims (a);
SCM_ASSERT ( scm_array_rank (a) != <rank> || dims[0].ubnd -
dims[0].lbnd + 1 != <dim1> || ...,
a, <pos>, <function name>);
scm_array_get_handle (a, &h);
This all of course depends on the first sentence in the description of
how to handle allocation of handles in the reference manual:
"You must take care to always unreserve an array after reserving it,
also in the presence of non-local exits. To simplify this, reserving
and unreserving work like a frame (*note Frames::): a call to
`scm_array_get_handle' can be thought of as beginning a frame and
`scm_array_handle_release' as ending it. When a non-local exit happens
between these two calls, the array is implicitely unreserved."
The part after the first sentence indicates that I don't have to call
scm_array_handle_release before throwing the error. Is that so? In
that case this suggestion of change is unnecessary.
Comments?
M
_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel
