David B. Serafini writes:
> This doesn't solve Richard Maine's problem, but nothing at the language level
> is going to solve his problem because he's reading data as one data type and
> using it as another (if I understand correctly).
Yep. Basically for decent performance. I suppose that one *COULD*
read the data in some non-advancing mode a variable at a time, after
you figured out the appropriate type. But that's no way to get
acceptable performance (which is definitely an issue for me). You
want to read decent-sized chunks into a buffer and not bug the rtl,
with all its overhead, for each data value.
> Embedding a self-documenting file format wiht meta-data (like HDF) into an I/O
> rtl has its own problems: what happens when the data types in a READ statement
> don't correspond to the data types in the meta data in hte file?...
which is pretty much the same kind of problem I was talking about.
--
Richard Maine
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|