[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: If you like "u8vector" ...

This page is part of the web mail archives of SRFI 66 from before July 7th, 2015. The new archives for SRFI 66 contain all messages, not just those from before July 7th, 2015.

Michael Sperber wrote:
Yes, that makes sense.  However, aren't you really arguing for
Taylor's idea of having a single type for "blobs" and then having
access methods for various sequences of bits inside them?

It might be nice - but I haven't seen a specific proposal.

For byte and other uniform vectors random acces is useful.
E.g. floating-point matrix operations.

For parsing a complicated "blob" (either in memory, on disk,
or a network stream) you normally want sequential access,
and being able to mix data types: e.g. a u32 count whose value
is N is followed by N instances of u8, followed by a u64 checksum.

> (I have to
admit that I'm warming increasingly to this idea.)  Otherwise, the
concept is going to bloat every API that deals with similar objects
beyond recognition.  This way, most of the bloat would be right with
the "blob API" and stay there.  What do you think?

The devil is in the details ...
	--Per Bothner
per@xxxxxxxxxxx   http://per.bothner.com/