Re: [Chicken-hackers] [PATCH] Restore row and column counting for ports (fixes #978)

[John Cowan]

Peter Bex scripsit:

> If this turns out to be true that this is a big bottleneck, it makes
> sense to leave the bookkeeping up to applications.  Perhaps it could be
> convenient to add a custom "counted port" that can wrap other port
> types.

That's what I would recommend, rather than (a) carefully tracking in
all locations, with resultant overhead, or (b) tracking in only some
locations, resulting in bogus row and column values.

> Our current implementation of column counting does not take into
> account multibyte UTF-8 characters.  I'm not sure if we need to take
> care of this.  An input file might not be in UTF-8 encoding, in which
> case "fixing" UTF-8 counting will cause the count to be incorrect on
> those files.  We could just assume all ports are UTF-8 since that's the
> "native" Chicken character set.  The only way to *truly* fix this is
> to make ports aware of their encoding, but that's just... annoying :)

What's more, column width becomes a tricky matter when you have to handle
the full repertoire of Unicode characters, for they can have widths of 0
(combining characters, non-printing characters), 1 (most characters),
or 2 (CJK ideographs, U+3000 IDEOGRAPHIC SPACE, and U+FF01 to U+FF5E,
the FULLWIDTH variants of the ASCII repertoire).  That's not baggage
that should be carted around everywhere, but it would be fine in a
counted-port wrapper.

See UAX #11 <http://www.unicode.org/reports/tr11>, which is a detached
part of the Unicode Standard, for details.

-- 
So they play that [tune] on                     John Cowan
their fascist banjos, eh?                       address@hidden
        --Great-Souled Sam                      http://www.ccil.org/~cowan