Unicode characters can sometimes not be read in the form they are emitted

[triska]

Let's ask Scryer Prolog whether there are any other cases like #1768, where the system itself emits a term that cannot be read back with read/1:

?- length(_, N),
   char_code(Char, N),
   write_term_to_chars(Char, [quoted(true)], Cs0),
   append(Cs0, " .", Cs),
   catch(read_from_chars(Cs, Term), error(syntax_error(_),_), portray_clause(cannot_read(N,Char))),
   false.

With #1799 and #1800 already applied, I get the following 17 remaining cases:

cannot_read(5760,' ').
cannot_read(8192,' ').
cannot_read(8193,' ').
cannot_read(8194,' ').
cannot_read(8195,' ').
cannot_read(8196,' ').
cannot_read(8197,' ').
cannot_read(8198,' ').
cannot_read(8199,' ').
cannot_read(8200,' ').
cannot_read(8201,' ').
cannot_read(8202,' ').
cannot_read(8232,'
').
cannot_read(8233,'
').
cannot_read(8239,' ').
cannot_read(8287,' ').
cannot_read(12288,'　').

For instance, taking the first example:

?- char_code(Char, 5760).
   Char = ' '.
?- Char = ' '.
   error(syntax_error(invalid_single_quoted_character),read_term/3).

The overall impression I get from this is that the situation is very hopeful: If these remaining cases cannot be meaningfully checked for with one of the available character categorizations, then we can simply add these rather few cases as special cases that need to be emitted as hexadecimal escape sequences.

Low priority. For syntactic ISO conformance, #1771, #1773, #1778 etc. are far more important.

[UWN]

It seems, all these characters make sense within a quoted context. I mean, 1/4 has always been a regular printable character in Latin1. Hexadecimal sequences should be reserved for the truly ambiguous cases, like the nbsp.

[triska]

I now found out that applying #1799 reduces the list to 17 remaining cases, I have updated the list in the original post!

[triska]

Resolved via #1805.