¤ cannot be used

[triska]

Currently, I get:

?- X = ¤.
   error(syntax_error(unexpected_char),read_term/3:0).
?- X = '¤'.
   error(syntax_error(invalid_single_quoted_character),read_term/3).
?- X = "¤".
   error(syntax_error(missing_quote),read_term/3:0).

Is there anything special about this character? Why cannot it be used like other symbols/currency characters such as $?

For comparison, I get with GNU Prolog:

| ?- X = '¤'.

X = '¤'

yes

[triska]

On the other hand, the toplevel reports the character when asked for it in a different way:

?- char_code(C, 164).
   C = '¤'.

This means that such answer substitutions currently cannot be pasted back as queries.

[UWN]

We need to identify that class of characters which can be written directly with single or double quotes.

non quote char (* 6.4.2.1 *)
   = graphic char (* 6.5.1 *)
   | alphanumeric char (* 6.5.2 *)
   | solo char (* 6.5.3 *)
   | space char (* 6.5.4 *)
   | meta escape sequence (* 6.4.2.1 *)
   | control escape sequence (* 6.4.2.1 *)
   | octal escape sequence (* 6.4.2.1 *)
   | hexadecimal escape sequence (* 6.4.2.1 *) ;

which is what is used to define those characters that can appear in such quoted context. So extended characters (6.5) that look like characters may be added to non quote char but neither to one of those defined above like graphic char or alphanumeric char.

[triska]

¬ has the same problem (see also #1591):

?- char_code(C, 172).
   C = '¬'.
?- C = '¬'.
   error(syntax_error(invalid_single_quoted_character),read_term/3).

[UWN]

You can also extend graphic char and alphanumeric char.

It is a possibility to do so. And for alphanumeric chars Scryer already does this. The question here is rather whether or not it makes sense to extend graphic char which may make the source code much less readable and reliable. Libraries like TPTP have refrained from this, and this although it was suggested in related discussions.

The other question is which non-terminals may be extended. In 6.5 graphic char, alphanumeric char, solo char, layout char and meta char are mentioned. From NOTE 2 it becomes evident that also small letter char and capital letter char can be extended separately. So far in the standard there is no character that can be used in a quoted context but not in another context. But here the mentioned non quote char may be a safer choice for extension. This would be also a bit closer to the way other programming languages like C (WG14 N1518) do it.

Things would get even more complex, when also solo char is extended. So some new characters would be graphic and others solo. Sticking to a more conservative extension seems preferable. Such a more conservative extension does not rule out the use of such symbols but requires that quotes are used to make them better visible.

[triska]

This works perfectly now, thank you a lot!

?- X = ¤.
   X = ¤.

[triska]

I am reopening this because I think it is expected (#1749 (comment)) that the symbol cannot be part of a letter token:

?- X = ¤a.
   X = ¤a, unexpected.

[triska]

Now it seems to work really perfectly, thank you a lot!

[hurufu]

I'm afraid there is a regression on the latest master (v0.9.4-165-g12a61cdf):

?- X = ¤.
   error(syntax_error(unexpected_char),read_term/3:0).
?- 

Also might be the cause of #2547

[bakaq]

Probably related to #2529 and #2496 too.

[triska]

I am reopening this issue because the original issue is now present again, for unclear reasons. At the very least, it requires a conscious decision on how to resolve it.