moo-discuss

[MOO-Discuss] Non-Printable Characters in Strings

[Luke-Jr <lu...@us...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

How should non-printable characters be represented in strings?
Is C-style escaping of common characters (\n, \r, \t, \v, \x##, etc) a good=
=20
idea? Including C-style's \### which uses octal or should \### use decimal=
=20
(or hex?)?
Should the following string be permitted (it is currently not)? "this is a \
multiline string" or require the programmer to use \n?
Would it be a good or bad idea to parse \^A through \^Z into character 0x01=
 to=20
0x1A? What about parsing \e to 0x1B and \s1 to \s4 and \sf \sg \sr \su=20
(seperator characters) to 0x1C to 0x1F?
Anything else?

Comments please.
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAFweNZl/BHdU+lYMRAt1XAKCWyf/N5U/tro37qvmRgXJPTjFgGQCgmK9y
=46cTTf+zmtu8A4iOs1CmOX/o=3D
=3DcxE5
=2D----END PGP SIGNATURE-----

Re: [MOO-Discuss] Non-Printable Characters in Strings

[Luke-Jr <lu...@ar...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wednesday 28 January 2004 12:51 am, Luke-Jr wrote:
> Is C-style escaping of common characters (\n, \r, \t, \v, \x##, etc) a go=
od
> idea? Including C-style's \### which uses octal or should \### use decimal
> (or hex?)?
Also, if number-based escaping is used, should it be octet-based or=20
character-based?
Character-based would translate the value to UTF-8 and could accept a=20
variable-length number up to a 16-bit value. For example, \xFFFF, \d65535,=
=20
\o200000, \xFF, \xF \x (for a null character). (Using variable-length numbe=
rs=20
would, of course, be discouraged since it could easilly create bugs if the=
=20
character following the escaped one was a valid digit).
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAJdTnZl/BHdU+lYMRAhehAJ9ACWql1esncEupVjXjtCrU+rInbgCffIKu
8utozoE08zAegJuBJifYvJc=3D
=3D3JQ/
=2D----END PGP SIGNATURE-----

Re: [MOO-Discuss] Non-Printable Characters in Strings

[Gavin L. <md...@mi...>]

At 00:51 28/01/2004 +0000, Luke-Jr wrote:
 >
 >How should non-printable characters be represented in strings?
 >Is C-style escaping of common characters (\n, \r, \t, \v, \x##,
 >etc) a good idea? Including C-style's \### which uses octal or
 >should \### use decimal (or hex?)?

I've always regarded C-style escaping as a little messy... but 
then, that's mostly because I need to use the backslash a lot for 
paths and the like.

One possible alternative (though I don't really know if it's any 
better) is to use Delphi-style escaping, a la:
   "This is a string with"#10"a newline in the middle."
   "This is a string with "" a double-quote in the middle."
   "This is a string that " +
     "spans two source lines."

Of course, the latter example is already valid in MOO code, but of 
course the internal compiler/decompiler will merge it back onto 
one source line.

-- 
Gavin Lambert, Mirality Systems
<http://www.mirality.co.nz/>
----
Pardon me, waiter. I like my water diluted.

Re: [MOO-Discuss] Non-Printable Characters in Strings

[Luke-Jr <lu...@ar...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Monday 09 February 2004 08:01 pm, Gavin Lambert wrote:
> At 00:51 28/01/2004 +0000, Luke-Jr wrote:
>  >How should non-printable characters be represented in strings?
>  >Is C-style escaping of common characters (\n, \r, \t, \v, \x##,
>  >etc) a good idea? Including C-style's \### which uses octal or
>  >should \### use decimal (or hex?)?
>
> I've always regarded C-style escaping as a little messy... but
> then, that's mostly because I need to use the backslash a lot for
> paths and the like.
That's probably partially due to your operating system not using standard p=
ath=20
syntax. Most systems would use / as a directory seperator which doesn't hav=
e=20
this problem. Further, since MOO already uses the \ character for escaping,=
 I=20
wouldn't suggest breaking it. I was referring to the sequences following th=
e=20
\ character such as 'n', 'r', 't', 'x__', etc...
>
> One possible alternative (though I don't really know if it's any
> better) is to use Delphi-style escaping, a la:
>    "This is a string with"#10"a newline in the middle."
>    "This is a string with "" a double-quote in the middle."
>    "This is a string that " +
>      "spans two source lines."
With regard to the \<newline> part, I probably wasn't clear... I was thinki=
ng=20
using \<newline> as a way to insert a *literal* newline into the string, no=
t=20
as a way to continue the string on another line. Such a possiblity is an id=
ea=20
worth considering, though.
>
> Of course, the latter example is already valid in MOO code, but of
> course the internal compiler/decompiler will merge it back onto
> one source line.
I don't bother going to extreme trouble to keep compile/decompile formattin=
g=20
the same... Using my patch which renders 0xFF as an integer, for example, t=
he=20
decompiler will still output 255. To render it as 0xFF would probably requi=
re=20
adding a flag on Vars of TYPE_INT or a new TYPE_INTBASE (which has the flag=
=20
instead, but is treated as INT within the MOO), but both cases would either=
=20
use a lot of RAM and/or a lot of extentions to existing code when it's=20
probably not worth it. More important would be to preserve /* this style */=
=20
of comments which the MOO currently forgets completely, I think.
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAKA1VZl/BHdU+lYMRAm91AJ9Mh1GOF44SMAHBj5RKQglYHualDwCeKeqB
swhLjeVp7XAHPM/pYcPmOPc=3D
=3DT+Qt
=2D----END PGP SIGNATURE-----

Re: [MOO-Discuss] Non-Printable Characters in Strings

[Luke-Jr <lu...@ar...>]

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Attached is a patch for a vanilla LambdaMOO server to add support for C-sty=
le=20
escape codes, \### (octal), \^@ through \^_ (see ASCII chart for details on=
=20
range), \e, \s1 to \s4, \sf, \sg, \sr, and \su. Octet-based, but VBR length=
=20
numbers for \x##, \o###, and \d### are supported.
This patch does not address \<newline>, and such a sequence will continue t=
o=20
error. It has the potential for memory leaking since character 0x00 can be=
=20
used and the LambdaMOO server cannot properly handle it in strings (yet).
\### can easilly be changed to use decimal, but unless there are any=20
objections, I think octal is more logical. Note that hex is not an option,=
=20
since it would conflict with things such as \a or \b, which are C-style=20
escape codes.
As before, comments are more than welcome. :)

On Wednesday 28 January 2004 12:51 am, Luke-Jr wrote:
> Is C-style escaping of common characters (\n, \r, \t, \v, \x##, etc) a go=
od
> idea? Including C-style's \### which uses octal or should \### use decimal
> (or hex?)?
> Should the following string be permitted (it is currently not)? "this is a
> \ multiline string" or require the programmer to use \n?
> Would it be a good or bad idea to parse \^A through \^Z into character 0x=
01
> to 0x1A? What about parsing \e to 0x1B and \s1 to \s4 and \sf \sg \sr \su
> (seperator characters) to 0x1C to 0x1F?
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAKEQsZl/BHdU+lYMRAuVRAJ4yHURT+Dz5HCwffbuc5jUS4vay6wCdHXLr
Y7dP1hFXts9RBbYJwaTDWYY=3D
=3DyNH5
=2D----END PGP SIGNATURE-----