User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; sv-SE; rv:1.9.2.8)
            Gecko/20100802 Thunderbird/3.1.2
MIME-Version: 1.0
References: <4C9905B4.2080802@morningstar2.co.uk>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
Message-ID:  <4C99F36F.5010106@residenset.net>
Date:         Wed, 22 Sep 2010 14:15:43 +0200
Reply-To: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
Sender: Mailing list for the LaTeX3 project <LATEX-L@listserv.uni-heidelberg.de>
From: =?ISO-8859-1?Q?Lars_Hellstr=F6m?= <Lars.Hellstrom@RESIDENSET.NET>
Subject: Re: Quarks: revisions of use in expl3
To: LATEX-L@listserv.uni-heidelberg.de
In-Reply-To:  <4C9905B4.2080802@morningstar2.co.uk>
Precedence: list
Status: R

Joseph Wright skrev 2010-09-21 21.21:
> Hello all,
>
> The documented use for different quarks in expl3 is broadly as follows:
> - \q_stop and \q_mark for delimiting arguments, with \q_stop used
> mainly
> - \q_nil as a nil pointer

\begin{remark}
Which seems a very odd concept to have within TeX, but should probably 
be interpreted as "something that would have been a nil pointer, had 
LaTeX been implemented in LISP" (which it of course isn't, but 
allegedly Lamport drafted LaTeX in some sort of LISP pseudocode).
\end{remark}

> - \q_no_value to indicate the absence of a value at all
>
> However, the use of these quarks in expl3 has been rather less well
> defined, with \q_nil in particular used in places where \q_stop is most
> appropriate. We've therefore revised the applications, and the SVN
> version of expl3 is now much more systematic in this regard.
>
> In the main, this change should not affect the code others are using as
> these are mainly internal features. However, one place that the
> difference may impact is \tl_head:w and \tl_tail:w. These now terminate
> with \q_stop not \q_nil.
>
> There will be a CTAN update of expl3 and xpackages soon (by the
> weekend, I hope). This will include these changes (along with others:
> see my other e-mails).

Since I was successful regarding the boolean syntax, I think I'd better 
try reraising a point regarding quarks as well.

Claim 1. The main property required for a quark is that it is a control 
sequence that is distinct from everything that could reasonably be 
constructed from normal materials; in particular, it must be such that 
the delimited argument in a \def\macro#1\quark{...} grabs material up 
to precisely that quark (and no other quark).

Claim 2. The only other property employed is that it is possible to 
test a sequence #1 of tokens for being \quark_X by doing
   \def\temp_a{#1} \ifx \temp_a \quark_X

Claim 3. These properties are obtained by defining a quark as a macro 
that expands to itself. Since the control sequence is contained in its 
own definition, the resulting macro contains the information of which 
quark it is, and it is thus distinct from other quarks.

This does however lead to the main problem with quarks, that the 
documentation sternly warns about: if you ever were to attempt a full 
expansion of a quark then TeX would get stuck in an infinite loop. This 
dire consequence will of course encourage programmers to get the 
details right whenever they work with quarks, but in my experience 
these are the kind of details many people have hard to get right 
(sometimes because they're applying the data model of LISP to a TeX 
language which is quite different), so There Will Be Bugs. It would 
therefore be nice if the bugs could have less draconic consequences.

The idea I have proposed is to change the definition of quarks from 
effectively

   \def\q_nil{\q_nil}
to
   \def\q_nil{\string\q_nil}

(this \string should probably be called \quark_protect:N or some such, 
but the primitive names are less obscure in a discussion) as a mistaken 
expansion of the latter does not lead to an infinite loop; even in 
corny contexts such as

   \expandafter \expandafter \q_nil

(spaces added after control sequence to distinguish from string of 
character tokens) expansion safely terminates via

   \expandafter \string \q_nil
   \string \string \q_nil
   \string\q_nil  % No space after \string, since this \string
                  % is not a control sequence
   \string\string \q_nil
   \string\q_nil% Only "other" tokens.

Claim 1 continues to hold for this modified definition. Claim 2 can be 
restored if the test for being \quark_X is changed to

   \def\temp_a{\string#1} \ifx \temp_a \quark_X

(I get the impression that this kind of test should anyway be handled 
by a expl3-provided command rather than explicit user code, so the 
number of places to change wouldn't be very large.)

So wouldn't it be better if this was changed?

Lars Hellström