User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:17.0)
            Gecko/20130216 Thunderbird/17.0.3
MIME-Version: 1.0
References: <CACKZBPXMfsWiaFgkHGG+Yb3kzfr1EXdJzA74i7C7ZfqVaGGPkw@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Message-ID:  <513B4E80.2070409@morningstar2.co.uk>
Date:         Sat, 9 Mar 2013 15:00:16 +0000
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: Joseph Wright <joseph.wright@MORNINGSTAR2.CO.UK>
Subject: Re: Bug + fix for \prop_map_tokens:Nn
To: LATEX-L@LISTSERV.UNI-HEIDELBERG.DE
In-Reply-To:  <CACKZBPXMfsWiaFgkHGG+Yb3kzfr1EXdJzA74i7C7ZfqVaGGPkw@mail.gmail.com>
Precedence: list
Envelope-To: <rainer.schoepf@GMX.NET>
Status: R

On 09/03/2013 11:55, Michiel Helvensteijn wrote:
> Hi all,
> 
> I encountered a bug in the \prop_map_tokens:Nn function in l3candidates.dtx.
> 
> Here's a not-working minimal example:
> 
> ----------------------------------------
> \ExplSyntaxOn
>     \cs_set:Nn \test_mapper:nn
>         { (#1,~#2)~ }
>     \prop_new:N \testprop
>     \prop_put:Nnn \testprop {1} {v1}
>     \prop_put:Nnn \testprop {222} {v2}
>     \prop_put:Nnn \testprop {3} {v3}
>     \prop_put:Nnn \testprop {4} {v4}
>     \prop_map_tokens:Nn \testprop { \test_mapper:nn }
> \ExplSyntaxOff
> ----------------------------------------
> 
> The output will be:
> 
> ----------------------------------------
> (1, v1)2
> ----------------------------------------
> 
> It only processes single-token keys correctly. At the first
> multi-token key, everything except for its first two tokens are
> processed directly, then the rest of the list is skipped.
> 
> Here's the fix for l3candidates.dtx:
> 
> ----------------------------------------
> \cs_new:Npn \@@_map_tokens:nwn #1 \q__prop #2 \q__prop #3
>   {
>     \__quark_if_recursion_tail_break:nN {#2} \prop_map_break:
>     \use:n {#1} {#2} {#3}
>     \@@_map_tokens:nwn {#1}
>   }
> ----------------------------------------
> 
> It just changes the tail-break test from :NN to :nN and adds braces around #2.
> 
> Cheers!

The full fix is actually a bit more involved as using
\__quark_if_recursion_tail_break:nN will give the wrong result if your
key is called \q_recursion_tail. In the SVN, this is fixed and I've also
added a test.

I'm expecting to do a CTAN update very soon: I just need to wait for one
third party to update their package on CTAN (so we don't break documents).
-- 
Joseph Wright