path: root/macros/latex-dev/base/ltsockets.dtx

% \iffalse meta-comment
%
% Copyright (C) 2023
%       Frank Mittelbach, The LaTeX Project
%
% This file is part of the LaTeX base system.
% -------------------------------------------
%
% It may be distributed and/or modified under the
% conditions of the LaTeX Project Public License, either version 1.3c
% of this license or (at your option) any later version.
% The latest version of this license is in
%    https://www.latex-project.org/lppl.txt
% and version 1.3c or later is part of all distributions of LaTeX
% version 2008 or later.
%
% This file has the LPPL maintenance status "maintained".
%
% The list of all files belonging to the LaTeX base distribution is
% given in the file `manifest.txt'. See also `legal.txt' for additional
% information.
%
% The list of derived (unpacked) files belonging to the distribution
% and covered by LPPL is defined by the unpacking scripts (with
% extension .ins) which are part of the distribution.
%
% \fi
%
% \iffalse
%
%%% From File: ltsockets.dtx
%
%    \begin{macrocode}
\def\ltsocketsversion{0.9a}
\def\ltsocketsdate{2023-08-21}
%    \end{macrocode}
%
%<*driver>
\documentclass{l3doc}


\providecommand\InternalDetectionOff{}
\providecommand\InternalDetectionOn{}

\EnableCrossrefs
\CodelineIndex
\begin{document}
  \DocInput{ltsockets.dtx}
\end{document}
%</driver>
%
% \fi
%
%
%
%
% \title{\LaTeX{}'s socket management\thanks{This module has version
%    \ltsocketsversion\ dated \ltsocketsdate, \copyright\ \LaTeX\
%    Project.}}
%
% \author{Frank Mittelbach}
%
% \maketitle
%
% ^^A this needs some cleanup:
% \long\def\fmi#1{\begin{quote} TODO: \itshape #1\end{quote}}
%
% \providecommand\pkg[1]{\texttt{#1}}
% \providecommand\hook[1]{\texttt{#1}}
% \providecommand\env[1]{\texttt{#1}}
% \providecommand\plug[1]{\texttt{#1}}
% \providecommand\socket[1]{\texttt{#1}}
%
%
% \begin{abstract}
%    This code implements sockets which are places in the code into
%    which predeclared chunks of code (plugs) can be placed. Both the sockets
%    and the plugs are \enquote{named} and each socket is
%    assigned exactly one plug at any given time.
% \end{abstract}
%
% \tableofcontents
%
% \section{Introduction}
%
% A \LaTeX{} source file is transformed into a typeset document by
% executing code for each command or environment in the document
% source. Through various steps this code transforms the input and
% eventually generates typeset output appearing in a \enquote{galley}
% from which individual pages are cut off in an asyncronous way. This
% page generating process is normally not directly associated with
% commands in the input\footnote{Excepts for directives such as
%   \cs{newpage}.} but is triggered whenever the galley has received
% enough material to form another page (giving current settings).
% 
% As part of this transformation input data may get stored in some form
% and later reused, for example, as part of the output routine
% processing.
% 
% \section{Configuration of the transformation process}
% 
% There are three different major methods offered by \LaTeX{} to
% configure the transformation process:
% \begin{itemize}
% \item through the template mechanism,
% \item through the hook mechanism, or
% \item through sockets and plugs.
% \end{itemize}
% They offer different possibilities (with different features and
% limitations) and are intended for specific use cases, though it is
% possible to combine them.
% 
% \subsection{The template mechanism}
% 
% The template mechanism is intended for more complex document-level
% elements (e.g., headings such as \verb=\section= or environments like
% \env{itemize}). The template code implements the overall processing
% logic for such an element and offers a set of parameters to influence
% the final result.
% 
% The document element is then implemented by (a) selecting a suitable
% template (there may be more than one available for the kind of
% document element) and (b) by setting its parameters to desired
% values. This then forms a so-called instance which is executed when
% the document element is found in the source.
% 
% By altering the parameter values (in a document class or in the
% document preamble) or, if more drastic layout changes are desired, by
% selecting a different template and then adjusting its parameters, a
% wide variety of layouts can be realized through simple configuration
% setups without the need to develop new code.
% 
% The target audience of this method are therefore document class
% developers or users who wish to alter an existing layout (implemented
% by a document class) in certain (minor) ways.
% 
% The template mechanism is currently documented as part of the
% \pkg{xtemplate} package and one more elaborate implementation can be
% found as part of the \texttt{latex-lab} code for lists (to be
% documented further).
% 
% \subsection{The hook mechanism}
% 
% Hooks are places in the kernel code (or in packages) that offer
% packages the possibility to inject additional code at specific
% points in the processing in a controlled way without the need to
% replace the existing code block (and thereby overwriting
% modifications/extensions made by other packages). The target
% audience is therefore mainly package developers, even though some
% hooks can be useful for document authors.
% 
% Obviously, what can reasonably be added into a hook depends on the
% individual hook (hopefully documented as part of the hook
% documentation), but in general the idea behind hooks is that more
% than one package could add code into the hook at the same
% time. Perhaps the most famous hook (that \LaTeX{} had for a very
% long time) is \hook{begindocument} into which many packages add code
% to through \cs{AtBeginDocument}\marg{code} (which is nowadays
% implemented as a shorthand for
% \cs{AddToHook}\texttt{\{\hook{begindocument}\}}\marg{code}). To
% resolve possible conflicts between injections by different packages
% there is a rule mechanism by which code chunks in a hook can be
% ordered in a certain way and by which incompatible packages can be
% detected if a resolution is impossible.
% 
% In contrast to template code, there is no standard configuration
% method through parameters for hooks, i.e., the code added to a hook
% \enquote{is} the configuration. If it wants to provide for
% configuration through parameters it has to also provide its own
% method to set such parameters in some way. However, in that case it
% is likely that using a hook is not the right approach and the
% developer better calls a template instance instead which then offers
% configuration through a key/value interface.
% 
% In most cases, hooks do not take any arguments as input. Instead, the data
% that they can (and are allowed to) access depends on the surrounding
% context.
% 
% For example, the various hooks available during the page shipout
% process in \LaTeX's output routine can (and have to) access the
% accumulated page material stored in a box named
% \verb=\ShipoutBox=. This way, code added to, say, the
% \hook{shipout/before} hook could access the page content, alter it,
% and then write it back into \verb=\ShipoutBox= and any other code
% added to this hook could then operate on the modified content.  Of
% course, for such a scheme to work the code prior to executing the hook
% would need to setup up data in appropriate places and the hook
% documentation would need to document what kind of storage can be
% accessed (and possibly altered) by the hook.
% 
% There are also hooks that take arguments (typically portions of
% document data) and in that case the hook code can access these
% arguments through \verb=#1=, \verb=#2=, etc.
% 
% The hook mechanism is documented in \texttt{lthooks-doc.pdf}.
%
%
%
% \subsection{The socket mechanism}
% 
% In some cases there is code that implements a certain programming
% logic (for example, combining footnotes, floats, and the text for the
% current page to be shipped out) and if this logic should change (e.g.,
% footnotes to be placed above bottom floats instead of below) then this
% whole code block needs to be replaced with different code.
% 
% In theory, this could be implemented with templates, i.e., the code
% simply calls some instance that implements the logic and that instance
% is altered by selecting a different templates and/or adjusting their
% parameters. However, in many cases customization through parameters is
% overkill in such a case (or otherwise awkward, because paramerization
% is better done on a higher level instead of individually for small
% blocks of code) and using the template mechanism just to replace one
% block of code with a different one results in a fairly high
% performance hit. It is therefore usually not a good choice.
% 
% In theory, it would also be possible to use a hook, but again that is
% basically a misuse of the concept, because in this use case there should
% never be more than one block of code inside the hook; thus, to alter
% the processing logic one would need to set up rules that replace code
% rather than (as intended) execute all code added to the hook.
% 
% 
% 
% For this reason \LaTeX{} now offers a third mechanism:
% \enquote{sockets} into which one can place exactly one code block
%   --- a \enquote{plug}.
%
% In a nutshell: instead of having a fixed code block somewhere as part
% of the code, implementing a certain programming logic there is a
% reference to a named socket at this point.
% This is done by first declaring the named socket with:
% \begin{quote}
% \cs{NewSocket}\marg{socket-name}\marg{number-of-inputs}
% \end{quote}
% This is then referenced at the point where the replaceable code block
% should be executed with:
% \begin{quote}
%  \cs{UseSocket}\marg{socket-name}
% \end{quote}
% or, if the socket should take a number of inputs (additional
%    arguments beside the name) with
% \begin{quote}
%   \cs{UseSocket}\marg{socket-name}\marg{arg\textsubscript{1}}\ldots
%   \marg{arg\textsubscript{\meta{number-of-inputs}}}
% \end{quote}
% 
% In addition, several code blocks (a.k.a.\ plugs) implementing different logic for this
% socket are set up, each with a declaration of the form:
% \begin{quote}
%   \cs{NewSocketPlug}\marg{socket-name}\marg{socket-plug-name}\marg{code}
% \end{quote}
% Finally,
% one of them is assigned to the socket:
% \begin{quote}
% \cs{AssignSocketPlug}\marg{socket-name}\marg{socket-plug-name}
% \end{quote}
% If the programming logic should change, then all that is necessary is
% to make a new assignment with \cs{AssignSocketPlug} to a different
% \marg{socket-plug-name}. This assignment obeys scope so that an
% environment can alter a socket without the need to restore the
% previous  setting manually.
% 
% If the socket takes inputs, then those need to be provided to
% \cs{UseSocket} and in that case they can be referenced in the \meta{code}
% argument of \cs{NewSocketPlug} with \verb=#1=, \verb=#2=, etc.
% 
% In most cases a named socket is used only in a single place, but there
% is, of course, nothing wrong with using it in several places, as long
% as the code in all places is supposed to change in the same way.
% 
% 
% 
% 
% 
% \subsubsection{Examples}
% 
% 
% We start by declaring a new socket named \socket{foo} that expects
% two inputs:
%\begin{verbatim}
%      \NewSocket{foo}{2}
%\end{verbatim}
% \NewSocket{foo}{2}
% 
% Such a declaration has do be unique across the whole \LaTeX{}
% run. Thus, if another package attempts to use the same name
% (regardless of the number of inputs) it will generate an error:
%\begin{verbatim}
%      \NewSocket{foo}{2}
%      \NewSocket{foo}{1}
% \end{verbatim}
% Both declarations would therefore produce:
% \begin{verbatim}
%      ! LaTeX socket Error: Socket 'foo' already declared!
%\end{verbatim}
% 
% You also get an error if you attempt to declare some socket plug and
% the socket name is not yet declared, e.g.,
%\begin{verbatim}
%      \NewSocketPlug{baz}{undeclared}{some code}
%\end{verbatim}
% generates
%\begin{verbatim}
%      ! LaTeX socket Error: Socket 'baz' undeclared!
%\end{verbatim}
% 
% 
% Setting up plugs for the socket is done like this:
%\begin{verbatim}
%      \NewSocketPlug{foo}{plug-A}
%         {\begin{quote}\itshape foo-A: #1!#2\end{quote}}
%      \NewSocketPlug{foo}{plug-B}
%         {\begin{quote}\sffamily foo-B: #2\textsuperscript{2}\end{quote}}
%\end{verbatim}
% This will set up the plugs \texttt{plug-A} and \texttt{plug-B} for
% this socket.
% 
%\NewSocketPlug{foo}{plug-A}{\begin{quote}\itshape foo-A: #1!#2\end{quote}}
%\NewSocketPlug{foo}{plug-B}{\begin{quote}\sffamily foo-B: #2\textsuperscript{2}\end{quote}}
% 
% We still have to assign one or the other to the socket, thus without
% doing that the line
% \begin{verbatim}
%      \UseSocket{foo}{hello}{world}
% \end{verbatim}
% produces nothing because the default plug for sockets with 2 inputs
% is \plug{noop} (which grabs the additional arguments and throws them
%   away).\footnote{If socket \socket{foo} would have been a socket with one
%   input, then the default plug would be \plug{identity}, in which case
%   the socket input would remain without braces and gets typeset!}
% 
% \UseSocket{foo}{hello}{world}     ^^A nothing happens
% 
% So let's do the assignment
% \begin{verbatim}
%      \AssignSocketPlug{foo}{plug-A}
% \end{verbatim}
% and then
% \begin{verbatim}
%      \UseSocket{foo}{hello}{world}
% \end{verbatim}
% will properly typeset
% \AssignSocketPlug{foo}{plug-A}\UseSocket{foo}{hello}{world}
% and after
% \begin{verbatim}
%      \AssignSocketPlug{foo}{plug-B}
% \end{verbatim}
% and another call to
% \begin{verbatim}
%      \UseSocket{foo}{hello}{world}
% \end{verbatim}
% we get
% \AssignSocketPlug{foo}{plug-B}\UseSocket{foo}{hello}{world}
% 
% If we attempt to assign a plug that was not defined, e.g.,
% \begin{verbatim}
%      \AssignSocketPlug{foo}{plug-C}
% \end{verbatim}
% then we get an error during the assignment
% \begin{verbatim}
%      ! LaTeX socket Error: Plug 'plug-C' for socket 'foo' undeclared!
% \end{verbatim}
% and the previous assignment remains in place.
% 
% To see what is known about a socket and its plugs you can use
% \cs{ShowSocket} or \cs{LogSocket} which displays information similar
%   to this on the terminal or in the transcript file:
% \begin{verbatim}
%     Socket foo:
%         number of inputs = 2
%         available plugs = noop, plug-A, plug-B
%         current plug = plug-B
%         definition = \protected\long macro:#1#2->\begin {quote}\sffamily
%     foo-B: #2\textsuperscript {2}\end {quote}
% \end{verbatim}
% \LogSocket{foo}
% 
%
% \subsubsection{Details and semantics}
% 
% In this section we collect some normative statements.
% 
% \begin{itemize}
% 
% \item
%  From a functional point of view sockets are like simple \TeX{} macros,
%   i.e., they expect 0 to 9 mandatory arguments (the socket inputs) and get replaced by
%   their \enquote{expansion}
% 
% \item
%   A socket is \enquote{named} and the name consists of ASCII letters
%   \texttt{[a-z]},
%   \texttt{[A-Z]}, \texttt{[0-9]}, \texttt{[-/@]} only
% 
% \item
%   Socket names have to be unique, i.e., there can be only one socket
%   named \meta{name}. This is ensured by declaring each socket with
%   \cs{NewSocket}.
%
%   However, there is no requirement that sockets and
%   hook names have to be different. In fact, if a certain action that
%   could overwise be specified as hook code has to be executed always
%   last (or first) one could ensure this by placing a socket (single
%   action) after a hook (or vice versa) and using the same name to
%   indicate the relationship, e.g., 
%\begin{verbatim}
%  \UseHook{foo}        % different package can add code here
%  \UseSocket{foo}      % only one package can assign a plug
%\end{verbatim}
%   This avoids the need to order the hook code to ensure that
%   something is always last.
% 
% \item
%  Best practice naming conventions are \ldots\ \emph{to be documented}
% 
% \item
%    A socket has documented inputs which are
% 
%     \begin{itemize}
%     \item
%       the positional arguments (if any) with a description of what
%       they contain when used
%     
%     \item
%       implicit data (registers and other 2e/expl3 data stores) that
%       the socket is allowed to make use of, with a documented description
%       of what they contain (if relevant for the task at hand---no need to
%       describe the whole \LaTeX{} universe)
% 
%     \item
%        information about the state of the \TeX{} engine (again when
%       relevant), e.g. is called in mmode or vmode or in the output routine or \ldots
%       
%     \item
%       \ldots\ \empty{anything missing?}
%     \end{itemize}
% 
%       
% \item
%     A socket has documented results/outputs which can be
% 
%     \begin{itemize}
%     \item
%        what kind of data it should write to the current list (if that
%        is part of its task)
% 
%     \item
%        what kind of registers and other 2e/expl3 data stores it should
%        modify and in what way
% 
%     \item
%        what kind of state changes it should do (if any)
% 
%     \item
%        \emph{\ldots\ anything else?}
%     \end{itemize}
% 
% \item
%
%   At any time a socket has one block of code (a plug \iffalse(\fi:-)\,)
%   associated with it. Such code is itself named and the association
%   is done by linking the socket name to the code name (putting a
%   plug into the socket).
% 
% \item
%
%   The name of a plug consists of ASCII letters \texttt{[a-z]},
%   \texttt{[A-Z]}, \texttt{[0-9]}, \texttt{[-/@]} only.
%
% \item
%   Socket plug names have to be unique within on a per socket basis,
%   but it is perfectly allowed (and sensible in some cases) to use
%   the same plug name with different sockets (where based on the
%   sockets' purposes, different actions may be associated with the plug
%   name). For example \plug{noop} is a plug name declared for every
%   socket, yet it action \enquote{grab the socket inputs and throw
%   them away} obviously differs depending on how many inputs the
%   socket has.
%
% \item
%
%   When declaring a plug it is stated for which socket it is meant
%   (i.e., its code can only be used with that socket). This means
%   that the same plug name can be used with different sockets
%   referring to different code in each case.
% 
% \item
%   Configuration of  a socket can only be done by
%   linking different code to it. Nevertheless the code linked to it can
%   provide its own means of configuration (but this is outside of the
%   spec).
% 
% \item
%   Technically execution of a socket (\cs{UseSocket}) involves
% 
%     \begin{itemize}
%     \item
%        doing any house keeping (like writing debugging info, \ldots);
% 
%     \item
%        looking up the current code association (what plug is in the socket);
%     
%     \item
%       executing this code which will pick up the mandatory arguments
%       (happens at this point, not
%        before), i.e., it is like calling a csname defined with
% \begin{verbatim}      
%   \def\foo#1#2...{...#1...#2...}
% \end{verbatim}
% 
%     \item
%       do some further house keeping (if needed).
%     \end{itemize}
% 
% \item
%   A socket is typically only used in one place in code, but this is not
%   a requirement, i.e., if the same operation with the same inputs need
%   to be carried out in several places the same named socket can be used.
% 
% \end{itemize}
% 
% 
%
%
%
% \subsubsection{Command syntax}
%
% We give both the \LaTeXe{} and the L3 programming layer command names.
% 
% \begin{function}{\NewSocket,\socket_new:nn}
% \begin{syntax}
% \cs{NewSocket}    \Arg{socket-name}\Arg{number-of-inputs}
% \cs{socket_new:nn}\Arg{socket-name}\Arg{number-of-inputs}
% \end{syntax}
%   Declares a new socket with name \meta{socket-name} having
%   \meta{number-of-inputs} inputs. There is automatically a
%   plug \plug{noop} declared for it, which does nothing, i.e., it gobbles
%   the socket inputs (if any). This is made the default plug except for
%   sockets with one input which additionally define the plug
%   \plug{identity} and assign that as their default.
%
%   This \plug{identity} plug simply returns the socket input without
%   its outer braces.
%   The use case for this plug are situations like this:
%   \begin{quote}
%      \cs{UseSocket}\verb={tagsupport/footnote}=\Arg{code}
%   \end{quote}
%   If tagging is not active and the socket contains the plug \plug{identity}
%   then this returns \meta{code} without the outer braces and to
%   activate tagging all that is necessary is to change the plug to
%   say \plug{tagpdf} so
%   that it surrounds \meta{code} by some tagging magic.
%   This is the most common use case for sockets with one input, which
%   is why they have this special default.
%
%   The socket documentation should describe its purpose, its inputs and the
%   expected results as discussed above.
%
%   The declaration is only allowed at top-level, i.e., not inside a group.
% \end{function}
%
% 
% \begin{function}{\NewSocketPlug,\socket_new_plug:nnn,\socket_set_plug:nnn}
% \begin{syntax}
% \cs{NewSocketPlug}      \Arg{socket-name}\Arg{socket-plug-name}\Arg{code}
% \cs{socket_new_plug:nnn}\Arg{socket-name}\Arg{socket-plug-name}\Arg{code}
% \cs{socket_set_plug:nnn}\Arg{socket-name}\Arg{socket-plug-name}\Arg{code}
% \end{syntax}
%   Declares a new plug for socket \meta{socket-name} that runs
%   \meta{code} when executing. It complains if the plug was already
%   declared previously.
%
%   The form \cs{socket_set_plug:nnn} changes an existing plug. As
%   this should normally not be necessary, we currently have only an L3
%   layer name for the few cases it might be useful.
%
%   The declarations can be made inside a group and obey scope, i.e.,
%   they vanish if the group ends.
% \end{function}
%
% 
% \begin{function}{\AssignSocketPlug,\socket_assign_plug:nn}
% \begin{syntax}
% \cs{AssignSocketPlug}     \Arg{socket-name}\Arg{socket-plug-name}
% \cs{socket_assign_plug:nn}\Arg{socket-name}\Arg{socket-plug-name}
% \end{syntax}
%   Assigns the plug \meta{socket-plug-name} to the socket
%   \meta{socket-name}. It errors if either socket or plug is not
%   defined.
%
%   The assignment is local, i.e., it obeys scope.
% \end{function}
%
% 
% \begin{function}{\UseSocket,\socket_use:nw,\socket_use:n,\socket_use:nn,\socket_use:nnn,\socket_use:nnnn}
% \begin{syntax}
% \cs{UseSocket}     \Arg{socket-name}
% \cs{socket_use:nnn}\Arg{socket-name} \Arg{socket-arg\textsubscript{1}} \Arg{socket-arg\textsubscript{2}}
% \end{syntax}
%   Executes the socket \meta{socket-name} by retrieving the
%   \meta{code} of the current plug assigned to the socket. This is
%   the only command that would appear inside macro code in packages.
%
%   For performance reasons there is no explicit check that the socket
%   was declared!
%
%   The different L3 programming layer commands are really doing the
%   same thing: they grab as many arguments as defined as inputs for the socket
%   and then pass them to the plug. The different names are only there
%   to make the code more readable, i.e., to indicate how many
%   arguments are grabbed in total (note that no runtime check is made to
%   verify that this is actually true).  We only provide them for
%   sockets with up to 3 inputs (most likely those with zero or one input would
%   have been sufficient). If you happen to have a socket with more
%   inputs, use \cs{socket_use:nw}.
% \end{function}
%
% 
% \begin{function}{\ShowSocket,\LogSocket,\socket_show:n,\socket_log:n}
% \begin{syntax}
% \cs{ShowSocket}   \Arg{socket-name}
% \cs{socket_show:n}\Arg{socket-name}
% \end{syntax}
%   Displays information about the socket \meta{socket-name} and its
%   state then stops and waits for further instructions --- at the
%   moment some what rudimentary.
%
%   \cs{LogSocket} and \cs{socket_log:n} only differ in that they don't stop.
% \end{function}
%
% 
% \begin{function}{\DebugSocketsOn,\DebugSocketsOff,
%                  \socket_debug_on:,\socket_debug_off:}
% \begin{syntax}
% \cs{DebugSocketsOn} \ldots\ \cs{DebugSocketsOff}
% \end{syntax}
%   Turns debugging of sockets on or off.
% \end{function}
%
% \subsubsection{Rationale for error handling}
% 
% The errors during the declarations are produced to help with
% typos---after all, such declarations might be part of a document
% preamble (not that likely, but possible). However, \cs{UseSocket} is
% not doing much checking, e.g.,
% \begin{verbatim}
%   \UseSocket{mispelled-socket}{hello}{world}
% \end{verbatim}
% will generate a rather low-level error and then typesets
% ``{hello}{world}'' because there is no dedicated runtime check if
% \texttt{mispelled-socket} is a known socket.
% 
% The reason is that if the misspelling is in the code, then this is a
% programming error in the package and for speed reasons \LaTeX{} does
% not repeately make runtime checks for coding errors unless they can or
% are likely to be user introduced.
% 
%
% \MaybeStop{\setlength\IndexMin{200pt}  \PrintIndex  }
%
%
% \section{The Implementation}
%
%    The implementation of the socket mechanism should be (partially)
%    redone and we should probably store the different code chunks in
%    a property list so that we can have a decent \cs{ShowSocket}
%    command that shows the available alternatives.\fmi{implement?}
%
%    \begin{macrocode}
%<*2ekernel|latexrelease>
%    \end{macrocode}
%
%    \begin{macrocode}
\ExplSyntaxOn
%    \end{macrocode}
%
%    \begin{macrocode}
%<@@=socket>
%    \end{macrocode}
%
%    \begin{macrocode}
%<latexrelease>\NewModuleRelease{2023/11/01}{ltsockets}
%<latexrelease>                 {The~socket~management~system}
%    \end{macrocode}
%
%
%
% \subsection{Debugging the socket structures}
%
%  Code and commands in this section are not final, it needs more
%  experimentation to see what kind of tracing information is going to
%  be useful in practice. For now the tracing is mainly meant to be used
%  for code testing and not so much for application testing.
%
%  It is quite likely that the
%  commands and the behavior of the tracing might change in the
%  future once we gained some experience with it.
%
%  \begin{macro}{\g_@@_debug_bool}
%    Holds the current debugging state.
%    \begin{macrocode}
\bool_new:N \g_@@_debug_bool
%    \end{macrocode}
%  \end{macro}
%
%  \begin{macro}{\socket_debug_on:,\socket_debug_off:}
%  \begin{macro}{\@@_debug:n, \@@_debug_term:n}
%  \begin{macro}{\@@_debug_gset:}
%    Turns debugging on and off by redefining \cs{@@_debug:n} and
%    \cs{@@_debug_term:n}. By default they do nothing.
%    \begin{macrocode}
\cs_new_eq:NN \@@_debug:n \use_none:n
\cs_new_eq:NN \@@_debug_term:n \use_none:n
%    \end{macrocode}
%
%    \begin{macrocode}
\cs_new_protected:Npn \socket_debug_on:
  {
    \bool_gset_true:N \g_@@_debug_bool
    \@@_debug_gset:
  }
\cs_new_protected:Npn \socket_debug_off:
  {
    \bool_gset_false:N \g_@@_debug_bool
    \@@_debug_gset:
  }
\cs_new_protected:Npn \@@_debug_gset:
  {
    \cs_gset_protected:Npx \@@_debug:n ##1
      { \bool_if:NT \g_@@_debug_bool {##1} }
    \cs_gset_protected:Npx \@@_debug_term:n ##1
      { \bool_if:NT \g_@@_debug_bool
        { \iow_term:x { ^^J [Sockets]~ ==>~ ##1} } }
  }
%    \end{macrocode}
%  \end{macro}
%  \end{macro}
%  \end{macro}
%
%
% \subsection{The L3 layer commands}
%
%  \begin{macro}{\socket_new:nn}
%    
%    Declaring a socket creates a str to hold the name (a pointer) to the
%    code that should be used when the socket is executed, and an integer to
%    hold the number of inputs of that socket.  Initially, an
%    ``empty'' code chunk is created and assigned so the socket
%    does nothing by default other than swallowing its inputs (if any).
%    \begin{macrocode}
\cs_new_protected:Npn \socket_new:nn #1 #2 {
  \str_if_exist:cTF { l_@@_#1_plug_str }
      {
        \msg_error:nnn { socket } { already-declared } {#1}
      }
      {
%    \end{macrocode}
%    We only support declarations on top-level.
%    \begin{macrocode}
        \int_if_zero:nTF \tex_currentgrouplevel:D
          {
            \str_new:c { l_@@_#1_plug_str }
            \seq_new:c { l_@@_#1_plugs_seq }
            \int_const:cn { c_@@_#1_args_int } {#2}
            \socket_new_plug:nnn {#1} { noop } {}
            \int_compare:nNnTF {#2} = 1
              {
                \socket_new_plug:nnn {#1} { identity } {##1}
                \socket_assign_plug:nn {#1} { identity }
              }
              { \socket_assign_plug:nn {#1} { noop } }
            \@@_debug_term:n
              { Socket~ '#1'~ declared~ with~ #2~ input(s) }
          }
          {
            \msg_error:nn { socket } { not-top-level }
          }
      }
}
%    \end{macrocode}
%  \end{macro}
%  
%  
%  \begin{macro}{\socket_log:n,\socket_show:n}
%    Show the current state of the socket --- for now this is just a
%    quick draft and should be redone and extended.
%    \begin{macrocode}
\cs_new_protected:Npn \socket_log:n #1 {
  \typeout{ Socket~ #1:}
  \str_if_exist:cTF { l_@@_#1_plug_str }
      {
        \typeout{ \@spaces number~ of~ inputs~ =~
                  \int_use:c { c_@@_#1_args_int } }
        \typeout{ \@spaces available~plugs~ =~
                  \seq_use:cnnn { l_@@_#1_plugs_seq }{,~}{,~}{,~} }
        \typeout{ \@spaces current~ plug~ =~
                  \str_use:c { l_@@_#1_plug_str } }
        \typeout{ \@spaces definition~ =~
                  \cs_meaning:c
                    { @@_#1_plug_ \str_use:c { l_@@_#1_plug_str } :w } }
        \typeout{}
      }
      {
%    \end{macrocode}
%    If we are showing a socket it is not an error if it doesn't exist.
%    \begin{macrocode}
        \typeout { Socket~ is~ not~ declared! }
      }
}
%    \end{macrocode}
%    And here the version that stops:
%    \begin{macrocode}
\cs_new_protected:Npn \socket_show:n #1 {\socket_log:n {#1} \errmessage{}}
%    \end{macrocode}
%  \end{macro}
%  
%  
%  
%  
%  
%  \begin{macro}{\socket_new_plug:nnn,\socket_set_plug:nnn}
%    
%    Declaring a code for a socket is just making a definition, taking
%    the number of arguments from the saved int.
%    \begin{macrocode}
\cs_new_protected:Npn \socket_new_plug:nnn #1#2#3 {
  \str_if_exist:cTF { l_@@_#1_plug_str }
      {
        \cs_if_exist:cTF { @@_#1_plug_#2:w }
          {
            \msg_error:nnnn { socket } { plug-already-declared } {#1} {#2}
          }
          {
            \cs_generate_from_arg_count:cNnn
              { @@_#1_plug_#2:w }
              \cs_new_protected:Npn
              { \int_use:c { c_@@_#1_args_int } }
              {#3}
%    \end{macrocode}
%    This is a new declaration so we add the name to a seq for the
%    debugging info.
%    \begin{macrocode}
            \seq_put_right:cn { l_@@_#1_plugs_seq } {#2}
            \@@_debug_term:n
              { Plug~ '#2'~ for~ socket~ '#1'~ declared. }
          }
      }
      {
        \msg_error:nnn { socket } { undeclared } {#1}
      }
}
%    \end{macrocode}
%    Changing the plug of an existing socket is rather similar, except
%    that we don't have to deal with adding it to the debugging
%    sequence.
%    \begin{macrocode}
\cs_new_protected:Npn \socket_set_plug:nnn #1#2#3 {
  \str_if_exist:cTF { l_@@_#1_plug_str }
      {
        \cs_if_exist:cTF { @@_#1_plug_#2:w }
          {
            \cs_generate_from_arg_count:cNnn
              { @@_#1_plug_#2:w }
              \cs_set_protected:Npn
              { \int_use:c { c_@@_#1_args_int } }
              {#3}
            \@@_debug_term:n
              {  Plug~ '#2'~ for~ socket~ '#1'~ changed. }
          }
          {
            \msg_error:nnnn { socket } { plug-undeclared } {#1} {#2}
          }
      }
      {
        \msg_error:nnn { socket } { undeclared } {#1}
      }
}
%    \end{macrocode}
%  \end{macro}
%  
%  
%  
%  \begin{macro}{\socket_assign_plug:nn}
%    
%    Assigning a plug to a socket just changes the name in
%    the socket string. The assignment is local to the current group.
%    \begin{macrocode}
\cs_new_protected:Npn \socket_assign_plug:nn #1 #2 {
  \str_if_exist:cTF { l_@@_#1_plug_str }
      {
        \cs_if_exist:cTF { @@_#1_plug_#2:w }
          {
            \@@_debug_term:n
              { Replacing~ plug~ '\str_use:c { l_@@_#1_plug_str }'~
                with~ '#2'~ in~ socket~ '#1'. }
            \str_set:cn { l_@@_#1_plug_str } {#2}
          }
          {
            \msg_error:nnnn { socket } { plug-undeclared } {#1} {#2}
          }
      }
      { \msg_error:nnn { socket } { undeclared } {#1} }
}
%    \end{macrocode}
%  \end{macro}
%  
%  
%  \begin{macro}{\socket_use:nw,\socket_use:n,\socket_use:nn,\socket_use:nnn,\socket_use:nnnn}
%    
%    And using it is more or less a \cs{use:c} so very lightweight. We do not
%    add a runtime check for speed reasons!
%
%    This command is named
%    \cs{socket_use:nw} because we don't know how many inputs the
%    socket has until we have looked at the socket name (in argument \verb=#1=).
%    But, of course, the developer knows so we also offer a few aliases
%    \cs{socket_use:nn}, etc.\ so that one can indicate the correct number of
%    arguments (socket inputs plus one) in the L3 layer code.
%
%    \begin{macrocode}
\cs_new_protected:Npn \socket_use:nw #1 {
  \@@_debug_term:n
    { Socket~ '#1'~ containing~ plug~
      '\str_use:c { l_@@_#1_plug_str }'~ used. }
  \use:c { @@_#1_plug_ \str_use:c { l_@@_#1_plug_str } :w }
}
%    \end{macrocode}
%    To make code a bit more readable we also define functions that
%    indicate how many arguments are picked up. However, this is just for
%    code documentation: internally they all do the same and the number
%    of arguments isn't checked by default.
%    \begin{macrocode}
\cs_new_eq:NN  \socket_use:n \socket_use:nw     % socket with no inputs
\cs_new_eq:NN  \socket_use:nn \socket_use:nw    % socket with one input
\cs_new_eq:NN  \socket_use:nnn \socket_use:nw   % socket with two inputs
\cs_new_eq:NN  \socket_use:nnnn \socket_use:nw  % socket with three inputs
%    \end{macrocode}
%    The above commands could be changed to check  how many inputs
%    the socket is declared with (for example, when checking is in
%    force).
%    \fmi{Implement?} 
%  \end{macro}
%
%
%
%
%
%
%  
% \subsection{Error messages}
%

%    
%    \begin{macrocode}
\msg_new:nnnn { socket } { already-declared }
    { Socket~ '#1'~ already~ declared! }
    { A~ socket~ can~ only~ be~ declared~ once.~ The~ name~ '#1'~ is~
      already~ taken.~ Use~ \ShowSocket{#1}~ to~ see~ its~ definition. }
      
\msg_new:nnnn { socket } { undeclared }
    { Socket~ '#1'~ undeclared! }
    { You~ tried~ to~ use~ a~ socket~ that~ was~ not~ declared~ before. }

\msg_new:nnnn { socket } { not-top-level }
    { Sockets~ can~ only~ be~ declared~ at~ top-level! }
    { It~ is~ not~ allowed~ to~ declare~ sockets~ inside~ a~
      group.~ Move~ the~ declaration~ to~ the~ top-level. }
%    \end{macrocode}
%
%    \begin{macrocode}
\msg_new:nnnn { socket } { plug-already-declared }
    { Plug~ '#2'~ for~ socket~ '#1'~ already~ declared! }
    { You~ can't~ change~ an~ existing~ plug~ with~ \NewSocketPlug~ and~ it~
      is~ normally~ not~ sensible~ to~ do~ so.~ Use~ the~ L3~ programming~
      layer~ function~ \socket_set_plug:nnn~ if~ you~ really~ have~ to. }
    
\msg_new:nnnn { socket } { plug-undeclared }
    { Plug~ '#2'~ for~ socket~ '#1'~ undeclared! }
    { The~ plug~ name~ is~ unknown.~ Is~ the~ name~ misspelled~ or~ did~ you~
      intend~ to~ assign~ it~ to~ a~ different~ socket? }
%    \end{macrocode}
%    
%    \begin{macrocode}
\prop_gput:Nnn \g_msg_module_type_prop { socket } { LaTeX }
%    \end{macrocode}
%
%
%
%  
% \subsection{The \LaTeXe{} interface commands}
%
%  \begin{macro}{\NewSocket,\NewSocketPlug,
%                \ShowSocket,\LogSocket,
%                \AssignSocketPlug,\UseSocket,
%                \DebugSocketsOn,\DebugSocketsOff}
%    As we expect that there are existing \LaTeXe{} packages that may
%   want to make use of the socket mechanism, we provide 2e names for
%   most of the commands.
%    \begin{macrocode}
\cs_new_eq:NN \NewSocket         \socket_new:nn 
\cs_new_eq:NN \ShowSocket        \socket_show:n 
\cs_new_eq:NN \LogSocket         \socket_log:n 
%    \end{macrocode}
%
%    \begin{macrocode}
\cs_new_eq:NN \NewSocketPlug     \socket_new_plug:nnn
\cs_new_eq:NN \AssignSocketPlug  \socket_assign_plug:nn
\cs_new_eq:NN \UseSocket         \socket_use:nw
%    \end{macrocode}
%
%    \begin{macrocode}
\cs_new_eq:NN \DebugSocketsOn    \socket_debug_on:
\cs_new_eq:NN \DebugSocketsOff   \socket_debug_off:
%    \end{macrocode}
%  \end{macro}
%
%
%  
%    \begin{macrocode}
%
%<latexrelease>\IncludeInRelease{0000/00/00}{ltsockets}
%<latexrelease>                 {The~socket~management~(undo)}%
%<latexrelease>
%<latexrelease>\let \NewSocket  \@undefined
%<latexrelease>\let \ShowSocket \@undefined
%<latexrelease>\let \LogSocket  \@undefined
%<latexrelease>
%<latexrelease>\let \NewSocketPlug    \@undefined
%<latexrelease>\let \AssignSocketPlug \@undefined
%<latexrelease>\let \UseSocket        \@undefined
%<latexrelease>
%<latexrelease>\let \DebugSocketsOn  \@undefined
%<latexrelease>\let \DebugSocketsOff \@undefined
%<latexrelease>
%<latexrelease>\EndModuleRelease
%    \end{macrocode}
%
%    \begin{macrocode}
\ExplSyntaxOff
%    \end{macrocode}
%
%    \begin{macrocode}
%</2ekernel|latexrelease>
%    \end{macrocode}
%
%    Reset module prefix:
%    \begin{macrocode}
%<@@=>
%    \end{macrocode}
%
%
% \Finale
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\endinput



^^A  Needed for emacs
^^A
^^A  Local Variables: 
^^A  mode: latex
^^A  coding: utf-8-unix
^^A  End: