DataMuseum.dk

top - metrics - download

    Length: 4842 (0x12ea)
    Types: TextFile
    Names: »tailor.tex«

└─⟦2d1937cfd⟧ Bits:30007241 EUUGD22: P.P 5.0
    └─⟦dc59850a2⟧ »EurOpenD22/pp5.0/pp-5.tar.Z« 
        └─⟦e5a54fb17⟧ 
            └─⟦this⟧ »pp-5.0/doc/tailor/tailor.tex« 

TextFile

\documentstyle [11pt,a4] {article}
\author {S.E. Kille}
\date {\today}
\title {A system for tailoring C programs}


\begin {document}
\maketitle

\begin {abstract}
This document defines a proposed design for tailoring C programs.
It is specifically intended for ISODE, QUIPU, and PP.
\end {abstract}


\section {Introduction}


It seems wrong to require a C format, a tailor format, and hand generated
code to map between them.  I propose that we devise a format which specifies
both, and can be used to generate what is needed. 


We need to be able to specify the following:

\begin {itemize}
\item
NUMBER types, which will be mapped onto int

\item STRING types, which will be mapped onto char*

\item ENUMERATED types, which will be mapped onto int

\item FLAG types, which will be mapped onto int

\item LIST types. These are structure types, which are linked into a list.
Thus, a "next" and a "key" is always generated.  
Can contain STRING, ENUMERATED or FLAG.
and also DEF, which points to a previously defined type (typically to get at 
taiored defines for ENUMERATED, or to point at identified LIST types)

\end {itemize}

\section {BNF Defintion}

The following BNF for the single file is proposed.  Tokens are separated by
LWSP or braces.

\begin {verbatim}

<tailor-definition> ::= <preamble> <CRLF> <line-list>
<lines> ::= <line> | <line> <CRLF> <line-list>
<line> ::= "--" <global-comment> 
	| "#" <local-comment>
	| <item>

<preamble> ::= "PREFIX" <word>

<item> ::= <number> | <enum> | <flag> | <string> | <list> |
		<enumvalues> | <flagvalues>

<optwordlist> ::= <optword> | <optword> "," <optwordlist>
<optword> ::= <word> ["(" <word> ")"]		

<number> ::= "NUMBER" <optword>

<enum> ::= "ENUM" <optword> <word>

<flag> ::= "FLAG" <optword> <word>

<string> ::= "STRING" <optword>

<enumvalues> ::= "ENUMVALUES" <word> <optwordlist>

<flagvalues> ::= "FLAGVALUES" <word> <optwordlist>

<list> ::= "LIST" <optword> "{" <listitemlist> "}"

<listitemlist> ::= <listitem> | <listitem> <CRLF> <listitemlist>

<listitem> ::= <number> | <enum> | <flag> | <string> | <ptr>

<ptr> ::= "PTR" <optword> <word>
\end{verbatim}

\section {Behaviour}


\begin {itemize}
\item
An h file is made with 
Typedefs for any structures generated, and 
extern references to all variables.


\item
A C file with variable definitions, initialiased to zero


\item
Documentation on the format of the tailor file.  This should be unix manual
or LaTeX.  {\tt <global-comment>} is interspersed into the mechanically
generated structure.

\item A routine {\tt <prefix>\_tai()} to parse a tailor file and initialise
each variables.

\item {\tt <prefix>\_tai\_build()} to take 
tailor file, and generate a file of static definitios
(esssentially to replace 3., with values frozen in).  This would allow a
site to ``compile in'' an operational tailor file, and thus gain performance.
Only deltas would need to be runtime tailored.  

\item Code to print/log the value of tailored variables ({\tt <prefix>\_<item>\_print()}).

\item Code to call all of the functions in 7 (list-config) - a sort of
print\_config ({\tt <prefix>\_tai\_print()}).

\item  
Code to select structures from lists 
 ({\tt  <item>\_nm2struct()) }).

\end {itemize}

Some more notes on building the structures:

\begin {itemize}
\item For variables or structure items, the name of the variable is used, or
the bracketed value if present.

\item For ENUMVALUES, the definitions for ENUM are associated with the
strings concerned.  If bracketed values are present, these are references to
external \#defines.  Otherwise, {\tt <prefix>\_<value>} is allocated
incrementally, and the \#defines created.
Similarly for ENUMFLAGS.  
\end {itemize}


\section {Examples}

Example format

\begin {verbatim}
	STRING authwarn
\end{verbatim}


generates char* authwarn; and recognises 'authwarn "joe foobar"' in a tailor
file.

\begin {verbatim}
	NUMBER mcount(maxcount)
\end{verbatim}

generates 'int maxcount;' and recognises 'mcount 22'

\begin {verbatim}

ENUMVALUES logvalues bst(LLOG_BST), fst(LLOG_FST)

log LIST {
    STRING show (ll_show) 
    ENUM level(ll_level) logvalues
    }
    
chan(ch_struct) LIST {
    STRING show(ch_show) 
    PTR log(ch_log) log
    ENUM loglevel(ch_level) logvalues
    }


generates strucutures:

log {
   struct *log log_next;
   char *log_key;
   char* ll_show;
   char ll_level;
};

ch_struct {
   struct *ch_struct ch_struct_next;
   char *ch_struct_key;
   char *ch_show;
   struct *log ch_log;
   char ch_level;
}
\end{verbatim}

It generates routines chan\_nm2struct (key) and log\_nm2struct (key).
One might initialise logs by (e.g.) authlog = log\_nm2struct ("auth");

It would recognise formats:

\begin {verbatim}
log auth show="authorisation log", level=bst

chan xchan show="funny channel", log=auth, level=fst
\end{verbatim}


\end {document}