alarsyo/spot
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

Fix the associativity of ->, <->, U, R, W, and M wrt the PSL standard.

Browse source 
* src/ltlparse/ltlparse.yy: Make all the above operators
right-associative.  Also let `:' have precedence over `;'.
* src/ltltest/reduccmp.test: Adjust for the `:' precedence.
* doc/tl/tl.tex, NEWS: Document this.
This commit is contained in:
Alexandre Duret-Lutz 2012-04-28 16:13:24 +02:00
parent 807dcefba4
commit ce437cd499
4 changed files with 43 additions and 45 deletions
Download patch file Download diff file Expand all files Collapse all files

65
doc/tl/tl.tex
View file

@ -873,51 +873,46 @@ operator, even if the operator has multiple synonyms (like \samp{|},
\section{Operator precedence}
\spottodo[inline]{The following operator precedence describe the
current parser of Spot, but this is not what we want eventually.
For instance $\IMPLIES$ should be right associative. $\U$ and $\W$
would probably make more sense as right associative operators too.\\
%
Tools that have $\U$, $\R$, $\W$, and $\M$ operators as left
associative: Spin, ltl2ba (same parser as spin), Wring, psl2ba,
Spot, Modella, NuSMV.\\
%
Tools (and doc) that have these operators as right associative: Goal
(hence Büchi Store), PSL reference manual, LTL2AUT, LTL2Büchi (from
JavaPathFinder).\\
%
Tools that have these operators as non-associative (parentheses are
required): Vis, LBTT.\\
%
While compiling these lists I have also discovered that not all
people aggreed on the associativity of $\IMPLIES$ and $\EQUIV$.
Some have both left-assoc, or both right-assoc, other have only
$\IMPLIES$ as right-assoc.\\
%
We want to get closer to the PSL standard eventually.}
The following operator precedence describes the current parser of
Spot. It has not always been this way. Especially, all operators
were left associative until version 0.9, when we changed the
associativity of $\IMPLIES$, $\EQUIV$, $\U$, $\R$, $\W$, and $\M$ to get closer
to the PSL standard~\cite{psl.04.lrm,eisner.06.psl}.
\begin{center}
\begin{tabular}{clc}
assoc. & operators & priority \\
assoc. & operators & priority \\
\midrule
left & $\Asuffix,\,\AsuffixEQ,\,\Esuffix,\,\EsuffixEQ$ & \tikz[remember picture,baseline]\node (lowest){lowest}; \\
left & $\CONCAT,\,\FUSION$ & \\
left & $\IMPLIES,\,\EQUIV$ & \\
left & $\XOR$ & \\
left & $\OR$ & \\
left & $\AND,\,\ANDALT$ & \\
left & $\U,\,\W,\,\M,\,\R$ & \\
& $\F,\,\G$ & \\
& $\X$ & \\
& $\STAR{\mvar{i}..\mvar{j}},\,\PLUS,\,\EQUAL{\mvar{i}..\mvar{j}},\,\GOTO{\mvar{i}..\mvar{j}}$ & \\
& $\NOT$ & \\
& $\code{=0},\,\code{=1}$ & \tikz[remember picture,baseline]\node (highest){highest}; \\
right & $\Asuffix,\,\AsuffixEQ,\,\Esuffix,\,\EsuffixEQ$ & \tikz[remember picture,baseline]\node (lowest){lowest}; \\
left & $\CONCAT$ & \\
left & $\FUSION$ & \\
right & $\IMPLIES,\,\EQUIV$ & \\
left & $\XOR$ & \\
left & $\OR$ & \\
left & $\AND,\,\ANDALT$ & \\
right & $\U,\,\W,\,\M,\,\R$ & \\
& $\F,\,\G$ & \\
& $\X$ & \\
& $\STAR{\mvar{i}..\mvar{j}},\,\PLUS,\,\EQUAL{\mvar{i}..\mvar{j}},\,\GOTO{\mvar{i}..\mvar{j}}$ & \\
& $\NOT$ & \\
& $\code{=0},\,\code{=1}$ & \tikz[remember picture,baseline]\node (highest){highest}; \\
\end{tabular}
\end{center}
\begin{tikzpicture}[remember picture,overlay,>=stealth',semithick]
\draw[->] (lowest) -- (highest);
\end{tikzpicture}
Beware that not all tools agree on the associativity of these
operators. For instance Spin, ltl2ba (same parser as spin), Wring,
psl2ba, Modella, and NuSMV all have $\U$ and $\R$ as left-associative,
while Goal (hence Büchi store), LTL2AUT, and LTL2Büchi (from
JavaPathFinder) have $\U$ and $\R$ as right-associative. Vis and LBTT
have these two operators as non-associative (parentheses required).
Similarly the tools do not aggree on the associativity of $\IMPLIES$
and $\EQUIV$: some tools handle both operators as left-associative, or
both right-associative, other have only $\IMPLIES$ as right-associative.
\chapter{Properties}
When Spot builds a formula (represented by an AST with shared