(*<*)
theory Slides2
imports "LaTeXsugar" "Nominal"
begin
notation (latex output)
set ("_") and
Cons ("_::/_" [66,65] 65)
(*>*)
text_raw {*
\renewcommand{\slidecaption}{Edinburgh, 11.~July 2010}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1>[t]
\frametitle{%
\begin{tabular}{@ {\hspace{-3mm}}c@ {}}
\\
\LARGE Proof Pearl:\\[-0mm]
\LARGE A New Foundation for\\[-2mm]
\LARGE Nominal Isabelle\\[12mm]
\end{tabular}}
\begin{center}
Brian Huf\!fman and {\bf Christian Urban}\\[0mm]
\end{center}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-2>[c]
\frametitle{Nominal Isabelle}
\begin{itemize}
\item \ldots is a definitional extension of Isabelle/HOL
(let-polymorphism and type classes)\medskip
\item \ldots provides a convenient reasoning infrastructure for
terms involving binders (e.g.~lambda calculus, variable convention)\medskip
\item<2-> \ldots mainly used to find errors in my own
(published) paper proofs and in that of others \alert{\bf ;o)}
\end{itemize}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-3>[c]
\frametitle{Nominal Theory}
by Pitts; at its core are:\bigskip
\begin{itemize}
\item sorted atoms and
\item sort-respecting permutations
\end{itemize}
\onslide<2->{
\begin{textblock}{8}(4,11)
\onslide<3->{$\text{inv\_of\_}\pi \,\act\, ($}\onslide<2->{$\pi \,\act\, x$}%
\onslide<3->{$) \,=\, x$}
\end{textblock}}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-4>[t]
\frametitle{The ``Old Way''}
\begin{itemize}
\item sorted atoms\\ \alert{$\quad\mapsto$ separate types}\; (``copies'' of nat)\bigskip
\item sort-respecting permutations\\ \alert{$\quad\mapsto$ lists of pairs of atoms (list swappings)}
\onslide<2->{
\begin{center}
\begin{tabular}{c@ {\hspace{7mm}}c}
$\text{[]} \,\act\, c = c$ &
$\swap{a}{b}\!::\!\pi \,\act\, c =
\begin{cases}
b & \text{if}\, \pi\act c = a\\
a & \text{if}\, \pi\act c = b\\
\pi\act c & \text{otherwise}
\end{cases}$
\end{tabular}
\end{center}}
\end{itemize}
\only<3>{
\begin{textblock}{14}(1,12.5)
\alert{The big benefit:} the type system takes care of the sort-respecting part.
\end{textblock}}
\only<4>{
\begin{textblock}{14}(1,12.5)
\alert{A small benefit:} permutation composition is \alert{list append}
and permutation inversion is \alert{list reversal}.
\end{textblock}
}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-4>
\frametitle{Problems}
\begin{itemize}
\item @{text "_ \<bullet> _ :: \<alpha> perm \<Rightarrow> \<beta> \<Rightarrow> \<beta>"}\bigskip
\item @{text "supp _ :: \<beta> \<Rightarrow> \<alpha> set"}
\begin{center}
$\text{finite} (\text{supp}\;x)_{\,\alpha_1\,\text{set}}$ \ldots
$\text{finite} (\text{supp}\;x)_{\,\alpha_n\,\text{set}}$
\end{center}\bigskip
\item $\forall \pi_{\alpha_1} \ldots \pi_{\alpha_n}\;.\; P$\bigskip
\item type-classes
\onslide<3>{\small\hspace{5mm}can only have \alert{\bf one} type parameter}
\begin{itemize}
\item<2-> $\text{[]}\,\act\, x = x$
\item<2-> $(\pi_1 @ \pi_2) \,\act \, x = \pi_1 \,\act\,(\pi_2 \,\act\, x)$
\item<2-> if $\pi_1 \sim \pi_2$ then $\pi_1 \,\act\, x = \pi_2 \,\act\, x$
\end{itemize}
\end{itemize}
\only<4->{
\begin{textblock}{9}(3,7)\begin{tikzpicture}
\draw (0,0) node[inner sep=3mm,fill=cream, ultra thick, draw=red, rounded corners=2mm]
{\normalsize\color{darkgray}
\begin{quote}
\begin{itemize}
\item \alert{lots} of ML-code
\item \alert{not} pretty
\item \alert{not a proof pearl} :o(
\end{itemize}
\end{quote}};
\end{tikzpicture}
\end{textblock}}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-5>
\frametitle{A Better Way}
*}
datatype atom = Atom string nat
text_raw {*
\mbox{}\bigskip
\begin{itemize}
\item<3-> permutations are (restricted) bijective functions from @{text "atom \<Rightarrow> atom"}
\begin{itemize}
\item sort-respecting \hspace{5mm}($\,\forall a.\;\text{sort}(\pi a) = \text{sort}(a)$)
\item finite domain \hspace{5mm}($\text{finite} \{a.\;\pi a \not= a\}$)
\end{itemize}\medskip
\item<4-> \alert{What about swappings?}
\small
\[
\begin{array}{l@ {\hspace{1mm}}l}
(a\;b) \dn & \text{if}\;\text{sort}(a) = \text{sort}(b)\\
& \text{then}\;\lambda c. \text{if}\;a = c\;\text{then}\;b\;\text{else}\;
\text{if}\;b = c\;\text{then}\;a\;\text{else}\;c\\
& \text{else}\;\only<4>{\raisebox{-5mm}{\textcolor{red}{\huge\bf $?$}}}
\only<5->{\text{\alert{\bf id}}}
\end{array}
\]
\end{itemize}
\only<3>{
\begin{textblock}{7}(4,11)
@{text "\<pi> \<bullet> _ :: perm \<Rightarrow> \<beta> \<Rightarrow> \<beta>"}
\end{textblock}}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-7>
\frametitle{A Smoother Nominal Theory}
From there it is essentially plain sailing:\bigskip
\begin{itemize}
\item<2-> $(a\;b) = (b\;a) \onslide<4->{= (a\;c) + (b\;c) + (a\;c)}$\bigskip
\item<3-> permutations are an instance of Isabelle's\\
group\_add ($0$, $\pi_1 + \pi_2$, $- \pi$)\bigskip
\item<6-> $\_\;\act\;\_ :: \text{perm} \Rightarrow \alpha \Rightarrow \alpha$\medskip
\begin{itemize}
\item $0\;\act\;x = x$\\
\item $(\pi_1 + \pi_2)\;\act\;x = \pi_1\;\act\;(\pi_2\;\act\;x)$
\end{itemize}\medskip
\onslide<7->{\alert{$\;\mapsto\;$}only one type class, $\text{finite}(\text{supp}\;x)$, $\forall \pi. P$}
\end{itemize}
\only<5>{
\begin{textblock}{6}(2.5,11)
\begin{tikzpicture}
\draw (0,0) node[inner sep=3mm,fill=cream, ultra thick, draw=red, rounded corners=2mm]
{\normalsize\color{darkgray}
\begin{minipage}{8cm}\raggedright
This is slightly odd, since in general:
\begin{center}$\pi_1 + \pi_2 \alert{\not=} \pi_2 + \pi_1$\end{center}
\end{minipage}};
\end{tikzpicture}
\end{textblock}}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-3>
\frametitle{One Snatch}
\begin{itemize}
\item \underline{concrete} atoms:
\end{itemize}
*}
(*<*)
consts sort :: "atom \<Rightarrow> string"
(*>*)
typedef name = "{a :: atom. sort a = ''name''}" (*<*)sorry(*>*)
typedef ident = "{a :: atom. sort a = ''ident''}" (*<*)sorry(*>*)
text_raw {*
\mbox{}\bigskip\bigskip
\begin{itemize}
\item<2-> there is an overloaded function \alert{\bf atom}, which injects concrete
atoms into generic ones\medskip
\begin{center}
\begin{tabular}{l}
$\text{atom}(a) \fresh x$\\
$(a \leftrightarrow b) \dn (\text{atom}(a)\;\;\text{atom}(b))$
\end{tabular}
\end{center}\bigskip\medskip
\onslide<3->
{One would like to have $a \fresh x$, $(a\; b)$, \ldots}
\end{itemize}
\only<1>{
\begin{textblock}{11}(2.5,8.5)
You like to get the advantages of the old way back: you \alert{cannot mix} atoms
of dif$\!$ferent sort:\bigskip
\small
e.g.~LF-objects:\\[-9mm]\mbox{}
\begin{center}$\quad M ::= c \mid x \mid \lambda x\!:\!A. M \mid M_1\;M_2$
\end{center}
\end{textblock}}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-4>
\frametitle{Sorts Reloaded}
*}
datatype atom\<iota> = Atom\<iota> string nat
text_raw {*
\isanewline\isanewline
\pause
\alert{Problem}: HOL-binders or Church-style lambda-terms
\begin{center}
$\lambda x_\alpha.\, x_\alpha\;x_\beta$
\end{center}
\pause
\isanewline\isanewline
\isacommand{datatype} ty = TVar string $\mid$ ty $\rightarrow$ ty\\
\isacommand{datatype} var = Var name ty\\
\pause
$(x \leftrightarrow y) \,\act\, (x_\alpha, x_\beta) = (y_\alpha, y_\beta)$
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-3>
\frametitle{Non-Working Solution}
Instead of\isanewline\isanewline
*}
datatype atom\<iota>\<iota> = Atom\<iota>\<iota> string nat
text_raw {*
\isanewline\isanewline
have
\isanewline\isanewline
*}
datatype 'a atom\<iota>\<iota>\<iota> = Atom\<iota>\<iota> 'a nat
text_raw {*
\pause
\isanewline\isanewline
But then
\begin{center}
@{text "_ \<bullet> _ :: \<alpha> perm \<Rightarrow> \<beta> \<Rightarrow> \<beta>"}
\end{center}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-3>
\frametitle{A Working Solution}
*}
datatype sort = Sort string "sort list"
datatype atom\<iota>\<iota> = Atom\<iota>\<iota> string nat
text_raw {*
\isanewline\isanewline
have
\isanewline\isanewline
*}
datatype 'a atom\<iota>\<iota>\<iota> = Atom\<iota>\<iota> 'a nat
text_raw {*
\pause
\isanewline\isanewline
But then
\begin{center}
@{text "_ \<bullet> _ :: \<alpha> perm \<Rightarrow> \<beta> \<Rightarrow> \<beta>"}
\end{center}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
text_raw {*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\mode<presentation>{
\begin{frame}<1-2>[c]
\frametitle{Conclusion}
\mbox{}\\[-3mm]
\begin{itemize}
\item the formalised version of the nominal theory is now much nicer to
work with (sorts are occasionally explicit)\bigskip
\item permutations: ``be as abstract as you can'' (group\_add is a slight oddity)\bigskip
\item allow sort-disrespecting swappings\onslide<2->{: just define them as the identity}
\end{itemize}
\end{frame}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
*}
(*<*)
end
(*>*)