Add conclusions

1 files changed, 32 insertions, 4 deletions

diff --git a/cloogle.tex b/cloogle.tex
index f0af4e2..f93c50f 100644
--- a/cloogle.tex
+++ b/cloogle.tex
@@ -92,6 +92,9 @@
 		\item With $\pm15,000$ functions, brute force unification search takes too long
 		\item Grouping types together still gives $\pm13,000$ unique types
 		\item If type $t$ generalises $u$, any $v$ that unifies with $u$ will also unify with $t$
+			\begin{itemize}
+				\item Conversely, if $v$ does not unify with $t$ it will not unify with $u$
+			\end{itemize}
 		\item This gives a partial order on types:
 
 			\begin{center}
@@ -115,7 +118,7 @@
 
 \begin{frame}[shrink]{Efficient unification search}
 	\begin{itemize}[<+->]
-		\item I would show you the real tree, but it is too large for most viewers
+		\item I would show the full tree, but didn't find a viewer that wouldn't crash
 		\item Here is a detail of the tree for 5,000 functions:
 
 			\medskip
@@ -217,9 +220,34 @@
 
 \section*{The end}
 
-\subsection*{Conclusions}
-\begin{frame}{Conclusions}
-	% TODO
+\subsection*{Discussion}
+\begin{frame}[fragile]{Discussion}
+	\begin{itemize}[<+->]
+		\item The first generation of students was very enthusiastic
+			\begin{itemize}
+				\item The second not so much...
+				\item But statistics show they used it more and more during the course
+			\end{itemize}
+
+		\item Unification search has some issues: you will never find
+
+			\begin{minted}[gobble=4]{clean}
+				fopen :: String Int *f -> (Bool, *File, *f)
+			\end{minted}
+
+			\begin{itemize}
+				\item But a unifier gives useful information, like required instances
+				\item Unification allows for optimisations that a distance measure does not
+			\end{itemize}
+
+		\item It's not just about search: indexing code has many use cases!
+
+			\begin{itemize}
+				\item Jumping to definitions (tagfiles)
+				\item Automatic imports (vim-clean)
+				\item Type-based code completion?
+			\end{itemize}
+	\end{itemize}
 \end{frame}
 
 \subsection*{Acknowledgements}