src/Doc/Tutorial/Inductive/Star.thy
 changeset 69609 ff784d5a5bfb parent 69517 cc2d676d5395
--- a/src/Doc/Tutorial/Inductive/Star.thy	Sat Jan 05 17:00:43 2019 +0100
+++ b/src/Doc/Tutorial/Inductive/Star.thy	Sat Jan 05 17:24:33 2019 +0100
@@ -22,12 +22,12 @@
| rtc_step:       "\<lbrakk> (x,y) \<in> r; (y,z) \<in> r* \<rbrakk> \<Longrightarrow> (x,z) \<in> r*"

text\<open>\noindent
-The function @{term rtc} is annotated with concrete syntax: instead of
-\<open>rtc r\<close> we can write @{term"r*"}. The actual definition
+The function \<^term>\<open>rtc\<close> is annotated with concrete syntax: instead of
+\<open>rtc r\<close> we can write \<^term>\<open>r*\<close>. The actual definition
consists of two rules. Reflexivity is obvious and is immediately given the
\<open>iff\<close> attribute to increase automation. The
second rule, @{thm[source]rtc_step}, says that we can always add one more
-@{term r}-step to the left. Although we could make @{thm[source]rtc_step} an
+\<^term>\<open>r\<close>-step to the left. Although we could make @{thm[source]rtc_step} an
introduction rule, this is dangerous: the recursion in the second premise
slows down and may even kill the automatic tactics.

@@ -44,7 +44,7 @@
text\<open>\noindent
Although the lemma itself is an unremarkable consequence of the basic rules,
it has the advantage that it can be declared an introduction rule without the
-danger of killing the automatic tactics because @{term"r*"} occurs only in
+danger of killing the automatic tactics because \<^term>\<open>r*\<close> occurs only in
the conclusion and not in the premise. Thus some proofs that would otherwise
need @{thm[source]rtc_step} can now be found automatically. The proof also
shows that \<open>blast\<close> is able to handle @{thm[source]rtc_step}. But
@@ -72,19 +72,19 @@
To understand what is going on, let us look again at @{thm[source]rtc.induct}.
In the above application of \<open>erule\<close>, the first premise of
@{thm[source]rtc.induct} is unified with the first suitable assumption, which
-is @{term"(x,y) \<in> r*"} rather than @{term"(y,z) \<in> r*"}. Although that
+is \<^term>\<open>(x,y) \<in> r*\<close> rather than \<^term>\<open>(y,z) \<in> r*\<close>. Although that
is what we want, it is merely due to the order in which the assumptions occur
in the subgoal, which it is not good practice to rely on. As a result,
-\<open>?xb\<close> becomes @{term x}, \<open>?xa\<close> becomes
-@{term y} and \<open>?P\<close> becomes @{term"\<lambda>u v. (u,z) \<in> r*"}, thus
+\<open>?xb\<close> becomes \<^term>\<open>x\<close>, \<open>?xa\<close> becomes
+\<^term>\<open>y\<close> and \<open>?P\<close> becomes \<^term>\<open>\<lambda>u v. (u,z) \<in> r*\<close>, thus
yielding the above subgoal. So what went wrong?

When looking at the instantiation of \<open>?P\<close> we see that it does not
depend on its second parameter at all. The reason is that in our original
-goal, of the pair @{term"(x,y)"} only @{term x} appears also in the
-conclusion, but not @{term y}. Thus our induction statement is too
+goal, of the pair \<^term>\<open>(x,y)\<close> only \<^term>\<open>x\<close> appears also in the
+conclusion, but not \<^term>\<open>y\<close>. Thus our induction statement is too
general. Fortunately, it can easily be specialized:
-transfer the additional premise @{prop"(y,z)\<in>r*"} into the conclusion:\<close>
+transfer the additional premise \<^prop>\<open>(y,z)\<in>r*\<close> into the conclusion:\<close>
(*<*)oops(*>*)
lemma rtc_trans[rule_format]:
"(x,y) \<in> r* \<Longrightarrow> (y,z) \<in> r* \<longrightarrow> (x,z) \<in> r*"
@@ -114,9 +114,9 @@
done

text\<open>
-Let us now prove that @{term"r*"} is really the reflexive transitive closure
-of @{term r}, i.e.\ the least reflexive and transitive
-relation containing @{term r}. The latter is easily formalized
+Let us now prove that \<^term>\<open>r*\<close> is really the reflexive transitive closure
+of \<^term>\<open>r\<close>, i.e.\ the least reflexive and transitive
+relation containing \<^term>\<open>r\<close>. The latter is easily formalized
\<close>

inductive_set
@@ -151,7 +151,7 @@
transitivity.  As a consequence, @{thm[source]rtc.induct} is simpler than
@{thm[source]rtc2.induct}. Since inductive proofs are hard enough
anyway, we should always pick the simplest induction schema available.
-Hence @{term rtc} is the definition of choice.
+Hence \<^term>\<open>rtc\<close> is the definition of choice.
\index{reflexive transitive closure!defining inductively|)}

\begin{exercise}\label{ex:converse-rtc-step}
@@ -160,7 +160,7 @@
\end{exercise}
\begin{exercise}
Repeat the development of this section, but starting with a definition of
-@{term rtc} where @{thm[source]rtc_step} is replaced by its converse as shown
+\<^term>\<open>rtc\<close> where @{thm[source]rtc_step} is replaced by its converse as shown
in exercise~\ref{ex:converse-rtc-step}.
\end{exercise}
\<close>