We had an index entry for "median" attached to the percentile_cont function
entry, which was pretty useless because a person following the link would
never realize that that function was the one they were being hinted to use.
Instead, make the index entry point at the example in syntax-aggregates,
and add a <seealso> link to "percentile".
Also, since that example explicitly claims to be calculating the median,
make it use percentile_cont not percentile_disc.  This makes no difference
in terms of the larger goals of that section, but so far as I can find,
nearly everyone thinks that "median" means the continuous not discrete
calculation.
Per gripe from Steven Winfield.  Back-patch to 9.4 where we introduced
percentile_cont.
Discussion: https://postgr.es/m/
20161223102056[email protected]
 
         <primary>percentile</primary>
         <secondary>continuous</secondary>
        </indexterm>
-       <indexterm>
-        <primary>median</primary>
-       </indexterm>
        <function>percentile_cont(<replaceable class="parameter">fraction</replaceable>) WITHIN GROUP (ORDER BY <replaceable class="parameter">sort_expression</replaceable>)</function>
       </entry>
       <entry>
 
     case, write just <literal>()</> not <literal>(*)</>.
     (<productname>PostgreSQL</> will actually accept either spelling, but
     only the first way conforms to the SQL standard.)
+   </para>
+
+   <para>
+    <indexterm>
+     <primary>median</primary>
+    </indexterm>
+    <indexterm>
+     <primary>median</primary>
+     <seealso>percentile</seealso>
+    </indexterm>
     An example of an ordered-set aggregate call is:
 
 <programlisting>
-SELECT percentile_disc(0.5) WITHIN GROUP (ORDER BY income) FROM households;
- percentile_disc
+SELECT percentile_cont(0.5) WITHIN GROUP (ORDER BY income) FROM households;
+ percentile_cont
 -----------------
            50489
 </programlisting>