New issue from Hewill: "Improve optional<T&>::or_else"

Author: Dani-Hub

xml/issue4367.xml

@@ -0,0 +1,58 @@
+<?xml version='1.0' encoding='utf-8' standalone='no'?>
+<!DOCTYPE issue SYSTEM "lwg-issue.dtd">
+
+<issue num="4367" status="New">
+<title>Improve <code>optional&lt;T&amp;&gt;::or_else</code></title>
+<section>
+<sref ref="[optional.ref.monadic]"/>
+</section>
+<submitter>Hewill Kang</submitter>
+<date>07 Sep 2025</date>
+<priority>99</priority>
+
+<discussion>
+<p>
+<code>optional&lt;T&amp;&gt;::or_else</code> currently returns <code>*<i>val</i></code> when it has a value,
+which calls the <code>optional(U&amp;&amp;)</code> constructor which in turn calls 
+<code><i>convert-ref-init-val</i></code> which in turn calls `addressof`.
+<p/>
+There's no reason to do that. It's much more efficient to call <code>optional&lt;T&amp;&gt;</code>'s 
+default copy constructor to just copy a pointer.
+</p>
+</discussion>
+
+<resolution>
+<p>
+This wording is relative to <paper num="N5014"/>.
+</p>
+
+<ol>
+
+<li><p>Modify <sref ref="[optional.ref.monadic]"/> as indicated:</p>
+
+<blockquote>
+<pre>
+template&lt;class F&gt; constexpr optional or_else(F&amp;&amp; f) const;
+</pre>
+<blockquote>
+<p>
+-7- <i>Constraints</i>: `F` models `invocable`.
+<p/>
+-8- <i>Mandates</i>: <code>is_same_v&lt;remove_cvref_t&lt;invoke_result_t&lt;F&gt;&gt;, optional&gt;</code> is `true`.
+<p/>
+-9- <i>Effects</i>: Equivalent to:
+</p>
+<blockquote><pre>
+if (has_value()) {
+  return *<ins>this</ins><del><i>val</i></del>;
+} else {
+  return std::forward&lt;F&gt;(f)();
+}
+</pre></blockquote></blockquote>
+</blockquote>
+
+</li>
+
+</ol></resolution>
+
+</issue>