feat: constexpr URL parsing for C++20

fix #890

Author: alandefreitas

.gitignore

@@ -9,7 +9,12 @@ x64
 
 cmake-build-*/
 cmake-build/
+build/*
+!build/Jamfile
+local/
+tmp/
 AGENTS.md
+CLAUDE.md
 .roadmap/
 .idea/
 doc/modules/reference/

doc/modules/ROOT/examples/unit/snippets.cpp

@@ -684,6 +684,38 @@ parsing_urls()
         // end::snippet_parsing_url_3[]
         BOOST_TEST(v.buffer() == "/path/to/file.txt#anchor");
     }
+
+#if defined(BOOST_URL_HAS_CXX20_CONSTEXPR)
+    {
+        // tag::snippet_constexpr_parsing_1[]
+        // Parse and validate a URL at compile time (C++20).
+        // value() on an error produces a compile error,
+        // so invalid URLs are caught at build time.
+        constexpr url_view api_base =
+            parse_uri("https://api.example.com/v2").value();
+
+        // A malformed literal would fail to compile:
+        // constexpr url_view bad =
+        //     parse_uri("ht tp://bad url").value();
+        // end::snippet_constexpr_parsing_1[]
+        BOOST_TEST(api_base.scheme() == "https");
+    }
+    {
+        // tag::snippet_constexpr_parsing_2[]
+        // Pre-parsed URL constants have zero runtime cost.
+        // Parsing happens entirely at compile time.
+        constexpr url_view default_endpoint =
+            parse_uri("https://api.example.com:8443/v1").value();
+        constexpr url_view fallback_endpoint =
+            parse_uri("http://localhost:9090/debug").value();
+
+        // Components are available at runtime with no parsing overhead
+        assert(default_endpoint.port_number() == 8443);
+        assert(fallback_endpoint.scheme() == "http");
+        // end::snippet_constexpr_parsing_2[]
+        BOOST_TEST(default_endpoint.port_number() == 8443);
+    }
+#endif
 }
 
 void

doc/modules/ROOT/pages/urls/parsing.adoc

@@ -246,3 +246,26 @@ include::example$unit/snippets.cpp[tag=snippet_parsing_url_1b,indent=0]
 Check the reference for cpp:result[] for a synopsis of the type.
 
 
+== Constexpr Parsing
+
+In {cpp}20 and later, all parse functions are `constexpr`.
+This means URLs can be parsed and validated at compile time, which is useful when one side of a comparison is a fixed string that should be pre-parsed and checked for correctness.
+
+For instance, a `constexpr` cpp:url_view[] parsed from a valid string literal is fully resolved at compile time.
+If the literal is malformed, `value()` attempts to throw, which is not permitted during constant evaluation, so the program fails to compile:
+
+[source,cpp]
+----
+include::example$unit/snippets.cpp[tag=snippet_constexpr_parsing_1,indent=0]
+----
+
+Pre-parsed `constexpr` URL views can also serve as zero-cost constants.
+All parsing is done at compile time; at runtime the components are accessed with no parsing overhead:
+
+[source,cpp]
+----
+include::example$unit/snippets.cpp[tag=snippet_constexpr_parsing_2,indent=0]
+----
+
+Because cpp:url_view[] does not own its character buffer, `constexpr` URL views always reference string literals.
+

include/boost/url/authority_view.hpp

@@ -1,5 +1,6 @@
 //
 // Copyright (c) 2019 Vinnie Falco (vinnie.falco@gmail.com)
+// Copyright (c) 2022 Alan de Freitas (alandefreitas@gmail.com)
 //
 // Distributed under the Boost Software License, Version 1.0. (See accompanying
 // file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
@@ -25,6 +26,15 @@
 namespace boost {
 namespace urls {
 
+class url_base;
+
+namespace implementation_defined {
+struct authority_rule_t;
+struct uri_rule_t;
+struct relative_ref_rule_t;
+struct absolute_uri_rule_t;
+} // implementation_defined
+
 /** A non-owning reference to a valid authority
 
     Objects of this type represent valid authority
@@ -79,17 +89,27 @@ namespace urls {
     @see
         @ref parse_authority.
 */
-class BOOST_URL_DECL
+class BOOST_SYMBOL_VISIBLE
     authority_view
     : private detail::parts_base
 {
     detail::url_impl u_;
 
     friend struct detail::url_impl;
-
+    friend struct implementation_defined::authority_rule_t;
+    friend struct implementation_defined::uri_rule_t;
+    friend struct implementation_defined::relative_ref_rule_t;
+    friend struct implementation_defined::absolute_uri_rule_t;
+    friend class url_view_base;
+    friend class url_base;
+
+    BOOST_URL_CXX14_CONSTEXPR
     explicit
     authority_view(
-        detail::url_impl const& u) noexcept;
+        detail::url_impl const& u) noexcept
+        : u_(u)
+    {
+    }
 
 public:
     //--------------------------------------------
@@ -100,8 +120,15 @@ class BOOST_URL_DECL
 
     /** Destructor
     */
+    BOOST_URL_CXX20_CONSTEXPR
     virtual
-    ~authority_view();
+    ~authority_view()
+    {
+        // Empty body instead of = default because
+        // some GCC versions reject a defaulted
+        // virtual constexpr destructor in constexpr
+        // evaluation ("used before its definition").
+    }
 
     /** Constructor
 
@@ -115,7 +142,11 @@ class BOOST_URL_DECL
 
         @par Specification
     */
-    authority_view() noexcept;
+    BOOST_URL_CXX14_CONSTEXPR
+    authority_view() noexcept
+        : u_(from::authority)
+    {
+    }
 
     /** Construct from a string.
 
@@ -154,13 +185,15 @@ class BOOST_URL_DECL
         @see
             @ref parse_authority.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     explicit
     authority_view(core::string_view s);
 
     /** Constructor
     */
+    BOOST_URL_CXX14_CONSTEXPR
     authority_view(
-        authority_view const&) noexcept;
+        authority_view const&) noexcept = default;
 
     /** Assignment
 
@@ -173,9 +206,10 @@ class BOOST_URL_DECL
         @par Exception Safety
         Throws nothing.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     authority_view&
     operator=(
-        authority_view const& other) noexcept;
+        authority_view const& other) noexcept = default;
 
     //--------------------------------------------
     //
@@ -317,6 +351,7 @@ class BOOST_URL_DECL
             @ref userinfo.
 
     */
+    BOOST_URL_CXX20_CONSTEXPR
     bool
     has_userinfo() const noexcept;
 
@@ -416,6 +451,7 @@ class BOOST_URL_DECL
             @ref user,
             @ref userinfo.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_userinfo() const noexcept;
 
@@ -519,6 +555,7 @@ class BOOST_URL_DECL
             @ref user,
             @ref userinfo.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_user() const noexcept;
 
@@ -562,6 +599,7 @@ class BOOST_URL_DECL
             @ref user,
             @ref userinfo.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     bool
     has_password() const noexcept;
 
@@ -659,6 +697,7 @@ class BOOST_URL_DECL
             @ref user,
             @ref userinfo.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_password() const noexcept;
 
@@ -782,6 +821,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_host() const noexcept;
 
@@ -901,6 +941,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_host_address() const noexcept;
 
@@ -942,6 +983,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     ipv4_address
     host_ipv4_address() const noexcept;
 
@@ -991,6 +1033,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     ipv6_address
     host_ipv6_address() const noexcept;
 
@@ -1025,6 +1068,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     core::string_view
     host_ipvfuture() const noexcept;
 
@@ -1112,6 +1156,7 @@ class BOOST_URL_DECL
         @li <a href="https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.2"
             >3.2.2. Host (rfc3986)</a>
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_host_name() const noexcept;
 
@@ -1155,6 +1200,7 @@ class BOOST_URL_DECL
             @ref port,
             @ref port_number.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     bool
     has_port() const noexcept;
 
@@ -1192,6 +1238,7 @@ class BOOST_URL_DECL
             @ref has_port,
             @ref port_number.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     core::string_view
     port() const noexcept;
 
@@ -1229,6 +1276,7 @@ class BOOST_URL_DECL
             @ref has_port,
             @ref port.
     */
+    BOOST_URL_CXX20_CONSTEXPR
     std::uint16_t
     port_number() const noexcept;
 
@@ -1270,6 +1318,7 @@ class BOOST_URL_DECL
 
         @return The host and port
     */
+    BOOST_URL_CXX20_CONSTEXPR
     pct_string_view
     encoded_host_and_port() const noexcept;
 
@@ -1537,14 +1586,19 @@ operator<<(
     @see
         @ref authority_view.
 */
-BOOST_URL_DECL
+BOOST_URL_CXX20_CONSTEXPR
 system::result<authority_view>
 parse_authority(
     core::string_view s) noexcept;
 
-//------------------------------------------------
-
 } // urls
 } // boost
 
+// When rfc/authority_rule.hpp is being processed,
+// it will include impl/authority_view.hpp itself
+// after declaring authority_rule.
+#if !defined(BOOST_URL_RFC_AUTHORITY_RULE_HPP)
+#include <boost/url/impl/authority_view.hpp>
+#endif
+
 #endif

include/boost/url/detail/any_params_iter.hpp

@@ -54,7 +54,6 @@ struct BOOST_SYMBOL_VISIBLE
     // True if the sequence is empty
     bool empty = false;
 
-    BOOST_URL_DECL
     virtual
     ~any_params_iter() noexcept = 0;
 
@@ -93,7 +92,6 @@ struct BOOST_SYMBOL_VISIBLE
     : any_params_iter
 {
     // ne = never empty
-    BOOST_URL_DECL
     explicit
     query_string_iter(
         core::string_view s,
@@ -154,14 +152,12 @@ struct params_iter_base
         {}
 
     // return encoded size
-    BOOST_URL_DECL
     void
     measure_impl(
         std::size_t& n,
         param_view const& p) noexcept;
 
     // encode to dest
-    BOOST_URL_DECL
     void
     copy_impl(
         char*& dest,
@@ -264,14 +260,12 @@ struct BOOST_SYMBOL_VISIBLE
 struct params_encoded_iter_base
 {
 protected:
-    BOOST_URL_DECL
     static
     void
     measure_impl(
         std::size_t& n,
         param_view const& v) noexcept;
 
-    BOOST_URL_DECL
     static
     void
     copy_impl(
@@ -430,4 +424,6 @@ make_params_encoded_iter(
 } // urls
 } // boost
 
+#include <boost/url/detail/impl/any_params_iter.hpp>
+
 #endif