Add phlex::identifier

phlex/configuration.cpp

@@ -1,14 +1,15 @@
 #include "phlex/configuration.hpp"
 #include "phlex/core/product_query.hpp"
+#include "phlex/model/identifier.hpp"
 #include "phlex/model/product_specification.hpp"
 
 #include <algorithm>
 #include <array>
 #include <string>
 #include <string_view>
 
-namespace {
-  [[maybe_unused]] std::optional<std::string> value_if_exists(
+namespace phlex::detail {
+  std::optional<phlex::experimental::identifier> value_if_exists(
     boost::json::object const& obj, // will be used later for new product_query
     std::string_view parameter)
   {
@@ -45,12 +46,12 @@ namespace {
         // std::unreachable();
         break;
       }
-      throw std::runtime_error(
-        fmt::format("Error retrieving parameter '{}'. Should be a string but is instead a {}",
-                    parameter,
-                    kind));
+      throw std::runtime_error(fmt::format(
+        "Error retrieving parameter '{}'. Should be an identifier string but is instead a {}",
+        parameter,
+        kind));
     }
-    return boost::json::value_to<std::string>(val);
+    return boost::json::value_to<phlex::experimental::identifier>(val);
   }
 }
 
@@ -79,4 +80,10 @@ namespace phlex {
     auto layer = value_decorate_exception<std::string>(query_object, "layer");
     return product_query{experimental::product_specification::create(product), layer};
   }
+
+  experimental::identifier experimental::tag_invoke(boost::json::value_to_tag<identifier> const&,
+                                                    boost::json::value const& jv)
+  {
+    return identifier{std::string_view(jv.as_string())};
+  }
 }

phlex/configuration.hpp

@@ -3,6 +3,7 @@
 
 #include "boost/json.hpp"
 #include "phlex/core/product_query.hpp"
+#include "phlex/model/identifier.hpp"
 
 #include <optional>
 #include <string>
@@ -20,6 +21,10 @@ namespace phlex {
       throw std::runtime_error("Error retrieving parameter '" + key + "':\n" + e.what());
     }
 
+    // Used later for product_query
+    std::optional<phlex::experimental::identifier> value_if_exists(boost::json::object const& obj,
+                                                                   std::string_view parameter);
+
     // helper for unpacking json array
     template <typename T, std::size_t... I>
     std::array<T, sizeof...(I)> unpack_json_array(boost::json::array const& array,
@@ -90,6 +95,11 @@ namespace phlex {
   product_query tag_invoke(boost::json::value_to_tag<product_query> const&,
                            boost::json::value const& jv);
 
+  namespace experimental {
+    identifier tag_invoke(boost::json::value_to_tag<identifier> const&,
+                          boost::json::value const& jv);
+  }
+
   template <std::size_t N>
   std::array<product_query, N> tag_invoke(
     boost::json::value_to_tag<std::array<product_query, N>> const&, boost::json::value const& jv)

phlex/model/CMakeLists.txt

@@ -7,6 +7,7 @@ cet_make_library(
   data_cell_counter.cpp
   data_layer_hierarchy.cpp
   data_cell_index.cpp
+  identifier.cpp
   product_matcher.cpp
   product_store.cpp
   products.cpp
@@ -30,6 +31,7 @@ install(
     data_cell_counter.hpp
     data_layer_hierarchy.hpp
     data_cell_index.hpp
+    identifier.hpp
     product_matcher.hpp
     product_specification.hpp
     product_store.hpp

phlex/model/identifier.cpp

@@ -0,0 +1,46 @@
+#include "identifier.hpp"
+
+#include <boost/hash2/hash_append.hpp>
+#include <boost/hash2/xxhash.hpp>
+
+namespace phlex::experimental {
+  identifier_query literals::operator""_idq(char const* lit, std::size_t len)
+  {
+    return {identifier::hash_string(std::string_view(lit, len))};
+  }
+
+  std::uint64_t identifier::hash_string(std::string_view str)
+  {
+    // Hash quality is very important here, since comparisons are done using only the hash
+    using namespace boost::hash2;
+    xxhash_64 h;
+    hash_append(h, {}, str);
+    return h.result();
+  }
+
+  identifier::identifier(std::string_view str) : content_(str), hash_(hash_string(content_)) {}
+  identifier& identifier::operator=(std::string_view str)
+  {
+    content_ = str;
+    hash_ = hash_string(content_);
+    return *this;
+  }
+
+  identifier::operator std::string_view() const noexcept { return std::string_view(content_); }
+  bool identifier::operator==(identifier const& rhs) const noexcept { return hash_ == rhs.hash_; }
+  std::strong_ordering identifier::operator<=>(identifier const& rhs) const noexcept
+  {
+    return hash_ <=> rhs.hash_;
+  }
+
+  bool operator==(identifier const& lhs, identifier_query rhs) { return lhs.hash_ == rhs.hash; }
+  std::strong_ordering operator<=>(identifier const& lhs, identifier_query rhs)
+  {
+    return lhs.hash_ <=> rhs.hash;
+  }
+
+  identifier literals::operator""_id(char const* lit, std::size_t len)
+  {
+    return identifier{std::string_view(lit, len)};
+  }
+}

phlex/model/identifier.hpp

@@ -0,0 +1,72 @@
+#ifndef PHLEX_MODEL_IDENTIFIER_H_
+#define PHLEX_MODEL_IDENTIFIER_H_
+
+#include <boost/json/fwd.hpp>
+
+#include <fmt/format.h>
+
+#include <cstdint>
+#include <string>
+#include <string_view>
+
+namespace phlex::experimental {
+  /// If you're comparing to an identifier you know at compile time, you're probably not going to need
+  /// to print it.
+  struct identifier_query {
+    std::uint64_t hash;
+  };
+
+  /// Carries around the string itself (as a shared_ptr to string to make copies lighter)
+  /// along with a precomputed hash used for all comparisons
+  class identifier {
+  public:
+    static std::uint64_t hash_string(std::string_view str);
+    identifier(identifier const& other) = default;
+    identifier(identifier&& other) noexcept = default;
+
+    explicit identifier(std::string_view str);
+
+    identifier& operator=(identifier const& rhs) = default;
+    identifier& operator=(identifier&& rhs) noexcept = default;
+
+    // Assignment for identifiers read from a file
+    // Not sure we really need this
+    identifier& operator=(std::string_view str);
+
+    ~identifier() = default;
+
+    // Conversion to std::string_view
+    explicit operator std::string_view() const noexcept;
+
+    bool operator==(identifier const& rhs) const noexcept;
+    std::strong_ordering operator<=>(identifier const& rhs) const noexcept;
+
+    // Comparison operators with _id queries
+    friend bool operator==(identifier const& lhs, identifier_query rhs);
+    friend std::strong_ordering operator<=>(identifier const& lhs, identifier_query rhs);
+    friend std::hash<identifier>;
+
+  private:
+    std::string content_;
+    std::uint64_t hash_;
+  };
+
+  // Identifier UDL
+  namespace literals {
+    identifier operator""_id(char const* lit, std::size_t len);
+    identifier_query operator""_idq(char const* lit, std::size_t len);
+  }
+
+  // Really trying to avoid the extra function call here
+  inline std::string_view format_as(identifier const& id) { return std::string_view(id); }
+}
+
+template <>
+struct std::hash<phlex::experimental::identifier> {
+  std::size_t operator()(phlex::experimental::identifier const& id) const noexcept
+  {
+    return id.hash_;
+  }
+};
+
+#endif // PHLEX_MODEL_IDENTIFIER_H_

test/CMakeLists.txt

@@ -23,6 +23,7 @@ cet_test(type_deduction SOURCE type_deduction.cpp LIBRARIES
          phlex::metaprogramming
 )
 cet_test(type_id SOURCE type_id.cpp LIBRARIES phlex::model fmt::fmt)
+cet_test(identifier SOURCE identifier.cpp LIBRARIES phlex::model phlex::configuration fmt::fmt)
 cet_test(
   yielding_driver
   SOURCE

test/identifier.cpp

@@ -0,0 +1,72 @@
+#include "phlex/model/identifier.hpp"
+#include <phlex/configuration.hpp>
+
+#include <algorithm>
+#include <array>
+#include <cassert>
+#include <string_view>
+
+#include <boost/json.hpp>
+#include <fmt/format.h>
+
+int main()
+{
+  using namespace phlex::experimental;
+  using namespace phlex::experimental::literals;
+  using namespace std::string_view_literals;
+  identifier a = "a"_id;
+  identifier a_copy = a;
+  identifier a2 = "a"_id;
+  identifier b = "b"_id;
+
+  boost::json::object parsed_json = boost::json::parse(R"( {"identifier": "b" } )").as_object();
+  auto b_from_json = phlex::detail::value_if_exists(parsed_json, "identifier");
+  assert(b_from_json);
+
+  fmt::print("a ({}) == \"a\"_idq: {}\n", a, a == "a"_idq);
+  fmt::print("a == a_copy ({}): {}\n", a_copy, a == a_copy);
+  fmt::print("a == a2 ({}): {}\n", a2, a == a2);
+  fmt::print("a != b ({}): {}\n", b, a != b);
+  fmt::print("b == *b_from_json ({}): {}\n", *b_from_json, b == *b_from_json);
+
+  assert(a == "a"_idq);
+  assert(a == a_copy);
+  assert(a == a2);
+  assert(a != b);
+  assert(b == *b_from_json);
+
+  // reassigning
+  a = "new a"_id;
+  a_copy = a;
+
+  // reassigning with a string_view
+  b = "new b"sv;
+
+  std::array<std::string_view, 9> strings{
+    "a"sv, "b"sv, "c"sv, "d"sv, "e"sv, "long-id-1"sv, "long-id-2"sv, "test"sv, "other_test"sv};
+  std::array<identifier, 9> identifiers{"a"_id,
+                                        "b"_id,
+                                        "c"_id,
+                                        "d"_id,
+                                        "e"_id,
+                                        "long-id-1"_id,
+                                        "long-id-2"_id,
+                                        "test"_id,
+                                        "other_test"_id};
+  std::ranges::sort(identifiers);
+  std::ranges::sort(strings);
+  fmt::print("Sorted identifiers (should not be in alphabetical order): \n");
+  for (auto const& id : identifiers) {
+    fmt::print("- {}\n", id);
+  }
+
+  bool ok = false;
+  for (std::size_t i = 0; i < identifiers.size(); ++i) {
+    if (strings.at(i) != std::string_view(identifiers.at(i))) {
+      ok = true;
+      break;
+    }
+  }
+  assert(ok);
+  return 0;
+}