C++ imrprovements and better alignment with the standard

bin/mzp-inspect.cpp

@@ -23,7 +23,7 @@ namespace po = boost::program_options;
 std::unique_ptr<MzPeak::Util::Parquet> open_parquet_file(MzPeak::Index& index,
                                                          const std::string& file)
 {
-  auto it = index.find(file);
+  auto it = index.manager()->find_file(file);
 
   if (it == index.files().end()) {
     std::println(stderr, "file \"{}\" is not in the mzPeak file index", file);
@@ -41,7 +41,7 @@ int print_array_index(MzPeak::Index& index, const std::string& file)
 
   auto fmd = parquet->file_metadata();
   auto et = parquet->index_file().entity_type();
-  auto key = MzPeak::Schema::entity_type_to_string(et) + "_array_index";
+  auto key = et.array_index_name();
   auto json = parquet->kv_string(fmd, key);
 
   if (!json.has_value()) {

include/mzpeak/data/array_index.h

@@ -62,7 +62,7 @@ class ArrayIndex final {
     BufferFormat buffer_format = BufferFormat::Point;
 
     /// The entity type this column belongs to.
-    EntityType context = EntityType::Other;
+    EntityType context = EntityType("other");
 
     /// The path from the *root* of the Parquet file's schema to this
     /// column.
@@ -128,6 +128,9 @@ class ArrayIndex final {
     /// The type of elements stored in this dimension.
     Schema::PSI::ArrayType array_type;
 
+    /// Is this the primary dimension for array type?
+    bool buffer_priority;
+
     /// Transformation information for this dimension.
     std::optional<Schema::PSI::Transform> transform;
 
@@ -204,7 +207,7 @@ class ArrayIndex final {
 
 private:
   // The entity type for the entire Parquet file.
-  EntityType entity_type_ = EntityType::Other;
+  EntityType entity_type_ = EntityType("other");
 
   // Root node.
   std::string prefix_ = "point";

include/mzpeak/data/encoding.h

@@ -159,7 +159,8 @@ void Decoder<T>::decode_with_nulls(const ArrayIndex::Dimension& dim,
   auto col = signals_->column(primary_entry);
 
   if (!col.has_value()) {
-    throw ParquetError("unable to decode dimension, not in schema: " + dim.name);
+    throw InvalidFormatError("unable to decode dimension, not in schema: " +
+                             dim.name);
   } else if (!slice_->has_column(col.value())) {
     return; // No data to decode so we can exit early.
   }

include/mzpeak/data/null_marking.h

@@ -142,7 +142,7 @@ std::optional<T> Decoder<T, U>::operator()(int64_t index)
 {
   // Sanity check.
   if (array_ == nullptr) {
-    throw("FIXME: assertion failed");
+    throw(Exception("while decoding null marking an nullptr was encountered"));
   };
 
   if (index != 0 && prior_.index == index - 1) {
@@ -200,7 +200,7 @@ std::optional<T> Decoder<T, U>::operator()(int64_t index)
     return prior_.value;
   } else {
     // Shouldn't happen.
-    throw("FIXME: failed to decode NULL marking value");
+    throw(Exception("failed to decode NULL marking value"));
   }
 }
 

include/mzpeak/exception.h

@@ -92,16 +92,16 @@ class UnknownLayoutError final : public Exception {
 /**
  * Attempt to access an invalid iterator.
  */
-class InvalidIterator final : public Exception {
+class InvalidIteratorError final : public Exception {
 public:
   /// Constructor.
-  InvalidIterator(const std::string& msg)
+  InvalidIteratorError(const std::string& msg)
       : Exception(msg)
   {
   }
 
   /// Destructor.
-  ~InvalidIterator() = default;
+  ~InvalidIteratorError() = default;
 };
 
 /**

include/mzpeak/index.h

@@ -9,19 +9,16 @@ directory of this repository.
 #pragma once
 
 #include <memory>
-#include <string_view>
 #include <vector>
 
+#include "mzpeak/schema/file.h"
+
 namespace MzPeak {
 
 namespace IO {
 class Archive;
 }
 
-namespace Schema {
-class File;
-}
-
 namespace Util {
 class Manager;
 }
@@ -42,9 +39,10 @@ class Index {
   const std::vector<Schema::File>& files() const;
 
   /**
-   * Find a file in the mzPeak archive with the given name.
+   * Find a file given its `EntityType` and `DataKind`.
    */
-  std::vector<Schema::File>::const_iterator find(std::string_view) const;
+  std::vector<Schema::File>::const_iterator find_file(Schema::EntityType::Type,
+                                                      Schema::DataKind::Type) const;
 
   /**
    * Access the spectra in the file.

include/mzpeak/schema/cv.h

@@ -46,6 +46,12 @@ class CV {
     return code_ == other.code_ && accession_ == other.accession_;
   }
 
+  /// Less than.
+  bool operator<(const CV& other) const
+  {
+    return code_ < other.code_ && accession_ < other.accession_;
+  }
+
 private:
   std::string code_;
   std::string accession_;

include/mzpeak/schema/entity_type.h

@@ -8,36 +8,65 @@ directory of this repository.
 
 #pragma once
 
+#include <optional>
 #include <string>
+#include <variant>
 
 namespace MzPeak::Schema {
 
 /**
  * The type of data entity stored in the file.
  */
-enum class EntityType {
-  /// Mass spectra.
-  Spectrum,
+class EntityType {
+public:
+  enum Type {
+    /// Mass spectra.
+    Spectrum,
 
-  /// Chromatograms or other measurements over time.
-  Chromatogram,
+    /// Chromatograms or other measurements over time.
+    Chromatogram,
 
-  /// Similar to Spectrum except the unit of measure is a wavelength
-  /// measurement.
-  WavelengthSpectrum,
+    /// Similar to Spectrum except the unit of measure is a wavelength
+    /// measurement.
+    WavelengthSpectrum,
+  };
 
-  /// Unspecified.
-  Other
-};
+  // Internal storage type.
+  using value_type = std::variant<Type, std::string>;
 
-/**
- * Convert a EntityType to a string.
- */
-std::string entity_type_to_string(EntityType);
+  /// Constructor.
+  EntityType(std::string_view);
 
-/**
- * Parse an EntityType from a string view.
- */
-EntityType entity_type_from_string(std::string_view);
+  /// Constructor.
+  EntityType(Type);
+
+  /**
+   * Return the string representation of an entity type.
+   */
+  std::string to_string() const;
+
+  /**
+   * Return the enumerated type if it is known.
+   */
+  std::optional<Type> type() const;
+
+  /**
+   * The name of the index column matching this entity.
+   */
+  std::string index_column_name() const;
+
+  /**
+   * The name of the array index metadata key.
+   */
+  std::string array_index_name() const;
+
+  /**
+   * The metadata key for the entity count.
+   */
+  std::string metadata_count_key() const;
+
+private:
+  value_type val_;
+};
 
 } // namespace MzPeak::Schema

include/mzpeak/schema/file.h

@@ -53,7 +53,7 @@ class File final {
 private:
   std::string file_name_;
   DataKind data_kind_ = DataKind("other");
-  EntityType entity_type_ = EntityType::Other;
+  EntityType entity_type_ = EntityType("other");
   std::vector<Column> columns_;
 };
 

include/mzpeak/schema/psi/data_type.h

@@ -95,6 +95,9 @@ class DataType {
   /// Equality.
   bool operator==(const DataType& other) const;
 
+  /// Less than (for sorting).
+  bool operator<(const DataType& other) const;
+
 private:
   value_type val_;
   std::optional<Util::Type> as_type_;

include/mzpeak/util/enumerable_proxy.h

@@ -101,7 +101,7 @@ class EnumerableProxy : public std::ranges::view_interface<EnumerableProxy<T>> {
         const_cast<Iterator*>(this)->cache_ = std::make_pair<>(n_, fetch_(n_));
         return cache_->second;
       } else {
-        throw InvalidIterator("attempt to dereference an invalid iterator");
+        throw InvalidIteratorError("attempt to dereference an invalid iterator");
       }
     }
 

include/mzpeak/util/manager.h

@@ -39,7 +39,7 @@ class Manager final {
   /**
    * Find a file given its `EntityType` and `DataKind`.
    */
-  std::vector<Schema::File>::const_iterator find_file(Schema::EntityType,
+  std::vector<Schema::File>::const_iterator find_file(Schema::EntityType::Type,
                                                       Schema::DataKind::Type) const;
 
   /**

include/mzpeak/util/query.h

@@ -9,6 +9,7 @@ top-level directory of this repository.
 #pragma once
 
 #include <any>
+#include <boost/compat/function_ref.hpp>
 #include <functional>
 #include <variant>
 
@@ -192,12 +193,14 @@ class Query final {
 
   /// A function that when given an column type, should return a single value.
   /// If this isn't possible it should return `Result<value_t>::skip()`.
-  using eval_callback_t = std::function<Result<value_t>(Schema::Column)>;
+  using eval_callback_t =
+      boost::compat::function_ref<Result<value_t>(Schema::Column)>;
 
   /// A func ion that when given an column type should return a min
   /// and max.  If this isn't possible it should return
   /// `Result<range_t>::skip()`.
-  using eval_range_callback_t = std::function<Result<range_t>(Schema::Column)>;
+  using eval_range_callback_t =
+      boost::compat::function_ref<Result<range_t>(Schema::Column)>;
 
   /**
    * Evaluate a query.

src/data/array_index.cpp

@@ -16,6 +16,27 @@ directory of this repository.
 
 namespace MzPeak::Data {
 
+/******************************************************************************/
+// How to compare entries.  Must match EntryChunkFn below.
+//
+// TODO: Unify these two types.
+struct EntryCmpFn {
+  bool operator()(const ArrayIndex::Entry& a, const ArrayIndex::Entry& b) const
+  {
+    return a.array_name < b.array_name && a.array_type < b.array_type &&
+           a.data_type < b.data_type && a.buffer_priority > b.buffer_priority;
+  }
+};
+
+// How to chunk entries.  Must match EntryCmpFn above.
+struct EntryChunkFn {
+  bool operator()(const ArrayIndex::Entry& a, const ArrayIndex::Entry& b) const
+  {
+    return a.array_name == b.array_name && a.array_type == b.array_type &&
+           a.data_type == b.data_type && a.buffer_priority == b.buffer_priority;
+  }
+};
+
 /******************************************************************************/
 ArrayIndex::Layout group_name_to_layout(const std::string& name)
 {
@@ -194,7 +215,7 @@ ArrayIndex::ArrayIndex(EntityType entity_type, const json::object& obj)
         entry.array_name = eo.at("array_name").as_string();
         entry.buffer_format =
             buffer_format_from_string(eo.at("buffer_format").as_string());
-        entry.context = entity_type_from_string(eo.at("context").as_string());
+        entry.context = EntityType(eo.at("context").as_string());
         entry.path = eo.at("path").as_string();
         entry.name = entry.path.substr(prefix_.size() + 1);
 
@@ -248,7 +269,7 @@ ArrayIndex::ArrayIndex(EntityType entity_type, const json::object& obj)
     }
   }
 
-  std::ranges::sort(entries_, {}, &Entry::array_name);
+  std::ranges::sort(entries_, EntryCmpFn());
 }
 
 /******************************************************************************/
@@ -279,10 +300,7 @@ std::optional<std::size_t> ArrayIndex::num_entities() const { return num_entitie
 std::vector<ArrayIndex::Dimension> ArrayIndex::dimensions() const
 {
   std::vector<std::vector<Entry>> groups =
-      entries_ | std::views::chunk_by([](auto& a, auto& b) {
-        return a.array_name == b.array_name && a.data_type == b.data_type &&
-               a.array_type == b.array_type;
-      }) |
+      entries_ | std::views::chunk_by(EntryChunkFn()) |
       std::ranges::to<std::vector<std::vector<Entry>>>();
 
   std::vector<Dimension> result;
@@ -292,8 +310,8 @@ std::vector<ArrayIndex::Dimension> ArrayIndex::dimensions() const
     if (group.empty()) continue;
     auto& head = group[0];
 
-    result.push_back({head.name, head.data_type, head.array_type, head.transform,
-                      std::move(group)});
+    result.push_back({head.name, head.data_type, head.array_type,
+                      head.buffer_priority, head.transform, std::move(group)});
   }
 
   return result;