ppforest2::serialization Namespace Reference

JSON serialization and deserialization for ppforest2 models. More...

Classes
struct	Export
	A model bundled with its export metadata and optional metrics. More...
class	JsonModelVisitor
	Visitor that serializes a model (Tree or Forest) to JSON. More...
class	JsonNodeVisitor
	Visitor that serializes a tree node to JSON. More...

Typedefs
using	json = nlohmann::json
using	Metrics = stats::Metrics
	Re-export of stats::Metrics for serialization callers.
using	Names = types::Names
	Re-export of types::Names for serialization callers.

Functions
bool	has_value (nlohmann::json const &j, std::string_view key)
	True iff j has a non-null value at key.
json	to_json (types::OutcomeVector const &y, types::Names const &names)
	Serialize a prediction vector as JSON.
void	validate_forest_export (nlohmann::json const &j)
	Validate the structural shape of a loaded model Export JSON.
void	validate_tree_export (nlohmann::json const &j)
	Validate the structural shape of a loaded model Export JSON.
Serialization
json	to_json (Model const &model)
json	to_json (TreeNode const &node)
json	to_json (Tree const &tree)
json	to_json (BaggedTree const &tree)
json	to_json (Forest const &forest)
json	to_json (stats::ConfusionMatrix const &cm)
json	to_json (stats::ClassificationMetrics const &cm)
json	to_json (VariableImportance const &vi)
json	to_json (stats::RegressionMetrics const &rm)
json	to_json (Metrics const &metrics)
json	to_json (types::FeatureMatrix const &matrix)
Labeled serialization (uses group names instead of integer codes)
json	to_json (Model const &model, Names const &group_names)
json	to_json (TreeNode const &node, Names const &group_names)
json	to_json (Tree const &tree, Names const &group_names)
json	to_json (BaggedTree const &tree, Names const &group_names)
json	to_json (Forest const &forest, Names const &group_names)
json	to_json (stats::ConfusionMatrix const &cm, Names const &group_names)
json	to_json (stats::ClassificationMetrics const &cm, Names const &group_names)
json	to_json (stats::RegressionMetrics const &rm, Names const &group_names)
json	to_json (Metrics const &metrics, Names const &group_names)
Optional serialization — <tt>nullopt</tt> round-trips as JSON <tt>null</tt>.
json	to_json (std::optional< VariableImportance > const &vi)
json	to_json (std::optional< Metrics > const &metrics)
json	to_json (std::optional< Metrics > const &metrics, Names const &group_names)
Deserialization
template<typename T>
T	from_json (json const &j)
	Deserialize a value block (confusion matrix, VI, metrics, …).
template<>
stats::ConfusionMatrix	from_json< stats::ConfusionMatrix > (json const &j)
template<>
stats::ClassificationMetrics	from_json< stats::ClassificationMetrics > (json const &j)
template<>
VariableImportance	from_json< VariableImportance > (json const &j)
template<>
stats::RegressionMetrics	from_json< stats::RegressionMetrics > (json const &j)
Metrics	metrics_from_json (json const &j, types::Mode mode)
	Deserialize a Metrics block; mode picks the variant alternative.
Stream operators
std::ostream &	operator<< (std::ostream &os, TreeNode const &node)
std::ostream &	operator<< (std::ostream &os, TreeBranch const &condition)
std::ostream &	operator<< (std::ostream &os, TreeLeaf const &response)
std::ostream &	operator<< (std::ostream &os, Tree const &tree)
std::ostream &	operator<< (std::ostream &os, Forest const &forest)
template<typename V>
std::ostream &	operator<< (std::ostream &ostream, std::vector< V > const &vec)
template<typename V, typename C1, typename C2>
std::ostream &	operator<< (std::ostream &ostream, std::set< V, C1, C2 > const &set)
template<typename K, typename V>
std::ostream &	operator<< (std::ostream &ostream, std::map< K, V > const &map)
template<typename V>
std::ostream &	operator<< (std::ostream &ostream, std::map< int, V > const &map)

Detailed Description

JSON serialization and deserialization for ppforest2 models.

Uses nlohmann/json. Provides to_json() for serializing trees, forests, confusion matrices, and variable importance to JSON. Deserialization uses j.get<T>() via nlohmann ADL:

// Serialize and write to file:
auto j = serialization::to_json(forest);
io::json::write_file(j, "model.json");
 
// Read a full export (model + config + meta):
auto j2 = io::json::read_file("model.json");
auto e  = j2.get<Export<Forest>>();  // e.model, e.groups, e.spec
 
// Read bare model block (integer labels):
Tree tree = model_json.get<Tree>();

Typedef Documentation

json

using ppforest2::serialization::json = nlohmann::json

Metrics

using ppforest2::serialization::Metrics = stats::Metrics

Re-export of stats::Metrics for serialization callers.

Names

using ppforest2::serialization::Names = types::Names

Re-export of types::Names for serialization callers.

Function Documentation

from_json()

template<typename T>

T ppforest2::serialization::from_json

(

json const &

j

)

Deserialize a value block (confusion matrix, VI, metrics, …).

Models go through j.get<Export<Model::Ptr>>() (or the typed Export<Tree::Ptr> / Export<Forest::Ptr> overloads) which rehydrate the real TrainingSpec from the JSON's config block. There's no standalone from_json<Tree::Ptr> because constructing a concrete tree without its spec would require a placeholder, which masks the missing config silently.

from_json< stats::ClassificationMetrics >()

template<>

stats::ClassificationMetrics ppforest2::serialization::from_json< stats::ClassificationMetrics >

(

json const &

j

)

from_json< stats::ConfusionMatrix >()

template<>

stats::ConfusionMatrix ppforest2::serialization::from_json< stats::ConfusionMatrix >

(

json const &

j

)

from_json< stats::RegressionMetrics >()

template<>

stats::RegressionMetrics ppforest2::serialization::from_json< stats::RegressionMetrics >

(

json const &

j

)

from_json< VariableImportance >()

template<>

VariableImportance ppforest2::serialization::from_json< VariableImportance >

(

json const &

j

)

has_value()

bool ppforest2::serialization::has_value ( nlohmann::json const & j, std::string_view key )

inline

True iff j has a non-null value at key.

Centralises the "key is present and not null" check used when reading optional fields. Callers that previously wrote j.contains(key) must now use this helper (or equivalent) to tolerate the null that the writer emits for std::nullopt.

metrics_from_json()

Metrics ppforest2::serialization::metrics_from_json	(	json const &	j,
		types::Mode	mode )

Deserialize a Metrics block; mode picks the variant alternative.

operator<<() [1/9]

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	os,
		Forest const &	forest )

operator<<() [2/9]

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	os,
		Tree const &	tree )

operator<<() [3/9]

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	os,
		TreeBranch const &	condition )

operator<<() [4/9]

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	os,
		TreeLeaf const &	response )

operator<<() [5/9]

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	os,
		TreeNode const &	node )

operator<<() [6/9]

template<typename V>

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	ostream,
		std::map< int, V > const &	map )

operator<<() [7/9]

template<typename K, typename V>

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	ostream,
		std::map< K, V > const &	map )

operator<<() [8/9]

template<typename V, typename C1, typename C2>

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	ostream,
		std::set< V, C1, C2 > const &	set )

operator<<() [9/9]

template<typename V>

std::ostream & ppforest2::serialization::operator<<	(	std::ostream &	ostream,
		std::vector< V > const &	vec )

to_json() [1/24]

json ppforest2::serialization::to_json

(

BaggedTree const &

tree

)

to_json() [2/24]

json ppforest2::serialization::to_json	(	BaggedTree const &	tree,
		Names const &	group_names )

to_json() [3/24]

json ppforest2::serialization::to_json

(

Forest const &

forest

)

to_json() [4/24]

json ppforest2::serialization::to_json	(	Forest const &	forest,
		Names const &	group_names )

to_json() [5/24]

json ppforest2::serialization::to_json

(

Metrics const &

metrics

)

to_json() [6/24]

json ppforest2::serialization::to_json	(	Metrics const &	metrics,
		Names const &	group_names )

to_json() [7/24]

json ppforest2::serialization::to_json

(

Model const &

model

)

to_json() [8/24]

json ppforest2::serialization::to_json	(	Model const &	model,
		Names const &	group_names )

to_json() [9/24]

json ppforest2::serialization::to_json

(

stats::ClassificationMetrics const &

cm

)

to_json() [10/24]

json ppforest2::serialization::to_json	(	stats::ClassificationMetrics const &	cm,
		Names const &	group_names )

to_json() [11/24]

json ppforest2::serialization::to_json

(

stats::ConfusionMatrix const &

cm

)

to_json() [12/24]

json ppforest2::serialization::to_json	(	stats::ConfusionMatrix const &	cm,
		Names const &	group_names )

to_json() [13/24]

json ppforest2::serialization::to_json

(

stats::RegressionMetrics const &

rm

)

to_json() [14/24]

json ppforest2::serialization::to_json	(	stats::RegressionMetrics const &	rm,
		Names const &	group_names )

to_json() [15/24]

json ppforest2::serialization::to_json

(

std::optional< Metrics > const &

metrics

)

to_json() [16/24]

json ppforest2::serialization::to_json	(	std::optional< Metrics > const &	metrics,
		Names const &	group_names )

to_json() [17/24]

json ppforest2::serialization::to_json

(

std::optional< VariableImportance > const &

vi

)

to_json() [18/24]

json ppforest2::serialization::to_json

(

Tree const &

tree

)

to_json() [19/24]

json ppforest2::serialization::to_json	(	Tree const &	tree,
		Names const &	group_names )

to_json() [20/24]

json ppforest2::serialization::to_json

(

TreeNode const &

node

)

to_json() [21/24]

json ppforest2::serialization::to_json	(	TreeNode const &	node,
		Names const &	group_names )

to_json() [22/24]

json ppforest2::serialization::to_json

(

types::FeatureMatrix const &

matrix

)

to_json() [23/24]

json ppforest2::serialization::to_json	(	types::OutcomeVector const &	y,
		types::Names const &	names )

Serialize a prediction vector as JSON.

With non-empty names, the predictions are interpreted as integer-coded class labels and mapped to label strings (names[code]). With empty names, the values are passed through verbatim as a flat JSON array — letting callers stay mode-agnostic (regression: continuous floats; classification without label metadata: integer codes).

to_json() [24/24]

json ppforest2::serialization::to_json

(

VariableImportance const &

vi

)

validate_forest_export()

void ppforest2::serialization::validate_forest_export

(

nlohmann::json const &

j

)

Validate the structural shape of a loaded model Export JSON.

Runs before any from_json decoding so failures surface with domain-specific, path-annotated messages instead of nlohmann's cryptic type errors.

Concentrates every cross-field and required-field check in one place. Strategy internals are not validated here — each strategy's own from_json performs that via JsonReader. This keeps the validator responsibilities flat (skeleton + config + meta + variant-specific model block) and avoids turning it into a schema library.

Each validator asserts model_type == "tree" / "forest" matches the call site; mismatches throw a UserError.

Exceptions

UserError

with a dotted path on validation failure.

validate_tree_export()

void ppforest2::serialization::validate_tree_export

(

nlohmann::json const &

j

)

Validate the structural shape of a loaded model Export JSON.

Runs before any from_json decoding so failures surface with domain-specific, path-annotated messages instead of nlohmann's cryptic type errors.

Concentrates every cross-field and required-field check in one place. Strategy internals are not validated here — each strategy's own from_json performs that via JsonReader. This keeps the validator responsibilities flat (skeleton + config + meta + variant-specific model block) and avoids turning it into a schema library.

Each validator asserts model_type == "tree" / "forest" matches the call site; mismatches throw a UserError.

Exceptions

UserError

with a dotted path on validation failure.