SeriesGroupBy

class pandas::SeriesGroupBy

GroupBy class for split-apply-combine operations.

Example

#include <pandas/pandas.h>
using namespace pandas;

// Use SeriesGroupBy
SeriesGroupBy obj;
// ... operations ...

Constructors

Signature	Location	Example
SeriesGroupBy(const Series<T>& series, const Series<GroupT>& by, bool sort = true)	pd_series_groupby.h:78
SeriesGroupBy(std::shared_ptr<Series<T>> owned, const Series<GroupT>& by, bool sort = true)	pd_series_groupby.h:89

Construction

Signature	Return Type	Location	Example
Series<T> create_result_series(const std::vector<T>& values) const	Series<T>	pd_series_groupby.h:1157
Series<double> create_result_series_double(const std::vector<double>& values) const	Series<double>	pd_series_groupby.h:1174
Series<int64_t> create_result_series_int64(const std::vector<int64_t>& values) const	Series<int64_t>	pd_series_groupby.h:1193

Indexing / Selection

Signature	Return Type	Location	Example
Series<T> first() const	Series<T>	pd_series_groupby.h:380	View
Series<T> get_group(const GroupT& key) const	Series<T>	pd_series_groupby.h:1063	View
std::optional<std::string> get_index_name() const	std::optional<std::string>	pd_series_groupby.h:128	View
cat_values, cats, false, get_index_name())	cat_values, cats, false,	pd_series_groupby.h:1108	View
std::optional<std::string> get_series_name() const	std::optional<std::string>	pd_series_groupby.h:131
pandas::Result idxmin_with_dtype() const	pandas::Result	pd_series_groupby.h:728	View
Series<T> last() const	Series<T>	pd_series_groupby.h:472	View

Data Manipulation

Signature	Return Type	Location	Example
void set_index_name(const std::string& name)	void	pd_series_groupby.h:120	View

Statistics

Signature	Return Type	Location	Example
Series<int64_t> count() const	Series<int64_t>	pd_series_groupby.h:276	View
Series<double> cummax() const	Series<double>	pd_series_groupby.h:852	View
Series<double> cummin() const	Series<double>	pd_series_groupby.h:877	View
Series<double> cumprod() const	Series<double>	pd_series_groupby.h:827	View
Series<double> cumsum() const	Series<double>	pd_series_groupby.h:802	View
Series<T> max() const	Series<T>	pd_series_groupby.h:342	View
Series<double> mean() const	Series<double>	pd_series_groupby.h:240	View
Series<double> median() const	Series<double>	pd_series_groupby.h:589	View
Series<T> min() const	Series<T>	pd_series_groupby.h:304	View
Series<int64_t> nunique(bool dropna = true) const	Series<int64_t>	pd_series_groupby.h:955	View
Series<double> std_(int ddof = 1) const	Series<double>	pd_series_groupby.h:510	View
auto sum() const	auto	pd_series_groupby.h:180	View
Series<int64_t> sum_int64_bool_() const	Series<int64_t>	pd_series_groupby.h:220	View
Series<double> var(int ddof = 1) const	Series<double>	pd_series_groupby.h:550	View

Aggregation

Signature	Return Type	Location	Example
Series<double> agg(const std::string& func) const	Series<double>	pd_series_groupby.h:634	View
DataFrame agg(const std::vector<std::string>& funcs) const	DataFrame	pd_series_groupby.h:708	View
pandas::Result agg_with_dtype(const std::string& how) const	pandas::Result	pd_series_groupby.h:716	View
pandas::Result agg_with_dtype_list(const std::vector<std::string>& funcs) const	pandas::Result	pd_series_groupby.h:722	View
auto apply(Func&& func) const -> Series<decltype(func(std::declval<Series<T>>()))>	auto	pd_series_groupby.h:737	View
void apply_result_index(Series<ResultT>& result) const	void	pd_series_groupby.h:1098	View
Series<T> transform(Func&& func) const	Series<T>	pd_series_groupby.h:767	View

Arithmetic

Signature	Return Type	Location	Example
const std::vector<std::string>& multiindex_names() const	const std::vector<std::string>&	pd_series_groupby.h:137	View

Comparison

Signature	Return Type	Location	Example
std::vector<std::vector<std::string>> level_arrays(nlevels)	std::vector<std::vector<std::string>>	pd_series_groupby.h:1114

Time Series

Signature	Return Type	Location	Example
Series<double> diff(int periods = 1) const	Series<double>	pd_series_groupby.h:903	View
Series<double> shift(int periods = 1) const	Series<double>	pd_series_groupby.h:935	View

Other Methods

Signature	Return Type	Location	Example
void build_groups()	void	pd_series_groupby.h:1136
const std::vector<std::string>& categorical_categories() const	const std::vector<std::string>&	pd_series_groupby.h:143	View
Series<int64_t> cumcount(bool ascending = true) const	Series<int64_t>	pd_series_groupby.h:991
const std::map<GroupT, std::vector<size_t>>& group_indices() const	const std::map<GroupT, std::vector<size_t>>&	pd_series_groupby.h:107
const std::vector<GroupT>& group_keys_order() const	const std::vector<GroupT>&	pd_series_groupby.h:112	View
const std::string& grouper_dtype() const	const std::string&	pd_series_groupby.h:149	View
std::vector<GroupT> groups() const	std::vector<GroupT>	pd_series_groupby.h:1046	View
const std::map<GroupT, std::vector<size_t>>& indices() const	const std::map<GroupT, std::vector<size_t>>&	pd_series_groupby.h:1054	View
Series<int64_t> ngroup(bool ascending = true) const	Series<int64_t>	pd_series_groupby.h:1015
size_t ngroups() const	size_t	pd_series_groupby.h:1038	View
Series<T> nth(int n, const std::string& dropna_mode = "") const	Series<T>	pd_series_groupby.h:419	View
const Series<T>& series() const	const Series<T>&	pd_series_groupby.h:117	View
bool series_name_is_int() const	bool	pd_series_groupby.h:125
void set_all_categories(const std::vector<std::string>& categories)	void	pd_series_groupby.h:153
void set_categorical_categories(const std::vector<std::string>& cats)	void	pd_series_groupby.h:140	View
void set_grouper_dtype(const std::string& dtype)	void	pd_series_groupby.h:146
void set_multiindex_names(const std::vector<std::string>& names)	void	pd_series_groupby.h:134	View
void set_series_name(const std::string& name)	void	pd_series_groupby.h:123
void set_series_name_is_int(bool flag)	void	pd_series_groupby.h:124
Series<int64_t> size() const	Series<int64_t>	pd_series_groupby.h:1082	View
std::string source_dtype() const	std::string	pd_series_groupby.h:95
std::string source_dtype_full() const	std::string	pd_series_groupby.h:102

Code Examples

The following examples are extracted from the test suite.

first (pd_test_1_all.cpp:11616)

        void pd_test_groupby_first_last() {
            std::cout << "========= GroupBy first/last ====================";

            std::map<std::string, std::vector<double>> data = {
                {"category", {1.0, 1.0, 2.0, 2.0}},
                {"value", {10.0, 20.0, 30.0, 40.0}}
            };
            pandas::DataFrame df(data);

            auto first_result = df.groupby("category").first();
            auto last_result = df.groupby("category").last();

            // First for group 1: 10, group 2: 30
            // Last for group 1: 20, group 2: 40
            double first1 = std::stod(first_result["value"].get_value_str(0));
            double first2 = std::stod(first_result["value"].get_value_str(1));

            bool passed = ((std::abs(first1 - 10.0) < 0.001 && std::abs(first2 - 30.0) < 0.001) ||
                          (std::abs(first1 - 30.0) < 0.001 && std::abs(first2 - 10.0) < 0.001));
            if (!passed) {

get_group (pd_test_2_all.cpp:20487)

        ++g_fail;
    }
}

static bool approx_eq(double a, double b, double tol = 1e-9) {
    if (std::isnan(a) && std::isnan(b)) return true;
    return std::abs(a - b) < tol;
}

// =====================================================================
// Test: get_group() with exclude_cols removes groupby columns
// =====================================================================

void pd_test_groupby_apply_get_group_exclude() {
    std::cout << "  -- pd_test_groupby_apply_get_group_exclude --" << std::endl;

    pandas::DataFrame df;
    df.add_column("key", std::vector<std::string>{"a", "a", "b", "b"});
    df.add_column("val1", std::vector<double>{1.0, 2.0, 3.0, 4.0});
    df.add_column("val2", std::vector<double>{10.0, 20.0, 30.0, 40.0});

get_index_name (pd_test_3_all.cpp:23398)

    std::vector<std::optional<std::string>> level_names = {"first", "second"};
    auto mi = pandas::MultiIndex::from_arrays<std::string>(level_values, level_names);
    s.set_multiindex(mi);

    auto gb = s.groupby_by_level(static_cast<size_t>(0), true);
    if (gb.group_keys_order().size() != 2)
        throw std::runtime_error("expected 2 groups");
    auto sums = gb.sum();
    if (sums[0] != 30.0 || sums[1] != 70.0)
        throw std::runtime_error("sum mismatch");
    if (!gb.get_index_name().has_value() || *gb.get_index_name() != "first")
        throw std::runtime_error("index name mismatch");

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_groupby_level_multi() {
    std::cout << "========= groupby_by_level(multi) =====================";

    pandas::Series<numpy::float64> s({1.0, 2.0, 3.0, 4.0});
    std::vector<std::vector<std::string>> level_values = {

get_index_name (pd_test_3_all.cpp:23398)

    std::vector<std::optional<std::string>> level_names = {"first", "second"};
    auto mi = pandas::MultiIndex::from_arrays<std::string>(level_values, level_names);
    s.set_multiindex(mi);

    auto gb = s.groupby_by_level(static_cast<size_t>(0), true);
    if (gb.group_keys_order().size() != 2)
        throw std::runtime_error("expected 2 groups");
    auto sums = gb.sum();
    if (sums[0] != 30.0 || sums[1] != 70.0)
        throw std::runtime_error("sum mismatch");
    if (!gb.get_index_name().has_value() || *gb.get_index_name() != "first")
        throw std::runtime_error("index name mismatch");

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_groupby_level_multi() {
    std::cout << "========= groupby_by_level(multi) =====================";

    pandas::Series<numpy::float64> s({1.0, 2.0, 3.0, 4.0});
    std::vector<std::vector<std::string>> level_values = {

idxmin_with_dtype (pd_test_5_all.cpp:95397)

void case_701_dfgb_idxmin_rangeindex(int& local_fail) {
    std::cout << "-- case_701_dfgb_idxmin_rangeindex\n";
    // Default RangeIndex (int64). Result columns must keep int64 dtype.
    pandas::DataFrame df;
    df.add_column<double>("v", std::vector<double>{3.0, 1.0, 2.0, 0.5});
    df.add_column<int64_t>("key", std::vector<int64_t>{0, 0, 1, 1});
    auto gb = df.groupby("key");
    pandas::DataFrame out;
    std::string err;
    try { out = gb.idxmin_with_dtype(); }
    catch (const std::exception& e) { err = e.what(); }
    catch (...) { err = "<unknown>"; }
    pandas_tests::check(err.empty(),
        "C_26_case_701_dfgb_idxmin_rangeindex()_no_throw", local_fail);
    if (!err.empty()) { std::cout << "  err: " << err << "\n"; return; }
    std::string got = df_col_dtype(out, "v");
    bool ok = (got == "int64");
    pandas_tests::check(ok,
        "C_26_case_701_dfgb_idxmin_rangeindex()_dtype", local_fail);
    if (!ok) std::cout << "  got=[" << got << "] expected=[int64]\n";

last (pd_test_1_all.cpp:11617)

        void pd_test_groupby_first_last() {
            std::cout << "========= GroupBy first/last ====================";

            std::map<std::string, std::vector<double>> data = {
                {"category", {1.0, 1.0, 2.0, 2.0}},
                {"value", {10.0, 20.0, 30.0, 40.0}}
            };
            pandas::DataFrame df(data);

            auto first_result = df.groupby("category").first();
            auto last_result = df.groupby("category").last();

            // First for group 1: 10, group 2: 30
            // Last for group 1: 20, group 2: 40
            double first1 = std::stod(first_result["value"].get_value_str(0));
            double first2 = std::stod(first_result["value"].get_value_str(1));

            bool passed = ((std::abs(first1 - 10.0) < 0.001 && std::abs(first2 - 30.0) < 0.001) ||
                          (std::abs(first1 - 30.0) < 0.001 && std::abs(first2 - 10.0) < 0.001));
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_groupby_first_last() : first values incorrect" << std::endl;

set_index_name (pd_test_2_all.cpp:20842)

void test_sgb_apply_result_index_categorical() {
    std::cout << "  -- test_sgb_apply_result_index_categorical --" << std::endl;

    std::vector<numpy::float64> values = {5.0, 10.0};
    pandas::Series<std::string> by({"A", "B"});
    pandas::Series<numpy::float64> data(values);

    auto sgb = data.groupby(by);
    sgb.set_categorical_categories({"A", "B", "C"});
    sgb.set_index_name("cat_key");

    pandas::Series<numpy::float64> result(values);
    std::vector<std::string> idx_labels = {"A", "B"};
    result.set_index(std::make_unique<pandas::Index<std::string>>(idx_labels));

    sgb.apply_result_index(result);

    // Should have CategoricalIndex (dtype_name() returns "category")
    check(result.index().dtype_name() == "category", "is_categorical_index");
}

count (pd_test_1_all.cpp:66)

        if (arr.is_na(0)) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_na_handling() : is_na(0) should be false" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_na_handling failed: is_na(0) should be false");
        }

        if (!arr.has_na()) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_na_handling() : has_na() should be true" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_na_handling failed: has_na() should be true");
        }

        if (arr.count() != 2) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_na_handling() : count() should be 2" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_na_handling failed: count() should be 2");
        }

        std::cout << " -> tests passed" << std::endl;
    }

    void pd_test_boolean_array_kleene_and() {
        std::cout << "========= BooleanArray: Kleene AND ======================= ";

cummax (pd_test_1_all.cpp:5152)

            // cummin: [1, 1, 1, 1]
            auto cmin = df.cummin();
            val = cmin["A"].get_value_str(3);
            passed = std::abs(std::stod(val) - 1.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cummin failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cummin failed");
            }

            // cummax: [1, 2, 3, 4]
            auto cmax = df.cummax();
            val = cmax["A"].get_value_str(2);
            passed = std::abs(std::stod(val) - 3.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cummax failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cummax failed");
            }

            std::cout << " -> tests passed" << std::endl;
        }

cummin (pd_test_1_all.cpp:5143)

            // cumprod: [1, 2, 6, 24]
            auto cp = df.cumprod();
            val = cp["A"].get_value_str(3);
            passed = std::abs(std::stod(val) - 24.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cumprod failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cumprod failed");
            }

            // cummin: [1, 1, 1, 1]
            auto cmin = df.cummin();
            val = cmin["A"].get_value_str(3);
            passed = std::abs(std::stod(val) - 1.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cummin failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cummin failed");
            }

            // cummax: [1, 2, 3, 4]
            auto cmax = df.cummax();
            val = cmax["A"].get_value_str(2);

cumprod (pd_test_1_all.cpp:5134)

            // cumsum: [1, 3, 6, 10]
            auto cs = df.cumsum();
            std::string val = cs["A"].get_value_str(2);
            bool passed = std::abs(std::stod(val) - 6.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cumsum failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cumsum failed");
            }

            // cumprod: [1, 2, 6, 24]
            auto cp = df.cumprod();
            val = cp["A"].get_value_str(3);
            passed = std::abs(std::stod(val) - 24.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cumprod failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cumprod failed");
            }

            // cummin: [1, 1, 1, 1]
            auto cmin = df.cummin();
            val = cmin["A"].get_value_str(3);

cumsum (pd_test_1_all.cpp:5125)

        }

        void pd_test_arithmetic_dataframe_cumulative() {
            std::cout << "========= DataFrame cumulative ==================";

            std::map<std::string, std::vector<double>> data;
            data["A"] = {1.0, 2.0, 3.0, 4.0};
            pandas::DataFrame df(data);

            // cumsum: [1, 3, 6, 10]
            auto cs = df.cumsum();
            std::string val = cs["A"].get_value_str(2);
            bool passed = std::abs(std::stod(val) - 6.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_cumulative() : cumsum failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_cumulative failed: cumsum failed");
            }

            // cumprod: [1, 2, 6, 24]
            auto cp = df.cumprod();
            val = cp["A"].get_value_str(3);

max (pd_test_1_all.cpp:771)

        pandas::CategoricalArray arr = pandas::CategoricalArray::from_codes(codes, cats, true);  // ordered

        // Test min
        std::optional<std::string> min_val = arr.min();
        if (!min_val.has_value() || *min_val != "low") {
            std::cout << "  [FAIL] : in pd_test_categorical_array_ordered_operations() : min != 'low'" << std::endl;
            throw std::runtime_error("pd_test_categorical_array_ordered_operations failed: min != 'low'");
        }

        // Test max
        std::optional<std::string> max_val = arr.max();
        if (!max_val.has_value() || *max_val != "high") {
            std::cout << "  [FAIL] : in pd_test_categorical_array_ordered_operations() : max != 'high'" << std::endl;
            throw std::runtime_error("pd_test_categorical_array_ordered_operations failed: max != 'high'");
        }

        // Test unordered throws for min/max
        pandas::CategoricalArray unordered = arr.as_unordered();
        bool threw = false;
        try {
            unordered.min();

mean (pd_test_1_all.cpp:282)

            std::optional<bool>(true),
            std::optional<bool>(true)
        });

        auto s = arr.sum();
        if (!s.has_value() || s.value() != 3) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_reductions() : sum should be 3" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_reductions failed: sum");
        }

        auto m = arr.mean();
        if (!m.has_value() || std::abs(m.value() - 0.75) > 0.001) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_reductions() : mean should be 0.75" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_reductions failed: mean");
        }

        std::cout << " -> tests passed" << std::endl;
    }

    void pd_test_boolean_array_dtype() {
        std::cout << "========= BooleanArray: dtype ======================= ";

median (pd_test_1_all.cpp:20910)

                throw std::runtime_error("pd_test_expanding_var failed: expanding var values incorrect");
            }

            std::cout << " -> tests passed" << std::endl;
        }

        void pd_test_expanding_median() {
            std::cout << "========= Expanding median ======================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0, 5.0});
            auto result = s.expanding().median();

            // Expanding median: 1, 1.5, 2, 2.5, 3
            bool passed = std::abs(result[0] - 1.0) < 0.001 &&
                          std::abs(result[1] - 1.5) < 0.001 &&
                          std::abs(result[2] - 2.0) < 0.001 &&
                          std::abs(result[3] - 2.5) < 0.001 &&
                          std::abs(result[4] - 3.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_expanding_median() : expanding median values incorrect" << std::endl;
                throw std::runtime_error("pd_test_expanding_median failed: expanding median values incorrect");

min (pd_test_1_all.cpp:764)

    }

    void pd_test_categorical_array_ordered_operations() {
        std::cout << "========= CategoricalArray: ordered operations (min/max) ======================= ";

        std::vector<std::string> cats = {"low", "medium", "high"};
        std::vector<numpy::int32> codes = {0, 2, 1, 0, -1};  // low, high, medium, low, NA
        pandas::CategoricalArray arr = pandas::CategoricalArray::from_codes(codes, cats, true);  // ordered

        // Test min
        std::optional<std::string> min_val = arr.min();
        if (!min_val.has_value() || *min_val != "low") {
            std::cout << "  [FAIL] : in pd_test_categorical_array_ordered_operations() : min != 'low'" << std::endl;
            throw std::runtime_error("pd_test_categorical_array_ordered_operations failed: min != 'low'");
        }

        // Test max
        std::optional<std::string> max_val = arr.max();
        if (!max_val.has_value() || *max_val != "high") {
            std::cout << "  [FAIL] : in pd_test_categorical_array_ordered_operations() : max != 'high'" << std::endl;
            throw std::runtime_error("pd_test_categorical_array_ordered_operations failed: max != 'high'");

nunique (pd_test_1_all.cpp:10604)

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_extension_index_nunique() {
    std::cout << "========= nunique =========================";

    pandas::CategoricalArray arr({"a", "b", "a", "c", "b", std::nullopt});
    pandas::CategoricalIndex idx(arr);

    bool passed = (idx.nunique(true) == 3 && idx.nunique(false) == 4);
    if (!passed) {
        std::cout << "  [FAIL] : in pd_test_extension_index_nunique() : nunique check failed" << std::endl;
        throw std::runtime_error("pd_test_extension_index_nunique failed");
    }

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_extension_index_factorize() {
    std::cout << "========= factorize =========================";

std_ (pd_test_1_all.cpp:20752)

                throw std::runtime_error("pd_test_rolling_min_periods failed: with min_periods=1, idx 1 should be 3.0");
            }

            std::cout << " -> tests passed" << std::endl;
        }

        void pd_test_rolling_std() {
            std::cout << "========= Rolling std ===========================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0, 5.0});
            auto result = s.rolling(3).std_();

            // std([1,2,3]) = 1.0 (ddof=1)
            // std([2,3,4]) = 1.0
            // std([3,4,5]) = 1.0
            bool passed = std::abs(result[2] - 1.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_rolling_std() : rolling std should be 1.0" << std::endl;
                throw std::runtime_error("pd_test_rolling_std failed: rolling std should be 1.0");
            }

sum (pd_test_1_all.cpp:276)

        }

        // Test sum/mean
        pandas::BooleanArray arr({
            std::optional<bool>(true),
            std::optional<bool>(false),
            std::optional<bool>(true),
            std::optional<bool>(true)
        });

        auto s = arr.sum();
        if (!s.has_value() || s.value() != 3) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_reductions() : sum should be 3" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_reductions failed: sum");
        }

        auto m = arr.mean();
        if (!m.has_value() || std::abs(m.value() - 0.75) > 0.001) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_reductions() : mean should be 0.75" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_reductions failed: mean");
        }

sum_int64_bool_ (pd_test_5_all.cpp:55457)

    check_col_dtype("caseG3", df, 0, "float64", local_fail);
}

static void f_seriesgroupby_agg_list_dtype_12_b3d2f7_caseH1_bool_sum(int& local_fail) {
    std::cout << "-- caseH1_bool_sum\n";
    pandas::Series<bool> v({true, false, true, true});
    pandas::Series<std::string> by({"a", "a", "b", "b"});
    auto sgb = v.groupby(by);
    pandas::DataFrame df = sgb.agg(std::vector<std::string>{"sum"});
    check_ncols("caseH1", df, 1, local_fail);
    // Plan 21: pandas widens bool sum to int64 — fixed via sum_int64_bool_().
    check_col_dtype("caseH1", df, 0, "int64", local_fail);
}

static void f_seriesgroupby_agg_list_dtype_12_b3d2f7_caseH2_bool_first(int& local_fail) {
    std::cout << "-- caseH2_bool_first\n";
    pandas::Series<bool> v({true, false, true, true});
    pandas::Series<std::string> by({"a", "a", "b", "b"});
    auto sgb = v.groupby(by);
    pandas::DataFrame df = sgb.agg(std::vector<std::string>{"first"});
    check_ncols("caseH2", df, 1, local_fail);

var (pd_test_1_all.cpp:20890)

                throw std::runtime_error("pd_test_expanding_std failed: expanding std values incorrect");
            }

            std::cout << " -> tests passed" << std::endl;
        }

        void pd_test_expanding_var() {
            std::cout << "========= Expanding var =========================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0, 5.0});
            auto result = s.expanding().var();

            // Expanding var (ddof=1): NaN, 0.5, 1.0, 1.6667, 2.5
            bool passed = std::isnan(result[0]) &&
                          std::abs(result[1] - 0.5) < 0.001 &&
                          std::abs(result[2] - 1.0) < 0.001 &&
                          std::abs(result[3] - 1.6667) < 0.001 &&
                          std::abs(result[4] - 2.5) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_expanding_var() : expanding var values incorrect" << std::endl;
                throw std::runtime_error("pd_test_expanding_var failed: expanding var values incorrect");

agg (pd_test_1_all.cpp:11100)

        }

        void pd_test_func_apply_series_agg() {
            std::cout << "========= Series agg ==================================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0, 5.0}, "values");

            bool passed = true;

            // Test string-based aggregation
            auto sum_result = s.agg("sum");
            if (!sum_result.has_value() || !approx_equal(sum_result.value(), 15.0)) {
                passed = false;
                std::cout << "  [FAIL] : in pd_test_func_apply_series_agg() : sum failed" << std::endl;
                throw std::runtime_error("pd_test_func_apply_series_agg failed: sum failed");
            }

            auto mean_result = s.agg("mean");
            if (!mean_result.has_value() || !approx_equal(mean_result.value(), 3.0)) {
                passed = false;
                std::cout << "  [FAIL] : in pd_test_func_apply_series_agg() : mean failed" << std::endl;

agg (pd_test_1_all.cpp:11100)

        }

        void pd_test_func_apply_series_agg() {
            std::cout << "========= Series agg ==================================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0, 5.0}, "values");

            bool passed = true;

            // Test string-based aggregation
            auto sum_result = s.agg("sum");
            if (!sum_result.has_value() || !approx_equal(sum_result.value(), 15.0)) {
                passed = false;
                std::cout << "  [FAIL] : in pd_test_func_apply_series_agg() : sum failed" << std::endl;
                throw std::runtime_error("pd_test_func_apply_series_agg failed: sum failed");
            }

            auto mean_result = s.agg("mean");
            if (!mean_result.has_value() || !approx_equal(mean_result.value(), 3.0)) {
                passed = false;
                std::cout << "  [FAIL] : in pd_test_func_apply_series_agg() : mean failed" << std::endl;

agg_with_dtype (pd_test_5_all.cpp:94652)

static void run_dfgb_case(const std::string& fn,
                          const std::string& col,
                          const std::string& expected_dtype,
                          const std::string& label,
                          int& local_fail) {
    pandas::DataFrame df = make_mixed_df();
    auto gb = df.groupby("key");
    pandas::DataFrame out;
    std::string err;
    try {
        out = gb.agg_with_dtype(fn);
    } catch (const std::exception& e) {
        err = e.what();
    } catch (...) {
        err = "<unknown>";
    }
    pandas_tests::check(err.empty(),
        label + "_no_throw",
        local_fail);
    if (!err.empty()) {
        std::cout << "  err: " << err << "\n";

agg_with_dtype_list (pd_test_5_all.cpp:94682)

static void run_dfgb_list_case(const std::vector<std::string>& fns,
                               const std::string& src_col,
                               const std::vector<std::string>& expected,
                               const std::string& label,
                               int& local_fail) {
    pandas::DataFrame df = make_mixed_df();
    auto gb = df.groupby("key");
    pandas::DataFrame out;
    std::string err;
    try {
        out = gb.agg_with_dtype_list(fns);
    } catch (const std::exception& e) {
        err = e.what();
    } catch (...) {
        err = "<unknown>";
    }
    pandas_tests::check(err.empty(),
        label + "_no_throw",
        local_fail);
    if (!err.empty()) {
        std::cout << "  err: " << err << "\n";

apply (pd_test_1_all.cpp:11244)

        void pd_test_func_apply_dataframe_apply_axis0() {
            std::cout << "========= DataFrame apply axis=0 ======================";

            std::map<std::string, std::vector<double>> data = {
                {"A", {1.0, 2.0, 3.0}},
                {"B", {4.0, 5.0, 6.0}}
            };
            pandas::DataFrame df(data);

            // apply axis=0 applies function to each column
            auto result = df.apply([](const std::vector<double>& col) {
                return std::accumulate(col.begin(), col.end(), 0.0);
            }, 0);

            bool passed = true;

            // Plan F·dtype: axis=0 reduce now returns a single "result" column
            // with the original column names ("A", "B") as the row index.
            // Sum of A: 1+2+3=6, Sum of B: 4+5+6=15
            const auto& result_col = result["result"];
            double sum_a = std::stod(result_col.get_value_str(0));

apply_result_index (pd_test_2_all.cpp:20781)

    pandas::Series<std::string> by(keys);
    pandas::Series<numpy::float64> data(values);

    auto sgb = data.groupby(by);
    sgb.set_multiindex_names({"level0", "level1"});

    // Create a "result" series with composite index
    pandas::Series<numpy::float64> result(values);
    result.set_index(std::make_unique<pandas::Index<std::string>>(keys));

    sgb.apply_result_index(result);

    // Should now have a MultiIndex
    check(result.has_multiindex(), "has_multiindex");
    check(result.multiindex().nlevels() == 2, "nlevels_2");
}

void test_sgb_apply_result_index_3level() {
    std::cout << "  -- test_sgb_apply_result_index_3level --" << std::endl;

    using std::string;

transform (pd_test_1_all.cpp:11071)

            std::cout << " -> tests passed" << std::endl;
        }

        void pd_test_func_apply_series_transform() {
            std::cout << "========= Series transform ============================";

            pandas::Series<double> s({1.0, 2.0, 3.0, 4.0}, "values");

            // Transform must return same shape
            auto result = s.transform([](double x) { return x * 2 + 1; });

            bool passed = true;
            if (result.size() != s.size()) {
                passed = false;
                std::cout << "  [FAIL] : in pd_test_func_apply_series_transform() : size changed" << std::endl;
                throw std::runtime_error("pd_test_func_apply_series_transform failed: size changed");
            }

            std::vector<double> expected = {3.0, 5.0, 7.0, 9.0};
            for (size_t i = 0; i < result.size(); ++i) {

multiindex_names (pd_test_3_all.cpp:23419)

        {"a", "a", "b", "b"}, {"x", "y", "x", "y"}
    };
    std::vector<std::optional<std::string>> level_names = {"L0", "L1"};
    auto mi = pandas::MultiIndex::from_arrays<std::string>(level_values, level_names);
    s.set_multiindex(mi);

    std::vector<size_t> levels = {0, 1};
    auto gb = s.groupby_by_level(levels, true);
    if (gb.group_keys_order().size() != 4)
        throw std::runtime_error("expected 4 composite groups");
    if (gb.multiindex_names().size() != 2 || gb.multiindex_names()[0] != "L0" || gb.multiindex_names()[1] != "L1")
        throw std::runtime_error("multiindex names mismatch");

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_groupby_by_index() {
    std::cout << "========= groupby_by_index() ==========================";

    pandas::Series<numpy::float64> s({10.0, 20.0, 30.0});
    s.set_index(pandas::Index<std::string>({"a", "b", "a"}));

diff (pd_test_1_all.cpp:5171)

        }

        void pd_test_arithmetic_dataframe_diff_shift() {
            std::cout << "========= DataFrame diff/shift ==================";

            std::map<std::string, std::vector<double>> data;
            data["A"] = {1.0, 3.0, 6.0, 10.0};
            pandas::DataFrame df(data);

            // diff: [NaN, 2, 3, 4]
            auto d = df.diff();
            std::string val = d["A"].get_value_str(1);
            bool passed = std::abs(std::stod(val) - 2.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_diff_shift() : diff failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_diff_shift failed: diff failed");
            }

            // First element should be NaN
            val = d["A"].get_value_str(0);
            passed = std::isnan(std::stod(val));

shift (pd_test_1_all.cpp:5188)

            // First element should be NaN
            val = d["A"].get_value_str(0);
            passed = std::isnan(std::stod(val));
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_diff_shift() : diff NaN failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_diff_shift failed: diff NaN failed");
            }

            // shift: [NaN, 1, 3, 6]
            auto s = df.shift();
            val = s["A"].get_value_str(1);
            passed = std::abs(std::stod(val) - 1.0) < 0.001;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_arithmetic_dataframe_diff_shift() : shift failed" << std::endl;
                throw std::runtime_error("pd_test_arithmetic_dataframe_diff_shift failed: shift failed");
            }

            std::cout << " -> tests passed" << std::endl;
        }

categorical_categories (pd_test_3_all.cpp:23513)

    pandas::CategoricalArray cat({"a", "b", "a"}, {"a", "b", "c"});

    auto gb_obs = s.groupby_by_categorical(cat, true, true);
    if (gb_obs.group_keys_order().size() != 2)
        throw std::runtime_error("expected 2 observed groups");

    auto gb_all = s.groupby_by_categorical(cat, true, false);
    if (gb_all.group_keys_order().size() != 3)
        throw std::runtime_error("expected 3 groups with observed=false");

    if (gb_obs.categorical_categories().size() != 3)
        throw std::runtime_error("categorical_categories not set");

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_groupby_by_labels() {
    std::cout << "========= groupby_by_labels() =========================";

    pandas::Series<numpy::float64> s({1.0, 2.0, 3.0, 4.0});
    std::vector<std::string> labels = {"X", "Y", "X", "Y"};

group_keys_order (pd_test_3_all.cpp:23393)

    pandas::Series<numpy::float64> s({10.0, 20.0, 30.0, 40.0});
    std::vector<std::vector<std::string>> level_values = {
        {"a", "a", "b", "b"}, {"x", "y", "x", "y"}
    };
    std::vector<std::optional<std::string>> level_names = {"first", "second"};
    auto mi = pandas::MultiIndex::from_arrays<std::string>(level_values, level_names);
    s.set_multiindex(mi);

    auto gb = s.groupby_by_level(static_cast<size_t>(0), true);
    if (gb.group_keys_order().size() != 2)
        throw std::runtime_error("expected 2 groups");
    auto sums = gb.sum();
    if (sums[0] != 30.0 || sums[1] != 70.0)
        throw std::runtime_error("sum mismatch");
    if (!gb.get_index_name().has_value() || *gb.get_index_name() != "first")
        throw std::runtime_error("index name mismatch");

    std::cout << " -> tests passed" << std::endl;
}

grouper_dtype (pd_test_3_all.cpp:23493)

    std::cout << "========= groupby_by_numeric() ========================";

    pandas::Series<numpy::float64> s({10.0, 20.0, 30.0, 40.0});
    pandas::Series<numpy::float64> by_s({1.0, 2.0, 1.0, 2.0});
    auto gb = s.groupby_by_numeric(by_s, true);
    if (gb.group_keys_order().size() != 2)
        throw std::runtime_error("expected 2 groups");
    auto sums = gb.sum();
    if (sums[0] != 40.0 || sums[1] != 60.0)
        throw std::runtime_error("sum mismatch");
    if (gb.grouper_dtype() != "float64")
        throw std::runtime_error("grouper_dtype mismatch");

    std::cout << " -> tests passed" << std::endl;
}

void pd_test_groupby_by_categorical() {
    std::cout << "========= groupby_by_categorical() ====================";

    pandas::Series<numpy::float64> s({10.0, 20.0, 30.0});
    pandas::CategoricalArray cat({"a", "b", "a"}, {"a", "b", "c"});

groups (pd_test_2_all.cpp:20864)

// =====================================================================
// Per-group expanding tests
// =====================================================================

void test_series_groupby_expanding_sum() {
    std::cout << "  -- test_series_groupby_expanding_sum --" << std::endl;

    // Two groups: A=[1,2,3], B=[10,20]
    std::vector<numpy::float64> vals = {1.0, 10.0, 2.0, 20.0, 3.0};
    pandas::Series<numpy::float64> data(vals);
    pandas::Series<std::string> groups({"A", "B", "A", "B", "A"});

    auto sgb = data.groupby(groups);
    pandas::SeriesGroupByExpandingWindow ew(sgb, 1);
    auto result = ew.sum();

    check(result.size() == 5, "size_5");
    // A group: expanding sum = 1, 3, 6
    // B group: expanding sum = 10, 30
    // Original order: [A:1, B:10, A:3, B:30, A:6]
    check(approx_eq(result[0], 1.0), "A_exp_sum_0");

indices (pd_test_1_all.cpp:14921)

                passed = passed && r2_tup1[0] == "b" && r2_tup1[1] == "x";
                passed = passed && r2_tup2[0] == "c" && r2_tup2[1] == "x";
            }

            // Test empty vector (no deletion)
            std::cout << "  Test 3: Empty delete_(std::vector<size_t>{})..." << std::endl;
            auto result3 = mi.delete_(std::vector<size_t>{});
            std::cout << "    Result size: " << result3.size() << " (expected " << mi.size() << ")" << std::endl;
            passed = passed && result3.size() == mi.size();

            // Test duplicate indices (should be deduplicated)
            std::cout << "  Test 4: Duplicate delete_({1, 1, 2})..." << std::endl;
            auto result4 = mi.delete_({1, 1, 2});
            std::cout << "    Result size: " << result4.size() << " (expected 3)" << std::endl;
            passed = passed && result4.size() == 3;

            // Test deleting all elements
            std::cout << "  Test 5: Delete all delete_({0,1,2,3,4})..." << std::endl;
            auto result5 = mi.delete_({0, 1, 2, 3, 4});
            std::cout << "    Result size: " << result5.size() << " (expected 0)" << std::endl;
            passed = passed && result5.size() == 0;

ngroups (pd_test_1_all.cpp:11497)

            // Create DataFrame with category column
            std::map<std::string, std::vector<double>> data = {
                {"category", {1.0, 1.0, 2.0, 2.0, 2.0}},
                {"value", {10.0, 20.0, 30.0, 40.0, 50.0}}
            };
            pandas::DataFrame df(data);

            // Test groupby
            auto grouped = df.groupby("category");

            bool passed = grouped.ngroups() == 2;
            if (!passed) {
                std::cout << "  [FAIL] : in pd_test_groupby_basic() : ngroups should be 2" << std::endl;
                throw std::runtime_error("pd_test_groupby_basic failed: ngroups should be 2");
            }

            std::cout << " -> tests passed" << std::endl;
        }

        void pd_test_groupby_multiple_columns() {
            std::cout << "========= GroupBy multiple columns ==============";

nth (pd_test_3_all.cpp:27491)

    check(result_cumsum["B"].get_value_double(1) == 2.0, "row 1 (bar) cumsum B = 2");
    check(result_cumsum["B"].get_value_double(3) == 6.0, "row 3 (bar) cumsum B = 6");
}

void pd_test_gb_nth_basic() {
    std::cout << "  -- pd_test_gb_nth_basic --" << std::endl;

    auto df = make_test_df();
    auto gb = df.groupby("A");

    auto result = gb.nth(0);
    check(result.nrows() == 2, "nth(0) returns 2 rows (one per group)");

    auto result_last = gb.nth(-1);
    check(result_last.nrows() == 2, "nth(-1) returns 2 rows");

    auto result_multi = gb.nth(std::vector<int>{0, -1});
    check(result_multi.nrows() == 4, "nth([0,-1]) returns 4 rows");
}

void pd_test_gb_nth_slice() {

series (pd_test_2_all.cpp:2307)

            std::vector<std::string> index = {"a", "b", "c", "d", "e"};

            std::map<std::string, std::vector<numpy::float64>> data1;
            data1["col1"] = {1.0, 2.0, 3.0, 4.0, 5.0};
            data1["col2"] = {2.0, 4.0, 6.0, 8.0, 10.0};  // Perfectly correlated with col1

            pandas::DataFrame df1(data1, std::make_unique<pandas::Index<std::string>>(index));

            // Series with same index and values that correlate with df columns
            pandas::Series<numpy::float64> series({1.0, 2.0, 3.0, 4.0, 5.0});
            series.set_index(pandas::Index<std::string>(index));

            pandas::Series<numpy::float64> result = df1.corrwith(series);

            bool passed = true;
            // col1 should have correlation 1.0 with series
            if (!approx_equal(result[0], 1.0)) {
                std::cout << "\n  [FAIL] : Expected correlation 1.0 for col1, got " << result[0] << std::endl;
                passed = false;
            }

set_categorical_categories (pd_test_2_all.cpp:20841)

}

void test_sgb_apply_result_index_categorical() {
    std::cout << "  -- test_sgb_apply_result_index_categorical --" << std::endl;

    std::vector<numpy::float64> values = {5.0, 10.0};
    pandas::Series<std::string> by({"A", "B"});
    pandas::Series<numpy::float64> data(values);

    auto sgb = data.groupby(by);
    sgb.set_categorical_categories({"A", "B", "C"});
    sgb.set_index_name("cat_key");

    pandas::Series<numpy::float64> result(values);
    std::vector<std::string> idx_labels = {"A", "B"};
    result.set_index(std::make_unique<pandas::Index<std::string>>(idx_labels));

    sgb.apply_result_index(result);

    // Should have CategoricalIndex (dtype_name() returns "category")
    check(result.index().dtype_name() == "category", "is_categorical_index");

set_multiindex_names (pd_test_2_all.cpp:20775)

    // Simulate a 2-level groupby result with composite \x1f keys
    using std::string;
    string sep(1, '\x1f');
    std::vector<string> keys = {"A" + sep + "X", "A" + sep + "Y", "B" + sep + "X", "B" + sep + "Y"};

    std::vector<numpy::float64> values = {1.0, 2.0, 3.0, 4.0};
    pandas::Series<std::string> by(keys);
    pandas::Series<numpy::float64> data(values);

    auto sgb = data.groupby(by);
    sgb.set_multiindex_names({"level0", "level1"});

    // Create a "result" series with composite index
    pandas::Series<numpy::float64> result(values);
    result.set_index(std::make_unique<pandas::Index<std::string>>(keys));

    sgb.apply_result_index(result);

    // Should now have a MultiIndex
    check(result.has_multiindex(), "has_multiindex");
    check(result.multiindex().nlevels() == 2, "nlevels_2");

size (pd_test_1_all.cpp:22)

#include "../pandas/pd_boolean_array.h"

namespace dataframe_tests {

namespace dataframe_tests_boolean_array {
    void pd_test_boolean_array_constructors() {
        std::cout << "========= BooleanArray: constructors ======================= ";

        // Default constructor
        pandas::BooleanArray arr1;
        if (arr1.size() != 0) {
            std::cout << "  [FAIL] : in pd_test_boolean_array_constructors() : default constructor size != 0" << std::endl;
            throw std::runtime_error("pd_test_boolean_array_constructors failed: default constructor size != 0");
        }

        // Initializer list constructor
        pandas::BooleanArray arr2({
            std::optional<bool>(true),
            std::optional<bool>(false),
            std::nullopt,
            std::optional<bool>(true)