effspm 0.1.5__cp310-cp310-win_amd64.whl → 0.3.0__cp310-cp310-win_amd64.whl

effspm/_effspm.cpp

@@ -0,0 +1,679 @@
+// _effspm.cpp
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+namespace py = pybind11;
+#include <iostream>
+
+
+// PrefixProjection headers
+#include "freq_miner.hpp"
+#include "load_inst.hpp"
+#include "utility.hpp"
+
+// BTMiner (wrapped in its own namespace in source files)
+#include "btminer/src/freq_miner.hpp"
+#include "btminer/src/load_inst.hpp"
+#include "btminer/src/utility.hpp"
+#include "btminer/src/build_mdd.hpp"
+
+// HTMiner (wrapped in its own namespace in source files)
+#include "htminer/src/build_mdd.hpp"    // ← ensure HTMiner MDD builder is available
+#include "htminer/src/freq_miner.hpp"
+#include "htminer/src/load_inst.hpp"
+#include "htminer/src/utility.hpp"
+
+
+#include "largepp/src/freq_miner.hpp"
+#include "largepp/src/load_inst.hpp"
+#include "largepp/src/utility.hpp"
+
+#include "largebm/src/freq_miner.hpp"
+#include "largebm/src/load_inst.hpp"
+#include "largebm/src/utility.hpp"
+#include "largebm/src/build_mdd.hpp"
+
+#include "largehm/src/freq_miner.hpp"
+#include "largehm/src/load_inst.hpp"
+#include "largehm/src/utility.hpp"
+#include "largehm/src/build_mdd.hpp"
+
+
+
+PYBIND11_MODULE(_effspm, m) {
+    m.doc() = "Unified SPM library: PrefixProjection, BTMiner, HTMiner";
+
+    // ─────────────────────────────────────────────────────────────
+    // PrefixProjection
+    // ─────────────────────────────────────────────────────────────
+    m.def("PrefixProjection",
+        [](py::object data,
+           double minsup,
+           unsigned int time_limit,
+           bool preproc,
+           bool use_dic,
+           bool verbose,
+           const std::string &out_file)
+        {
+            ::time_limit = time_limit;
+            ::pre_pro    = preproc;
+            ::use_dic    = use_dic;
+            ::use_list   = false;
+            ::b_disp     = verbose;
+            ::b_write    = !out_file.empty();
+            ::out_file   = out_file;
+
+            ClearCollected();
+            start_time = std::clock();
+
+            if (py::isinstance<py::str>(data)) {
+                std::string path = data.cast<std::string>();
+                if (!Load_instance(path, minsup))
+                    throw std::runtime_error("Failed to load file: " + path);
+            } else {
+                auto seqs = data.cast<std::vector<std::vector<int>>>();
+                items = std::move(seqs);
+                N = items.size();
+
+                int max_id = 0;
+                for (auto &seq : items)
+                    for (int x : seq)
+                        max_id = std::max(max_id, std::abs(x));
+                L = max_id;
+
+                theta = (minsup < 1.0) ? std::ceil(minsup * N) : minsup;
+
+                DFS.clear();
+                DFS.reserve(L);
+                for (unsigned int i = 0; i < L; ++i)
+                    DFS.emplace_back(-static_cast<int>(i) - 1);
+
+                M = 0;
+                E = 0;
+                for (auto &seq : items) {
+                    M = std::max<unsigned int>(M, seq.size());
+                    E += seq.size();
+                }
+            }
+
+            Freq_miner();
+
+            py::dict out;
+            out["patterns"] = GetCollected();
+            out["time"]     = give_time(std::clock() - start_time);
+            return out;
+        },
+        py::arg("data"),
+        py::arg("minsup")     = 0.01,
+        py::arg("time_limit") = 36000,
+        py::arg("preproc")    = false,
+        py::arg("use_dic")    = false,
+        py::arg("verbose")    = false,
+        py::arg("out_file")   = ""
+    );
+m.def("BTMiner",
+    [](py::object data,
+       double minsup,
+       unsigned int time_limit,
+       bool preproc,
+       bool use_dic,
+       bool verbose,
+       const std::string &out_file)
+    {
+        // We are calling the *professor* BTMiner, now namespaced as btminer::.
+        // So we only set the globals the professor code actually has.
+
+        // 1) configure professor globals
+        btminer::time_limit = static_cast<int>(time_limit);
+        btminer::pre_pro    = preproc;
+        btminer::use_dic    = use_dic;
+        btminer::b_disp     = verbose;
+        btminer::b_write    = !out_file.empty();
+        btminer::out_file   = out_file;
+        btminer::N_mult     = 1;         // professor uses these too
+        btminer::M_mult     = 1;
+        btminer::just_build = false;     // we want full mining
+
+        btminer::start_time = std::clock();
+
+        // 2) load data
+        //
+        // Professor’s code is primarily file-based (Load_instance(const string&, double)).
+        // So: if user passes a file path → use the professor loader directly.
+        // If user passes a Python list-of-lists → we will build the MDD the same
+        // way professor’s loader does, but without changing his logic.
+        if (py::isinstance<py::str>(data)) {
+            // ----- FILE MODE -----
+            std::string path = data.cast<std::string>();
+
+            if (verbose) {
+                std::cerr << "[BT][binding] file=" << path
+                          << " minsup=" << minsup
+                          << " preproc=" << preproc << std::endl;
+            }
+
+            if (!btminer::Load_instance(path, minsup)) {
+                throw std::runtime_error("BTMiner: failed to load file: " + path);
+            }
+        } else {
+            // ----- PYTHON LIST MODE -----
+            //
+            // We mimic professor’s loader:
+            //   - create root in Tree
+            //   - compute N, M, L
+            //   - compute theta from minsup
+            //   - seed DFS (one Pattern per item, as in Preprocess branch)
+            //   - call Build_MDD(...) for each sequence
+            //
+            // This DOES NOT change his mining logic; it just drives it from memory.
+
+            auto seqs = data.cast<std::vector<std::vector<int>>>();
+
+            // clear MDD and globals to a known state
+            btminer::Tree.clear();
+            btminer::Tree.emplace_back(0, 0, 0);      // root (exactly like professor)
+
+            // compute basic stats
+            int max_id = 0;
+            int max_len = 0;
+            int seq_count = 0;
+            long long entries = 0;
+
+            for (const auto &s : seqs) {
+                if (s.empty()) continue;
+                ++seq_count;
+                max_len = std::max<int>(max_len, static_cast<int>(s.size()));
+                for (int x : s) {
+                    max_id = std::max(max_id, std::abs(x));
+                    ++entries;
+                }
+            }
+
+            btminer::N = seq_count;
+            btminer::M = max_len;
+            btminer::L = max_id;
+            btminer::E = static_cast<int>(entries);
+
+            // theta = abs support
+            if (minsup < 1.0)
+                btminer::theta = static_cast<int>(std::ceil(minsup * btminer::N * btminer::N_mult));
+            else
+                btminer::theta = static_cast<int>(minsup);
+
+            // seed DFS exactly like professor does in the preprocessed branch:
+            btminer::DFS.clear();
+            btminer::DFS.reserve(btminer::L);
+            for (int i = 0; i < btminer::L; ++i)
+                btminer::DFS.emplace_back(-i - 1);
+
+            // now build the MDD, sequence by sequence
+            for (const auto &s : seqs) {
+                if (s.empty()) continue;
+                // professor’s Build_MDD takes a vector<int> by non-const ref
+                std::vector<int> tmp = s;
+                btminer::Build_MDD(tmp);
+            }
+
+            if (verbose) {
+                std::cerr << "[BT][binding] PY mode: N=" << btminer::N
+                          << " L=" << btminer::L
+                          << " M=" << btminer::M
+                          << " E=" << btminer::E
+                          << " theta=" << btminer::theta
+                          << " Tree.size()=" << btminer::Tree.size()
+                          << std::endl;
+            }
+        }
+
+        // 3) run professor’s miner
+        btminer::Freq_miner();
+
+        // 4) build python result
+        // 4) build python result
+            py::dict out;
+            out["patterns"]     = btminer::GetCollected();    // ← NEW
+            out["num_patterns"] = btminer::num_patt;
+            out["time"]         = btminer::give_time(std::clock() - btminer::start_time);
+            out["N"]            = btminer::N;
+            out["L"]            = btminer::L;
+            out["theta"]        = btminer::theta;
+            return out;
+
+    },
+    py::arg("data"),
+    py::arg("minsup")     = 0.01,
+    py::arg("time_limit") = 36000,
+    py::arg("preproc")    = false,
+    py::arg("use_dic")    = false,
+    py::arg("verbose")    = false,
+    py::arg("out_file")   = ""
+);
+
+
+
+       
+// HTMiner
+// ─────────────────────────────────────────────────────────────
+// HTMiner
+m.def("HTMiner",
+    [](py::object data,
+       double minsup, unsigned int time_limit,
+       bool preproc, bool use_dic,
+       bool verbose, const std::string &out_file)
+    {
+        htminer::time_limit = time_limit;
+        htminer::pre_pro    = preproc;
+        htminer::use_dic    = use_dic;
+        htminer::just_build = false;
+        htminer::use_list   = false;
+        htminer::b_disp     = verbose;
+        htminer::b_write    = !out_file.empty();
+        htminer::out_file   = out_file;
+        htminer::ClearCollected();
+        htminer::start_time = std::clock();
+
+        if (py::isinstance<py::str>(data)) {
+            std::string path = data.cast<std::string>();
+            if (!htminer::Load_instance(path, minsup))
+                throw std::runtime_error("Failed to load file: " + path);
+        } else {
+            auto seqs = data.cast<std::vector<std::vector<int>>>();
+            htminer::items = std::move(seqs);
+            htminer::N = htminer::items.size();
+
+            int max_id = 0;
+            htminer::M = 0;
+            htminer::E = 0;
+            for (auto &seq : htminer::items) {
+                htminer::M = std::max<unsigned int>(htminer::M, seq.size());
+                for (int x : seq)
+                    max_id = std::max(max_id, std::abs(x));
+                htminer::E += seq.size();
+            }
+            htminer::L = max_id;
+            htminer::theta = (minsup < 1.0)
+                              ? static_cast<unsigned long long>(std::ceil(minsup * htminer::N))
+                              : static_cast<unsigned long long>(minsup);
+
+            htminer::DFS.clear();
+            htminer::DFS.reserve(htminer::L);
+            for (unsigned int i = 0; i < static_cast<unsigned int>(htminer::L); ++i)
+                htminer::DFS.emplace_back(-static_cast<int>(i) - 1);
+
+            htminer::VDFS.clear();
+            htminer::VDFS.resize(htminer::L);
+        }
+
+        htminer::Freq_miner();
+
+        // 👇 now really respects verbose
+        if (verbose) {
+            std::cout << " total patterns = "
+                      << htminer::collectedPatterns.size() << "\n";
+        }
+
+        py::dict out;
+        out["patterns"] = htminer::GetCollected();
+        out["time"]     = htminer::give_time(std::clock() - htminer::start_time);
+        return out;
+    },
+    py::arg("data"),
+    py::arg("minsup")     = 0.01,
+    py::arg("time_limit") = 36000,
+    py::arg("preproc")    = false,
+    py::arg("use_dic")    = false,
+    py::arg("verbose")    = false,
+    py::arg("out_file")   = ""
+);
+
+
+ m.def("LargePrefixProjection",
+    [](py::object data,
+       double minsup,
+       unsigned int time_limit,
+       bool preproc,
+       bool use_dic,
+       bool verbose,
+       const std::string &out_file)
+    {
+        largepp::time_limit = time_limit;
+        largepp::pre_pro    = preproc;
+        largepp::use_dic    = use_dic;
+        largepp::use_list   = true;
+        largepp::b_disp     = verbose;
+        largepp::b_write    = !out_file.empty();
+        largepp::out_file   = out_file;
+        largepp::just_build = false;
+
+        largepp::ClearCollected();
+        largepp::start_time = std::clock();
+
+        // 👇 this was the last noisy one
+        if (verbose) {
+            std::cerr << " minsup=" << minsup
+                      << " preproc=" << preproc
+                      << " verbose=" << verbose
+                      << " out_file=" << (out_file.empty() ? "(none)" : out_file)
+                      << " use_dic=" << use_dic << "\n";
+        }
+
+        if (py::isinstance<py::str>(data)) {
+            std::string fname = data.cast<std::string>();
+            largepp::Load_instance(fname, minsup);
+        } else {
+            largepp::Load_py(data, minsup);
+        }
+
+        largepp::Freq_miner();
+
+        py::dict out;
+        out["patterns"] = largepp::GetCollected();
+        out["time"]     = largepp::give_time(std::clock() - largepp::start_time);
+        return out;
+    },
+    py::arg("data"),
+    py::arg("minsup")     = 0.01,
+    py::arg("time_limit") = 36000,
+    py::arg("preproc")    = false,
+    py::arg("use_dic")    = false,
+    py::arg("verbose")    = false,
+    py::arg("out_file")   = ""
+);
+
+
+
+// ─────────────────────────────────────────────────────────────
+// LargeBTMiner  -- Python wrapper for the largebm implementation
+// ─────────────────────────────────────────────────────────────
+// m.def(
+//     "LargeBTMiner",
+//     [](py::object        data,
+//        double            minsup ,
+//        unsigned int      time_limit,
+//        bool              preproc ,
+//        bool              use_dic,
+//        bool              verbose,
+//        const std::string &out_file ) 
+//     {
+//         /* 1) Global flags */
+//         largebm::time_limit = time_limit;
+//         largebm::pre_pro    = preproc;
+//         largebm::use_dic    = use_dic;
+//         largebm::use_list   = false;          // large-mode → always MDD
+//         largebm::just_build = false;
+//         largebm::b_disp     = verbose;
+//         largebm::b_write    = !out_file.empty();
+//         largebm::out_file   = out_file;
+
+//         /* 2) Reset per-run state */
+//         largebm::ClearCollected();
+//         largebm::start_time = std::clock();
+
+//         /* 3) Load the DB (file path or in-memory list<list<int>>) */
+//         if (py::isinstance<py::str>(data)) {
+//             std::string path = data.cast<std::string>();
+//             if (!largebm::Load_instance(path, minsup))
+//                 throw std::runtime_error("Failed to load file: " + path);
+//         } else {
+//             // In-memory sequences
+//             largebm::items = std::move(data.cast<std::vector<std::vector<int>>>());
+//             largebm::N     = static_cast<unsigned int>(largebm::items.size());
+
+//             /* -- basic stats -- */
+//             int max_id = 0;
+//             largebm::M = 0;
+//             largebm::E = 0;
+//             for ( auto &seq : largebm::items) {
+//                 largebm::M = std::max<unsigned int>(largebm::M,
+//                                                     static_cast<unsigned int>(seq.size()));
+//                 largebm::E += static_cast<unsigned long long>(seq.size());
+//                 for (int x : seq) max_id = std::max(max_id, std::abs(x));
+//             }
+//             largebm::L     = static_cast<unsigned int>(max_id);
+//             largebm::theta = (minsup < 1.0)
+//                                ? static_cast<unsigned long long>(std::ceil(minsup * largebm::N))
+//                                : static_cast<unsigned long long>(minsup);
+
+//             /* -- DFS buffer (size = L) -- */
+//             largebm::DFS.clear();
+//             largebm::DFS.reserve(largebm::L);
+//             for (unsigned int i = 0; i < largebm::L; ++i)
+//                 largebm::DFS.emplace_back(-static_cast<int>(i) - 1);
+
+//             /* -- Build the MDD -- */
+//             largebm::Tree.clear();
+//             largebm::Tree.emplace_back(0, 0, 0);              // dummy root
+//             for ( auto &seq : largebm::items)
+//                 largebm::Build_MDD(seq);
+//         }
+
+//         /* 4) Mine and return results */
+//         largebm::Freq_miner();
+
+//         py::dict out;
+//         out["patterns"] = largebm::GetCollected();
+//         out["time"]     = largebm::give_time(std::clock() - largebm::start_time);
+//         return out;
+//     },
+//     py::arg("data"),
+//     py::arg("minsup")     = 0.01,
+//     py::arg("time_limit") = 36000,
+//     py::arg("preproc")    = false,
+//     py::arg("use_dic")    = false,
+//     py::arg("verbose")    = false,
+//     py::arg("out_file")   = ""
+// );
+
+
+
+    // ─────────────────────────────────────────────────────────────────────────
+    // LargeBTMiner (MDD-based)
+    // ─────────────────────────────────────────────────────────────────────────
+ m.def("LargeBTMiner",
+    [](py::object data,
+       double minsup,
+       unsigned int time_limit,
+       bool preproc,
+       bool use_dic,
+       bool verbose,
+       const std::string &out_file)
+    {
+        using namespace largebm;
+
+        // 0) Set global flags and timers
+        largebm::time_limit = time_limit;
+        largebm::pre_pro    = preproc;
+        largebm::use_dic    = use_dic;
+        largebm::use_list   = false;         // large-mode → always MDD
+        largebm::b_disp     = verbose;
+        largebm::b_write    = !out_file.empty();
+        largebm::out_file   = out_file;
+        largebm::just_build = false;
+
+        // 0.1) Clear any leftover data/state from previous runs
+        largebm::items.clear();
+        largebm::item_dic.clear();
+        largebm::inv_item_dic.clear();
+        largebm::Tree.clear();
+        largebm::DFS.clear();
+        largebm::ClearCollected();
+
+        // 1) Load sequences (either from filename or from Python list)
+        if (py::isinstance<py::str>(data)) {
+            // ─────────── FILE-BASED MODE ───────────
+            std::string path = data.cast<std::string>();
+            if (!largebm::Load_instance(path, minsup))
+                throw std::runtime_error("Failed to load file: " + path);
+
+        } else {
+            // ────────── IN-MEMORY MODE ──────────
+            auto seqs = data.cast<std::vector<std::vector<int>>>();
+            largebm::items = std::move(seqs);
+            largebm::N     = largebm::items.size();
+
+            // 1.1) Compute basic DB statistics (M, E, L) and absolute support θ
+            int max_id = 0;
+            largebm::M = 0;
+            largebm::E = 0;
+            for (auto &seq : largebm::items) {
+                largebm::M = std::max<unsigned int>(largebm::M, static_cast<unsigned int>(seq.size()));
+                largebm::E += static_cast<unsigned long long>(seq.size());
+                for (int x : seq) max_id = std::max(max_id, std::abs(x));
+            }
+            largebm::L = static_cast<unsigned int>(max_id);
+            largebm::theta = (minsup < 1.0)
+                               ? static_cast<unsigned long long>(std::ceil(minsup * largebm::N))
+                               : static_cast<unsigned long long>(minsup);
+
+            // 1.2) Initialize DFS buffer (size = L)
+            largebm::DFS.reserve(largebm::L);
+            for (unsigned int i = 0; i < largebm::L; ++i)
+                largebm::DFS.emplace_back(-static_cast<int>(i) - 1);
+
+            // 1.3) Build the MDD “Tree”
+            // Insert one dummy root node (item=0, freq=0, anct=0)
+            largebm::Tree.emplace_back(0, 0, 0);
+            for (auto &seq : largebm::items)
+                largebm::Build_MDD(const_cast<std::vector<int>&>(seq));
+        }
+
+        // 2) Rebuild inverse-dictionary from fresh item_dic
+        {
+            std::vector<int> inv(largebm::item_dic.size() + 1);
+            for (int old = 1; old <= static_cast<int>(largebm::item_dic.size()); ++old) {
+                int cid = largebm::item_dic[old - 1];
+                if (cid > 0) inv[cid] = old;
+            }
+            largebm::inv_item_dic = std::move(inv);
+        }
+
+        // 3) Start timing and run the miner
+        largebm::start_time = std::clock();
+        largebm::Freq_miner();
+
+        // 4) Collect results and elapsed time
+        const auto& pats = largebm::GetCollected();
+
+        py::dict out;
+        out["patterns"] = pats;
+        out["time"]     = largebm::give_time(std::clock() - largebm::start_time);
+        return out;
+    },
+    py::arg("data"),
+    py::arg("minsup")     = 0.01,
+    py::arg("time_limit") = 36000,
+    py::arg("preproc")    = false,
+    py::arg("use_dic")    = false,
+    py::arg("verbose")    = false,
+    py::arg("out_file")   = ""
+);
+
+
+   
+m.def("LargeHTMiner",
+    [](py::object data,
+       double minsup,
+       unsigned int time_limit,
+       bool preproc,
+       bool use_dic,
+       bool verbose,
+       const std::string &out_file)
+    {
+        // 0) Set global flags and timers:
+        largehm::time_limit = time_limit;
+        largehm::pre_pro    = preproc;
+        largehm::use_dic    = use_dic;
+        largehm::use_list   = true;     // force in‐memory mode
+        largehm::b_disp     = verbose;
+        largehm::b_write    = !out_file.empty();
+        largehm::out_file   = out_file;
+        largehm::just_build = false;
+
+        largehm::ClearCollected();
+        largehm::start_time = std::clock();
+
+        if (py::isinstance<py::str>(data)) {
+            // ───────────── FILE‐BASED MODE ─────────────
+            // Force mlim so that every item lands in temp_vec (never temp_lim):
+            largehm::mlim = UINT_MAX;
+
+            std::string path = data.cast<std::string>();
+            if (! largehm::Load_instance(path, minsup))
+                throw std::runtime_error("Failed to load file: " + path);
+        }
+        else {
+            // ───────────── IN‐MEMORY MODE ─────────────
+            auto seqs = data.cast<std::vector<std::vector<int>>>();
+            largehm::items = std::move(seqs);
+            largehm::N     = largehm::items.size();
+
+            // 1) Compute L = maximum absolute item ID
+            int max_id = 0;
+            for (auto &seq : largehm::items)
+                for (int x : seq)
+                    max_id = std::max(max_id, std::abs(x));
+            largehm::L = static_cast<unsigned int>(max_id);
+
+            // 2) Compute theta as absolute support threshold
+            largehm::theta = (minsup < 1.0)
+                               ? static_cast<unsigned long long>(std::ceil(minsup * largehm::N))
+                               : static_cast<unsigned long long>(minsup);
+
+            // 3) Initialize DFS (size = L)
+            largehm::DFS.clear();
+            largehm::DFS.reserve(largehm::L);
+            for (unsigned int i = 0; i < largehm::L; ++i)
+                largehm::DFS.emplace_back(-static_cast<int>(i) - 1);
+
+            // 4) Compute M (max sequence length) and E (total entries)
+            largehm::M = 0;
+            largehm::E = 0;
+            for (auto &seq : largehm::items) {
+                largehm::M = std::max<unsigned int>(
+                    largehm::M, static_cast<unsigned int>(seq.size()));
+                largehm::E += seq.size();
+            }
+
+            // 5) ─── Build the MDD “manually” ───
+            largehm::Tree.clear();
+            largehm::VTree.clear();
+            largehm::CTree.clear();
+
+            // Insert exactly one dummy root node (chld=0, sibl=0, freq=0):
+            largehm::Tree.emplace_back(0,0,0);
+
+            // For each sequence “seq”, insert into MDD by placing a single −1 sentinel:
+            for (auto &seq : largehm::items) {
+                // Copy the item IDs:
+                std::vector<int> temp_vec = seq;
+                // Only a single “−1” is needed to force the suffix insertion:
+                std::vector<int> temp_lim(1, -1);
+
+                largehm::Build_MDD(temp_vec, temp_lim);
+            }
+
+            
+        }
+
+        // 6) Run the frequency miner (Tree is now properly built):
+        largehm::Freq_miner();
+
+        // 7) Return results to Python:
+        py::dict out;
+        out["patterns"] = largehm::GetCollected();
+        out["time"]     = largehm::give_time(std::clock() - largehm::start_time);
+        return out;
+    },
+    py::arg("data"),
+    py::arg("minsup")     = 0.01,
+    py::arg("time_limit") = 36000,
+    py::arg("preproc")    = false,
+    py::arg("use_dic")    = false,
+    py::arg("verbose")    = false,
+    py::arg("out_file")   = ""
+);
+
+
+
+
+}