Refactoring, test updates

Author: dyashuni

.github/workflows/build.yml

@@ -19,10 +19,13 @@ jobs:
         run: python -m pip install .
 
       - name: Test
-        run: python -m unittest discover --start-directory python_bindings/tests --pattern "*_test*.py"
+        run: python -m unittest discover -v --start-directory python_bindings/tests --pattern "*_test*.py"
 
   test_cpp:
-    runs-on: ubuntu-latest
+    runs-on: ${{matrix.os}}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest]
     steps:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4
@@ -34,17 +37,27 @@ jobs:
           mkdir build
           cd build
           cmake ..
-          make
+          if [ "$RUNNER_OS" == "Linux" ]; then
+            make
+          elif [ "$RUNNER_OS" == "Windows" ]; then
+            cmake --build ./ --config Release
+          fi
+        shell: bash
 
       - name: Prepare test data
         run: |
           pip install numpy
           cd examples
           python update_gen_data.py
+        shell: bash
 
       - name: Test
         run: |
           cd build
+          if [ "$RUNNER_OS" == "Windows" ]; then
+            cp ./Release/* ./
+          fi
           ./searchKnnCloserFirst_test
           ./test_updates
           ./test_updates update
+        shell: bash

.gitignore

@@ -8,4 +8,4 @@ hnswlib.cpython*.so
 var/
 .idea/
 .vscode/
-
+.vs/

examples/git_tester.py

@@ -1,34 +1,40 @@
+import os
+import shutil
+
+from sys import platform
 from pydriller import Repository
-import os 
-import datetime
-os.system("cp examples/speedtest.py examples/speedtest2.py") # the file has to be outside of git
-for idx, commit in enumerate(Repository('.', from_tag="v0.6.0").traverse_commits()):    
-    name=commit.msg.replace('\n', ' ').replace('\r', ' ')
-    print(idx, commit.hash, name)
 
 
+speedtest_src_path = os.path.join("examples", "speedtest.py")
+speedtest_copy_path = os.path.join("examples", "speedtest2.py")
+shutil.copyfile(speedtest_src_path, speedtest_copy_path) # the file has to be outside of git
+
+commits = list(Repository('.', from_tag="v0.6.0").traverse_commits())
+print("Found commits:")
+for idx, commit in enumerate(commits):
+    name = commit.msg.replace('\n', ' ').replace('\r', ' ')
+    print(idx, commit.hash, name)
 
-for commit in Repository('.', from_tag="v0.6.0").traverse_commits():
-    
-    name=commit.msg.replace('\n', ' ').replace('\r', ' ')
-    print(commit.hash, name)
-    
-    os.system(f"git checkout {commit.hash}; rm -rf build; ")
+for commit in commits:
+    name = commit.msg.replace('\n', ' ').replace('\r', ' ')
+    print("\nProcessing", commit.hash, name)
+
+    if os.path.exists("build"):
+        shutil.rmtree("build")
+    os.system(f"git checkout {commit.hash}")
     print("\n\n--------------------\n\n")
-    ret=os.system("python -m pip install .")
-    print(ret)
-    
-    if ret != 0:
-        print ("build failed!!!!")
-        print ("build failed!!!!")
-        print ("build failed!!!!")
-        print ("build failed!!!!")
-        continue    
-    
-    os.system(f'python examples/speedtest2.py -n "{name}" -d 4 -t 1')
-    os.system(f'python examples/speedtest2.py -n "{name}" -d 64 -t 1')
-    os.system(f'python examples/speedtest2.py -n "{name}" -d 128 -t 1')
-    os.system(f'python examples/speedtest2.py -n "{name}" -d 4 -t 24')
-    os.system(f'python examples/speedtest2.py -n "{name}" -d 128 -t 24')
+    ret = os.system("python -m pip install .")
+    print("Install result:", ret)
 
+    if ret != 0:
+        print("build failed!!!!")
+        print("build failed!!!!")
+        print("build failed!!!!")
+        print("build failed!!!!")
+        continue
 
+    os.system(f'python {speedtest_copy_path} -n "{name}" -d 4 -t 1')
+    os.system(f'python {speedtest_copy_path} -n "{name}" -d 64 -t 1')
+    os.system(f'python {speedtest_copy_path} -n "{name}" -d 128 -t 1')
+    os.system(f'python {speedtest_copy_path} -n "{name}" -d 4 -t 24')
+    os.system(f'python {speedtest_copy_path} -n "{name}" -d 128 -t 24')

examples/updates_test.cpp

@@ -1,5 +1,7 @@
 #include "../hnswlib/hnswlib.h"
 #include <thread>
+
+
 class StopW
 {
     std::chrono::steady_clock::time_point time_begin;
@@ -22,6 +24,7 @@ class StopW
     }
 };
 
+
 /*
  * replacement for the openmp '#pragma omp parallel for' directive
  * only handles a subset of functionality (no reductions etc)
@@ -81,8 +84,6 @@ inline void ParallelFor(size_t start, size_t end, size_t numThreads, Function fn
             std::rethrow_exception(lastException);
         }
     }
-
-
 }
 
 
@@ -94,7 +95,7 @@ std::vector<datatype> load_batch(std::string path, int size)
     assert(sizeof(datatype) == 4);
 
     std::ifstream file;
-    file.open(path);
+    file.open(path, std::ios::binary);
     if (!file.is_open())
     {
         std::cout << "Cannot open " << path << "\n";
@@ -107,15 +108,14 @@ std::vector<datatype> load_batch(std::string path, int size)
     return batch;
 }
 
+
 template <typename d_type>
 static float
 test_approx(std::vector<float> &queries, size_t qsize, hnswlib::HierarchicalNSW<d_type> &appr_alg, size_t vecdim,
             std::vector<std::unordered_set<hnswlib::labeltype>> &answers, size_t K)
 {
     size_t correct = 0;
     size_t total = 0;
-    //uncomment to test in parallel mode:
-    
 
     for (int i = 0; i < qsize; i++)
     {
@@ -137,10 +137,16 @@ test_approx(std::vector<float> &queries, size_t qsize, hnswlib::HierarchicalNSW<
     return 1.0f * correct / total;
 }
 
+
 static void
-test_vs_recall(std::vector<float> &queries, size_t qsize, hnswlib::HierarchicalNSW<float> &appr_alg, size_t vecdim,
-               std::vector<std::unordered_set<hnswlib::labeltype>> &answers, size_t k)
-{
+test_vs_recall(
+    std::vector<float> &queries,
+    size_t qsize,
+    hnswlib::HierarchicalNSW<float> &appr_alg,
+    size_t vecdim,
+    std::vector<std::unordered_set<hnswlib::labeltype>> &answers,
+    size_t k) {
+
     std::vector<size_t> efs = {1};
     for (int i = k; i < 30; i++)
     {
@@ -155,6 +161,8 @@ test_vs_recall(std::vector<float> &queries, size_t qsize, hnswlib::HierarchicalN
         efs.push_back(i);
     }
     std::cout << "ef\trecall\ttime\thops\tdistcomp\n";
+
+    bool test_passed = false;
     for (size_t ef : efs)
     {
         appr_alg.setEf(ef);
@@ -171,20 +179,24 @@ test_vs_recall(std::vector<float> &queries, size_t qsize, hnswlib::HierarchicalN
         std::cout << ef << "\t" << recall << "\t" << time_us_per_query << "us \t"<<hops_per_query<<"\t"<<distance_comp_per_query << "\n";
         if (recall > 0.99)
         {
+            test_passed = true;
             std::cout << "Recall is over 0.99! "<<recall << "\t" << time_us_per_query << "us \t"<<hops_per_query<<"\t"<<distance_comp_per_query << "\n";
             break;
         }
     }
+    if (!test_passed)
+    {
+        std::cerr << "Test failed\n";
+        exit(1);
+    }
 }
 
+
 int main(int argc, char **argv)
 {
-
     int M = 16;
     int efConstruction = 200;
     int num_threads = std::thread::hardware_concurrency();
-  
-
 
     bool update = false;
 
@@ -207,7 +219,6 @@ int main(int argc, char **argv)
 
     std::string path = "../examples/data/";
 
-
     int N;
     int dummy_data_multiplier;
     int N_queries;
@@ -216,8 +227,7 @@ int main(int argc, char **argv)
     {
         std::ifstream configfile;
         configfile.open(path + "/config.txt");
-        if (!configfile.is_open())
-        {
+        if (!configfile.is_open()) {
             std::cout << "Cannot open config.txt\n";
             return 1;
         }
@@ -237,11 +247,9 @@ int main(int argc, char **argv)
 
     StopW stopw = StopW();
 
-    if (update)
-    {
+    if (update) {
         std::cout << "Update iteration 0\n";
 
-        
         ParallelFor(1, N, num_threads, [&](size_t i, size_t threadId) {
             appr_alg.addPoint((void *)(dummy_batch.data() + i * d), i);
         });
@@ -252,13 +260,12 @@ int main(int argc, char **argv)
         });
         appr_alg.checkIntegrity();
 
-        for (int b = 1; b < dummy_data_multiplier; b++)
-        {
+        for (int b = 1; b < dummy_data_multiplier; b++) {
             std::cout << "Update iteration " << b << "\n";
             char cpath[1024];
             sprintf(cpath, "batch_dummy_%02d.bin", b);
             std::vector<float> dummy_batchb = load_batch<float>(path + cpath, N * d);
-            
+
             ParallelFor(0, N, num_threads, [&](size_t i, size_t threadId) {            
                 appr_alg.addPoint((void *)(dummy_batch.data() + i * d), i);
             });
@@ -268,7 +275,7 @@ int main(int argc, char **argv)
 
     std::cout << "Inserting final elements\n";
     std::vector<float> final_batch = load_batch<float>(path + "batch_final.bin", N * d);
-    
+
     stopw.reset();
     ParallelFor(0, N, num_threads, [&](size_t i, size_t threadId) {
                     appr_alg.addPoint((void *)(final_batch.data() + i * d), i);
@@ -280,19 +287,16 @@ int main(int argc, char **argv)
     std::vector<int> gt = load_batch<int>(path + "gt.bin", N_queries * K);
 
     std::vector<std::unordered_set<hnswlib::labeltype>> answers(N_queries);
-    for (int i = 0; i < N_queries; i++)
-    {
-        for (int j = 0; j < K; j++)
-        {
+    for (int i = 0; i < N_queries; i++) {
+        for (int j = 0; j < K; j++) {
             answers[i].insert(gt[i * K + j]);
         }
     }
 
-    for (int i = 0; i < 3; i++)
-    {
+    for (int i = 0; i < 3; i++) {
         std::cout << "Test iteration " << i << "\n";
         test_vs_recall(queries_batch, N_queries, appr_alg, d, answers, K);
     }
 
     return 0;
-};
+};