branch completed

.gitignore

@@ -7,6 +7,7 @@
 
 # Distribution / packaging
 build/
+dist/
 *.egg-info/
 
 # ctags

CMakeLists.txt

@@ -1,19 +1,31 @@
 cmake_minimum_required(VERSION 3.10)
 
 project(nndescent
-    VERSION 0.0.0
+    VERSION 1.0.3
     LANGUAGES CXX
     DESCRIPTION "A C++ implementation of the nearest neighbour descent algorithm")
 
 set(CMAKE_CXX_STANDARD 11)
 
-# Standard mode including all distance functions (slow)
-set(CMAKE_CXX_FLAGS
+
+# Enable the fast mode option that uses only euclidean and cosine metrics
+option(FAST_MODE "Enable fast compile mode" OFF)
+
+# Standard mode includes all distance functions (slow)
+set(STANDARD_CXX_FLAGS
     "-Wall -g -pg -Ofast -DALL_METRICS -march=native -flto -fno-math-errno -fopenmp -pg")
 
-# Debug compile mode (only euclidean/cosine)
-# set(CMAKE_CXX_FLAGS
-    # "-Wall -g -pg -Ofast -march=native -flto -fno-math-errno -fopenmp -pg")
+# Fast mode (only euclidean/cosine distance)
+set(FAST_CXX_FLAGS
+    "-Wall -g -pg -Ofast -march=native -flto -fno-math-errno -fopenmp -pg")
+
+# Select the appropriate compiler flags
+if(FAST_MODE)
+    set(CMAKE_CXX_FLAGS "${FAST_CXX_FLAGS}")
+else()
+    set(CMAKE_CXX_FLAGS "${STANDARD_CXX_FLAGS}")
+endif()
+
 
 # Collect source files
 file(GLOB_RECURSE SRC_FILES src/*.cpp)
@@ -24,6 +36,7 @@ add_library(nndescent STATIC ${SRC_FILES})
 # Source code
 target_include_directories(nndescent PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/src)
 
+
 # Tests
 add_executable(test_distances tests/test_distances.cpp)
 target_link_libraries(test_distances PRIVATE nndescent)

MANIFEST.in

@@ -0,0 +1,9 @@
+# To upload this project to PyPI, follow these steps:
+#
+# python setup.py sdist
+# twine upload dist/*
+#
+# When prompted for the username, provide __token__
+
+# Includes the header files in the resulting .gz file for PyPI
+include src/*.h

README.md

@@ -46,14 +46,30 @@ Please note that not all distances have undergone thorough testing. Therefore, i
 
 ## Installation
 
+### From PyPI
+
+You can install nndescent directly from PyPI using pip:
+
+```sh
+pip install nndescent
+```
+
+If you want to run the examples in `tests`, additional packages are needed. You can install them manually or install nndescent with the full option:
+
+```sh
+pip install nndescent[full]
+```
+
+### From Source
+
 1. Clone the repository:
 
 ```sh
 git clone https://github.com/brj0/nndescent.git
 cd nndescent
 ```
 
-2. The project can by build with:
+2. Build and install the package:
 
 ```sh
 pip install .
@@ -71,7 +87,6 @@ pip install .[full]
 python tests/make_test_data.py
 ```
 
-
 ## Usage
 
 In Python you can utilize the nndescent library in the following way:

pybindings/pybind11ings.cpp

@@ -19,7 +19,7 @@ using namespace nndescent;
 const Parms DEFAULT_PARMS;
 
 
-/**
+/*
  * @brief Convert a nndescent::Matrix<T> object to a NumPy array in Python.
  */
 template<class T>
@@ -71,7 +71,7 @@ CSRMatrix<T> to_csr_matrix(py::object &py_obj)
 }
 
 
- /**
+/**
  * @brief Wrapper class for binding the NND (Nearest Neighbor Descent)
  * algorithm in Python.
  */

setup.py

@@ -1,5 +1,7 @@
 from setuptools import Extension, setup
+import glob
 import numpy as np
+import os
 import re
 import pybind11
 
@@ -18,43 +20,49 @@ def get_version_string():
     np.get_include(),
 ]
 
-module = Extension(
-    name="nndescent",
-    sources=[
-        "pybindings/pybind11ings.cpp",
-        "src/dtypes.cpp",
-        "src/nnd.cpp",
-        "src/rp_trees.cpp",
-        "src/utils.cpp",
-        "src/distances.cpp",
-    ],
-    include_dirs=include_dirs,
-    extra_compile_args=[
+# Get the value of the NND_DEBUG environment variable
+debug = os.getenv("NND_DEBUG") == "1"
+
+# Set the compiler flags based on the debug flag
+if debug:
+    print("\n*** Debugging compiler flags are used. ***\n")
+    compile_args = [
+        "-O0",
+        "-Wall",
+        "-Wextra",
+        "-fno-stack-protector",
+        "-g",
+        "-pg",
+    ]
+else:
+    compile_args = [
+        "-DALL_METRICS",
         "-Ofast",
         "-flto",
-        "-DALL_METRICS",
         "-fno-math-errno",
         "-fopenmp",
         "-g",
         "-march=native",
-    ],
-    # CXXFLAGS for debugging
-    # extra_compile_args=[
-        # "-O0",
-        # "-Wall",
-        # "-Wextra",
-        # "-fno-stack-protector",
-        # "-g",
-        # "-pg",
-    # ],
+    ]
+
+module = Extension(
+    name="nndescent",
+    sources=glob.glob("pybindings/*.cpp") + glob.glob("src/*.cpp"),
+    include_dirs=include_dirs,
+    extra_compile_args=compile_args,
     extra_link_args=["-fopenmp"],
     language="c++",
 )
 
+with open("README.md", "r") as f:
+    long_description = f.read()
+
 setup(
     name="nndescent",
     version=get_version_string(),
     description="C++ extension implementing nearest neighbour descent",
+    long_description=long_description,
+    long_description_content_type="text/markdown",
     install_requires=[
         "numpy",
     ],
@@ -69,4 +77,19 @@ def get_version_string():
         ],
     },
     ext_modules=[module],
+    keywords="nearest neighbor, knn, ANN",
+    license="BSD 2-Clause License",
+    author="Jon Brugger",
+    url="https://github.com/brj0/nndescent",
+    classifiers=[
+        "Development Status :: 3 - Alpha",
+        "Intended Audience :: Developers",
+        "Intended Audience :: Science/Research",
+        "License :: OSI Approved :: BSD License",
+        "Operating System :: POSIX",
+        "Programming Language :: C++",
+        "Programming Language :: Python :: 3",
+        "Topic :: Scientific/Engineering",
+        "Topic :: Software Development",
+    ],
 )

src/distances.h

@@ -2201,7 +2201,6 @@ float dot(Iter0 first0, Iter0 last0, Iter1 first1)
 }
 
 
-
 /*
  * @brief Sparse dot.
  */

src/dtypes.h

@@ -1001,8 +1001,8 @@ class HeapList
     int checked_push(size_t i, int idx, KeyType key, char flag);
 
     /*
-     * Pushes a node with the specified index, key, and flag into the
-     * specified heap if its key is smaller and it is not already in the heap.
+     * Pushes a node with the specified index and key into the specified heap
+     * if its key is smaller and it is not already in the heap.
      *
      * @param i The index of the heap.
      * @param idx The index of the node.