Add python bindings (#69)

Co-authored-by: Daniel Mann <[email protected]>
Co-authored-by: Eugen Beck <[email protected]>

Author: 3 people

.gitignore

@@ -22,6 +22,7 @@
 *.pyc
 *.o
 *.so
+/RASR.egg-info
 
 # For exec-files, the postfix is ".$(OS)-$(PROC)-$(COMPILE)[-$(SUFFIX)]".
 *.*-*-*
@@ -53,6 +54,9 @@
 .history*
 *.autosave
 
+# Singularity/Apptainer images
+*.sif
+
 # Broken NFS.
 .nfs*
 

Modules.make

@@ -123,6 +123,7 @@ TOOLS += CorpusStatistics
 TOOLS += FeatureExtraction
 TOOLS += FeatureStatistics
 TOOLS += Fsa
+TOOLS += LibRASR
 TOOLS += Lm
 TOOLS += SpeechRecognizer
 TOOLS += Xml

apptainer/2023-11-08_tensorflow-2.14_v1/image.def

@@ -9,6 +9,8 @@ Stage: build
     DEBIAN_FRONTEND=noninteractive apt install -y wget git unzip gzip libssl-dev lsb-release zsh \
         bison libxml2-dev libopenblas-dev libsndfile1-dev libcrypto++-dev libcppunit-dev \
         parallel xmlstarlet python3-lxml htop strace gdb sox python3-pip cmake ffmpeg vim
+    
+    pip3 install pybind11==2.11.1
 
     # download the cache manager and place in /usr/local
     cd /usr/local

apptainer/2023-11-08_tensorflow-2.14_v1/makefiles/Modules.make

@@ -110,6 +110,9 @@ MODULES += MODULE_TEST
 # Tensorflow integration
 MODULES += MODULE_TENSORFLOW
 
+# ONNX integration
+# MODULES += MODULE_ONNX
+
 # define variables for the makefiles
 $(foreach module, $(MODULES), $(eval $(module) = 1))
 
@@ -120,6 +123,8 @@ TOOLS += CorpusStatistics
 TOOLS += FeatureExtraction
 TOOLS += FeatureStatistics
 TOOLS += Fsa
+TOOLS += LibRASR
+TOOLS += Lm
 TOOLS += SpeechRecognizer
 TOOLS += Xml
 

apptainer/2023-11-08_tensorflow-2.14_v1/makefiles/config/cc-gcc.make

@@ -22,6 +22,7 @@ CXX_MINOR = $(shell $(CXX) --version | head -n 1 | sed -e 's/.*[ \t]\([0-9]\)\.\
 # -----------------------------------------------------------------------------
 # compiler options
 DEFINES		+= -D_GNU_SOURCE
+CCFLAGS		+= -fPIC
 CCFLAGS		+= -pipe
 CCFLAGS		+= -funsigned-char
 CCFLAGS		+= -fno-exceptions

apptainer/2023-11-08_tensorflow-2.14_v1/makefiles/config/os-linux.make

@@ -144,6 +144,8 @@ LDFLAGS     += `${PYTHON_PATH}/bin/python3.11-config --ldflags --embed 2>/dev/nu
 LDFLAGS     += -Wl,-rpath -Wl,${PYTHON_PATH}/lib
 else
 INCLUDES    += `python3.11-config --includes 2>/dev/null`
+INCLUDES    += -I$(shell python3.11 -c 'import numpy as np; print(np.get_include())')
+INCLUDES    += `python3.11 -m pybind11 --includes 2>/dev/null`
 LDFLAGS     += `python3.11-config --ldflags --embed 2>/dev/null`
 # IF you want to use Python2 for whatever reason:
 # INCLUDES    += `pkg-config --cflags python`

config/cc-gcc.make

@@ -22,6 +22,7 @@ CXX_MINOR = $(shell $(CXX) --version | head -n 1 | sed -e 's/.*[ \t]\([0-9]\)\.\
 # -----------------------------------------------------------------------------
 # compiler options
 DEFINES		+= -D_GNU_SOURCE
+CCFLAGS		+= -fPIC
 CCFLAGS		+= -pipe
 CCFLAGS		+= -funsigned-char
 CCFLAGS		+= -fno-exceptions

config/os-linux.make

@@ -152,7 +152,9 @@ LDFLAGS     += `${PYTHON_PATH}/bin/python3-config --ldflags 2>/dev/null`
 LDFLAGS     += -Wl,-rpath -Wl,${PYTHON_PATH}/lib
 else
 INCLUDES    += `python3-config --includes 2>/dev/null`
-LDFLAGS     += `python3-config --ldflags 2>/dev/null`
+INCLUDES    += -I$(shell python3 -c 'import numpy as np; print(np.get_include())')
+INCLUDES    += `python3 -m pybind11 --includes 2>/dev/null`
+LDFLAGS     += `python3-config --ldflags --embed 2>/dev/null`
 # IF you want to use Python2 for whatever reason:
 # INCLUDES    += `pkg-config --cflags python`
 # LDFLAGS     += `pkg-config --libs python`

setup.py

@@ -0,0 +1,40 @@
+"""Setup.py inspired by https://thomastrapp.com/posts/building-a-pypi-package-for-a-modern-cpp-project/"""
+import os
+from setuptools import setup, dist
+
+PACKAGE_NAME = "librasr" # keyword to import
+LIBRASR_SUBDIR = os.path.join("src", "Tools", "LibRASR")
+LIBRASR_SO = "librasr.so"
+
+assert os.path.exists(os.path.join(LIBRASR_SUBDIR, LIBRASR_SO)), \
+    "Could not find librasr shared library. Did you compile it correctly?"
+class BinaryDistribution(dist.Distribution):
+    """Helper class: We assume that librasr.so has already been compiled."""
+    def has_ext_modules(self) -> bool:
+        return True
+
+# get README content for long description
+this_directory = os.path.abspath(os.path.dirname(__file__))
+with open(os.path.join(this_directory, 'README')) as f:
+    long_description = f.read()
+
+setup(
+    name='RASR',
+
+    # include shared library and link to lower case package name
+    packages=[PACKAGE_NAME],
+    package_dir={PACKAGE_NAME: LIBRASR_SUBDIR},
+    package_data={PACKAGE_NAME: [LIBRASR_SO]},
+    include_package_data=True,
+
+    description="RASR as a python module.",
+    long_description=long_description,
+    long_description_content_type="text/markdown",
+
+    # use custom distribution class to export librasr.so
+    distclass=BinaryDistribution,
+
+    version='0.0.1',
+    url='https://github.com/rwth-i6/rasr',
+    author_email='[email protected]',
+)

src/Core/Configuration.cc

@@ -169,7 +169,7 @@ void Configuration::Resource::writeUsage(XmlWriter& os) const {
  * Central storage place for all resources.
  */
 
-class Configuration::ResourceDataBase {
+class Configuration::ResourceDataBase : public ReferenceCounted {
 private:
     std::set<Resource> resources;
     Resource           noResource_;
@@ -401,24 +401,19 @@ bool Configuration::isWellFormedParameterName(const std::string& s) {
 }
 
 Configuration::Configuration()
-        : db_(0),
-          isDataBaseOwner_(true),
+        : db_(new ResourceDataBase),
           selection_("UNNAMED"),
           name_("UNNAMED") {
-    // define new root configuration
-    db_ = new ResourceDataBase;
 }
 
 Configuration::Configuration(const Configuration& c)
         : db_(c.db_),
-          isDataBaseOwner_(false),
           selection_(c.selection_),
           name_(c.name_) {
 }
 
 Configuration::Configuration(const Configuration& c, const std::string& add_selection)
-        : db_(c.db_),
-          isDataBaseOwner_(false) {
+        : db_(c.db_) {
     require(add_selection.length());
     require(add_selection.find(resource_separation_char) == std::string::npos);
     require(add_selection.find(resource_wildcard_char) == std::string::npos);
@@ -428,16 +423,9 @@ Configuration::Configuration(const Configuration& c, const std::string& add_sele
     ensure(isWellFormedParameterName(selection_));
 }
 
-Configuration::~Configuration() {
-    if (isDataBaseOwner_)
-        delete db_;
-}
+Configuration::~Configuration() {}
 
 Configuration& Configuration::operator=(const Configuration& c) {
-    if (isDataBaseOwner_) {
-        delete db_;
-        isDataBaseOwner_ = false;
-    }
     db_        = c.db_;
     selection_ = c.selection_;
     name_      = c.name_;