Use ryu and fastfloat for double serialization/deserialization

.github/workflows/build-and-test.yaml

@@ -45,6 +45,30 @@ jobs:
           cd build
           GEOARROW_TESTING_DIR=$GITHUB_WORKSPACE/testing ctest -T test --output-on-failure .
 
+      - name: Install lcov
+        run: |
+          sudo apt-get install lcov
+
+      - name: Calculate coverage
+        run: |
+          cd build
+          lcov --capture --directory . \
+            --exclude "*_test.cc" \
+            --exclude "/usr/*" \
+            --exclude "*/gtest/*" \
+            --exclude "*/ryu/*" \
+            --exclude "*/fast_float.h" \
+            --exclude "*/nanoarrow.*" \
+            --output-file coverage.info
+
+          lcov --list coverage.info
+
+      - name: Upload coverage
+        if: success()
+        uses: codecov/codecov-action@v2
+        with:
+          files: build/coverage.info
+
       - name: Test with memcheck
         run: |
           cd build
@@ -56,17 +80,3 @@ jobs:
         with:
           name: geoarrow-memcheck
           path: build/Testing/Temporary/MemoryChecker.*.log
-
-      - name: Calculate coverage
-        run: |
-          SOURCE_PREFIX=`pwd`
-          mkdir build/cov
-          cd build/cov
-          gcov -abcfu --source-prefix=$SOURCE_PREFIX `find ../CMakeFiles/geoarrow.dir/ -name "*.gcno"`
-          rm nanoarrow.h.gcov nanoarrow.c.gcov
-
-      - name: Upload coverage
-        if: success()
-        uses: codecov/codecov-action@v2
-        with:
-          directory: build/cov

CMakeLists.txt

@@ -6,7 +6,7 @@ if(NOT DEFINED CMAKE_C_STANDARD)
   set(CMAKE_C_STANDARD 99)
 endif()
 
-set(GEOARROW_VERSION "0.2.0-SNAPSHOT")
+set(GEOARROW_VERSION "0.1.0-SNAPSHOT")
 string(REGEX MATCH "^[0-9]+\\.[0-9]+\\.[0-9]+" GEOARROW_BASE_VERSION "${GEOARROW_VERSION}")
 project(geoarrow VERSION "${GEOARROW_BASE_VERSION}")
 
@@ -16,6 +16,8 @@ set(GEOARROW_VERSION_PATCH "${geoarrow_VERSION_PATCH}")
 
 option(GEOARROW_BUILD_TESTS "Build tests" OFF)
 option(GEOARROW_CODE_COVERAGE "Enable coverage reporting" OFF)
+option(GEOARROW_USE_FAST_FLOAT "Use fast_float for numeric value parsing" ON)
+option(GEOARROW_USE_RYU "Use ryu for numeric value printing" ON)
 
 add_library(coverage_config INTERFACE)
 
@@ -24,6 +26,27 @@ if(GEOARROW_CODE_COVERAGE)
   target_link_options(coverage_config INTERFACE --coverage)
 endif()
 
+if(GEOARROW_USE_FAST_FLOAT)
+  if(NOT DEFINED CMAKE_CXX_STANDARD)
+    set(CMAKE_CXX_STANDARD 11)
+  endif()
+  set(GEOARROW_DOUBLE_PARSE_SOURCE src/geoarrow/double_parse_fast_float.cc)
+  set(GEOARROW_USE_FAST_FLOAT_DEFINE "#define GEOARROW_USE_FAST_FLOAT 1")
+else()
+  set(GEOARROW_DOUBLE_PARSE_SOURCE src/geoarrow/double_parse_std.c)
+  set(GEOARROW_USE_FAST_FLOAT_DEFINE "#define GEOARROW_USE_FAST_FLOAT 0")
+endif()
+
+if(GEOARROW_USE_RYU)
+  set(GEOARROW_DOUBLE_PRINT_SOURCE src/geoarrow/double_print.c src/geoarrow/ryu/d2s.c)
+  set(GEOARROW_USE_RYU_DEFINE "#define GEOARROW_USE_RYU 1")
+else()
+  set(GEOARROW_DOUBLE_PRINT_SOURCE src/geoarrow/double_print.c)
+  set(GEOARROW_USE_RYU_DEFINE "#define GEOARROW_USE_RYU 0")
+endif()
+
+configure_file(src/geoarrow/geoarrow_config.h.in generated/geoarrow_config.h)
+
 add_library(
   geoarrow
   src/geoarrow/schema.c
@@ -38,11 +61,16 @@ add_library(
   src/geoarrow/wkb_writer.c
   src/geoarrow/wkt_reader.c
   src/geoarrow/wkt_writer.c
+  ${GEOARROW_DOUBLE_PARSE_SOURCE}
+  ${GEOARROW_DOUBLE_PRINT_SOURCE}
   src/geoarrow/nanoarrow.c)
 
 target_include_directories(geoarrow PUBLIC
-  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>
+  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src/geoarrow>
   $<INSTALL_INTERFACE:include>)
+target_include_directories(geoarrow
+  PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}/generated>
+)
 target_compile_definitions(geoarrow PUBLIC
   "$<$<CONFIG:Debug>:NANOARROW_DEBUG>"
   "$<$<CONFIG:Debug>:GEOARROW_DEBUG>")
@@ -75,6 +103,8 @@ endif()
 
 install(TARGETS geoarrow DESTINATION lib)
 install(DIRECTORY src/ DESTINATION include FILES_MATCHING PATTERN "*.h")
+install(FILES ${CMAKE_CURRENT_BINARY_DIR}/generated/geoarrow_config.h
+        DESTINATION include/geoarrow)
 
 if(GEOARROW_BUILD_TESTS)
   set(MEMORYCHECK_COMMAND_OPTIONS "--leak-check=full --suppressions=${CMAKE_CURRENT_LIST_DIR}/valgrind.supp --error-exitcode=1")

python/bootstrap.py

@@ -19,21 +19,25 @@
 import glob
 import shutil
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     this_dir = os.path.abspath(os.path.dirname(__file__))
-    vendor_dir = os.path.join(this_dir, 'src', 'geoarrow', 'geoarrow')
-    vendor_source_dir = os.path.join(this_dir, '..', 'src', 'geoarrow')
+    vendor_dir = os.path.join(this_dir, "src", "geoarrow", "geoarrow")
+    vendor_source_dir = os.path.join(this_dir, "..", "src", "geoarrow")
 
     if os.path.exists(vendor_dir):
         shutil.rmtree(vendor_dir)
 
-    vendor_source_files = glob.glob(os.path.join(vendor_source_dir, '*.c'))
-    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, '*.h'))
-    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, 'geoarrow.hpp'))
+    vendor_source_files = glob.glob(os.path.join(vendor_source_dir, "*.c"))
+    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, "*.h"))
+    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, "geoarrow.hpp"))
+    vendor_source_files += glob.glob(
+        os.path.join(vendor_source_dir, "double_parse_fast_float.cc")
+    )
+    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, "ryu/*.h"))
+    vendor_source_files += glob.glob(os.path.join(vendor_source_dir, "ryu/*.c"))
 
     os.mkdir(vendor_dir)
+    os.mkdir(os.path.join(vendor_dir, "ryu"))
     for source_file in vendor_source_files:
-        shutil.copyfile(
-            source_file,
-            os.path.join(vendor_dir, os.path.basename(source_file))
-        )
+        dst_file = source_file.removeprefix(os.path.join(vendor_source_dir, ""))
+        shutil.copyfile(source_file, os.path.join(vendor_dir, dst_file))

python/setup.py

@@ -32,8 +32,17 @@
     subprocess.run([sys.executable, bootstrap_py])
 
 vendor_dir = os.path.join(this_dir, "src", "geoarrow", "geoarrow")
-vendored_files = os.listdir(vendor_dir)
-sources = [f"src/geoarrow/geoarrow/{f}" for f in vendored_files if f.endswith(".c")]
+sources = [
+    f"src/geoarrow/geoarrow/{f}"
+    for f in os.listdir(vendor_dir)
+    if f.endswith(".c") or f.endswith(".cc")
+]
+
+sources += [
+    f"src/geoarrow/geoarrow/ryu/{f}"
+    for f in os.listdir(os.path.join(vendor_dir, "ryu"))
+    if f.endswith(".c")
+]
 
 
 # Workaround because setuptools has no easy way to mix C and C++ sources