feat: Dynamically Linked Library in CPP #11439

soumiiow · 2024-11-05T13:11:52Z

Related to prestodb/presto#23634 in the Prestissimo space
and based off of the following PR: https://github.com/facebookincubator/velox/pull/1005/files

These changes will allow users to dynamically load functions in prestissimo using cpp. The Presto Server will use this library to dynamically load User Defined Functions (UDFs), connectors, or types.

an example of dynamically registering a function is also provided for reference, along with a unit test

netlify · 2024-11-05T13:12:12Z

✅ Deploy Preview for meta-velox canceled.

Name	Link
🔨 Latest commit	`21dae50`
🔍 Latest deploy log	https://app.netlify.com/sites/meta-velox/deploys/67a403bb4132430008afcd5c

pedroerp · 2024-11-05T15:50:57Z

@soumiiow thanks for looking into this. Out of curiosity, why doesn't this work in MacOS?

pedroerp · 2024-11-05T15:52:59Z

velox/common/CMakeLists.txt

@@ -15,6 +15,7 @@ add_subdirectory(base)
 add_subdirectory(caching)
 add_subdirectory(compression)
 add_subdirectory(config)
+add_subdirectory(dynamicRegistry)


we use snake case for directory names "dynamic_registry"

pedroerp

Very cool! I few small comments but overall looks good.

pedroerp · 2024-11-05T15:53:27Z

velox/common/dynamicRegistry/DynamicLibraryLoader.cpp

+#include <dlfcn.h>
+#include <iostream>
+#include "velox/common/base/Exceptions.h"
+namespace facebook::velox {


nit: new line before namespace definition.

pedroerp · 2024-11-05T15:55:15Z

velox/common/dynamicRegistry/DynamicLibraryLoader.cpp

+    VELOX_USER_FAIL("Couldn't find Velox registry symbol: {}", error);
+  }
+  registryItem();
+  std::cout << "LOADED DYLLIB 1" << std::endl;


for consistency, could you use LOG(INFO) and print the file name / path of the library loaded?

pedroerp · 2024-11-05T15:56:10Z

velox/common/dynamicRegistry/DynamicLibraryLoader.cpp

+
+static constexpr const char* kSymbolName = "registry";
+
+void loadDynamicLibraryFunctions(const char* fileName) {


we can probably omit the "Functions" from the name, and this can be used to really load anything, as long as you provide the registration functions. Let's name it loadDynamicLibrary()

pedroerp · 2024-11-05T15:57:02Z

velox/common/dynamicRegistry/README.md

+### 1. Create a cpp file for your dynamic library
+For dynamically loaded function registration, the format followed is mirrored of that of built-in function registration with some noted differences. Using [MyDynamicTestFunction.cpp](tests/MyDynamicTestFunction.cpp) as an example, the function uses the extern "C" keyword to protect against name mangling. A registry() function call is also necessary here.
+
+### 2. Register functions dynamically by creating .dylib or .so shared libraries and dropping them in a plugin directory


nit: the titles are too long; maybe just add the docs as a refular numbered list?

I pushed it out without the title formatting but does this look a bit cluttered now?

pedroerp · 2024-11-05T15:58:20Z

velox/common/dynamicRegistry/tests/DynamicLinkTest.cpp

+  auto signaturesBefore = getFunctionSignatures().size();
+
+  // Function does not exist yet.
+  EXPECT_THROW(dynamicFunction(0), VeloxUserError);


can you use VELOX_ASSERT_THROW() instead to validate the right exception is being thrown?

pedroerp · 2024-11-05T15:59:22Z

velox/common/dynamicRegistry/tests/CMakeLists.txt

+# `MyDynamicFunction.cpp` as a small .so library, and use the
+# MY_DYNAMIC_FUNCTION_LIBRARY_PATH macro to locate the .so binary.
+add_compile_definitions(
+  MY_DYNAMIC_FUNCTION_LIBRARY_PATH="${CMAKE_CURRENT_BINARY_DIR}")


please vendor the macro. Maybe something like VELOX_TEST_DYNAMIC_LIBRARY_PATH

majetideepak · 2024-11-05T17:57:06Z

velox/common/dynamicRegistry/DynamicLibraryLoader.cpp

+ * limitations under the License.
+ */
+
+#include "velox/common/dynamicRegistry/DynamicLibraryLoader.h"


this header include is not required.

majetideepak · 2024-11-05T18:02:38Z

velox/common/dynamicRegistry/tests/DynamicLinkTest.cpp

+
+  // Dynamically load the library.
+  std::string libraryPath = MY_DYNAMIC_FUNCTION_LIBRARY_PATH;
+  libraryPath += "/libvelox_function_my_dynamic.so";


Use CMAKE_SHARED_LIBRARY_SUFFIX and CMAKE_SHARED_LIBRARY_PREFIX to support MacOS.

https://stackoverflow.com/questions/32445070/how-does-cmake-know-which-prefixes-and-suffixes-to-add-to-shared-libraries
https://cmake.org/cmake/help/v3.0/variable/CMAKE_SHARED_LIBRARY_PREFIX.html
https://cmake.org/cmake/help/v3.0/variable/CMAKE_SHARED_LIBRARY_SUFFIX.html

What else is an issue for MacOS?

assignUser · 2024-11-12T19:54:41Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+  ${GMock}
+  ${GTEST_BOTH_LIBRARIES})


please use the GTest:: targets

assignUser · 2024-11-12T19:56:46Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+# To test functions being added by dynamically linked libraries, we compile
+# `MyDynamicFunction.cpp` as a small .so library, and use the
+# VELOX_TEST_DYNAMIC_LIBRARY_PATH macro to locate the .so binary.
+add_compile_definitions(


Please use target_compile_definitions( on the relevant target instead.

assignUser · 2024-11-12T19:58:15Z

velox/common/dynamic_registry/CMakeLists.txt

+if(${VELOX_BUILD_TESTING})
+  add_subdirectory(tests)
+endif()
+velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)


Suggested change

velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)

velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)

velox_link_libraries(velox_dynamic_function_loader PRIVATE velox_exception)

assignUser · 2024-11-12T19:59:38Z

velox/common/dynamic_registry/README.md

Thanks for adding docs!

velox/common/dynamic_registry/tests/CMakeLists.txt

majetideepak · 2024-11-12T19:59:51Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+# `MyDynamicFunction.cpp` as a small .so library, and use the
+# VELOX_TEST_DYNAMIC_LIBRARY_PATH macro to locate the .so binary.
+add_compile_definitions(
+  VELOX_TEST_DYNAMIC_LIBRARY_PATH="${CMAKE_CURRENT_BINARY_DIR}")


MacOS support is still missing. You can create the full library path here based on the CMake options I shared earlier.

velox/common/dynamic_registry/tests/MyDynamicFunction.cpp

majetideepak · 2024-11-12T20:02:07Z

velox/common/dynamic_registry/README.md

@@ -0,0 +1,22 @@
+# Velox: Dynamically Loading Registry Libraries in C++


"Dynamic Loading of Velox Extensions" is probably a better title.

majetideepak · 2024-11-12T20:03:01Z

velox/common/dynamic_registry/README.md

@@ -0,0 +1,22 @@
+# Velox: Dynamically Loading Registry Libraries in C++
+
+This library adds the ability to load User Defined Functions (UDFs), connectors, or types without having to fork and build Prestissimo, through the use of shared libraries that a Prestissimo worker can access. These are to be loaded on launch of the Presto server. The Presto server searches for any .so or .dylib files and loads them using this library.


Prestissimo -> Velox. Remaining paragraph as well.

majetideepak · 2024-11-12T20:04:24Z

velox/common/dynamic_registry/README.md

+target_link_libraries(name_of_dynamic_fn PRIVATE xsimd fmt::fmt velox_expression)
+```
+
+3. In the Prestissimo worker's config.properties file, set the plugin.dir property


Move this to Prestissimo.

majetideepak · 2024-11-12T20:04:33Z

velox/common/dynamic_registry/README.md

+```
+plugin.dir="User\Test\Path\plugin"
+```
+4. When the worker or the sidecar process starts, it will scan the plugin directory and attempt to dynamically load all shared libraries


Move this to Prestissimo.

majetideepak · 2024-11-12T20:05:33Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+
+namespace facebook::velox {
+
+/// Dynamically opens and registers functions defined in a shared library (.so)


Remove (.so)
Add fullstop.

majetideepak · 2024-11-12T20:07:50Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+
+/// Dynamically opens and registers functions defined in a shared library (.so)
+///
+/// Given a shared library name (.so), this function will open it using dlopen,


Opens a shared library using dlopen, looks for the symbol registry, and invokes it.

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

aditi-pandit

Thanks @soumiiow. Had bunch of minor comments, except for a bigger one around testing.

velox/common/dynamic_registry/CMakeLists.txt

aditi-pandit · 2024-11-13T05:35:23Z

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

+
+  // Lookup the symbol.
+  void* registrySymbol = dlsym(handler, kSymbolName);
+  auto registryItem = reinterpret_cast<void (*)()>(registrySymbol);


Nit : rename registryFunction

hey!! so i got some previous feedback to stay away from the "registryFunction" in the naming so as to not make it seem like this library is to be used exclusively for functions, and to move away from our initial design which was made with only the function loading in mind. Perhaps, would there be a better name for this variable than the work "item"? I can only rlly think of registryItem or registryPtr but would love to hear your suggestions too

@soumiiow : To me this is almost like the "main" function in a executable program. How about "loadLibrary" or "loadUserLibrary" or "enterUserLibrary" ? There could be code beyond registration here as well.

aditi-pandit · 2024-11-13T05:36:07Z

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

+  if (error != nullptr) {
+    VELOX_USER_FAIL("Couldn't find Velox registry symbol: {}", error);
+  }
+  registryItem();


Add a comment "Invoke the registry function"

velox/common/dynamic_registry/tests/MyDynamicFunction.cpp

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

velox/common/dynamic_registry/tests/MyDynamicFunction.cpp

aditi-pandit · 2024-11-13T06:07:54Z

velox/common/dynamic_registry/README.md

+target_link_libraries(name_of_dynamic_fn PRIVATE xsimd fmt::fmt velox_expression)
+```
+
+3. In the Prestissimo worker's config.properties file, set the plugin.dir property


This is not relevant in Velox. And also since its not used anywhere in the current code, its hard to put this in picture.

aditi-pandit · 2024-11-13T06:18:03Z

velox/common/dynamic_registry/README.md

@@ -0,0 +1,22 @@
+# Velox: Dynamically Loading Registry Libraries in C++
+
+This library adds the ability to load User Defined Functions (UDFs), connectors, or types without having to fork and build Prestissimo, through the use of shared libraries that a Prestissimo worker can access. These are to be loaded on launch of the Presto server. The Presto server searches for any .so or .dylib files and loads them using this library.


It might be good to not talk about Prestissimo in this README.

This is a generic utility for dynamically loading a "registry" function from a library. Its sufficient to just say that this is for "Extensibility" features that add custom user code which could include new Velox types, functions, operators and connectors.

aditi-pandit

@soumiiow : Thanks for updating your tests. Have a bunch of questions.

aditi-pandit · 2025-01-07T21:10:17Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+///
+///   void registry();
+///
+/// The registration function needs to be defined in the top-level namespace,


What do you mean by top-level namespace ? Is that common terminology ?

velox/common/dynamic_registry/DynamicLibraryLoader.h

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

velox/common/dynamic_registry/CMakeLists.txt

aditi-pandit · 2025-01-07T21:39:45Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+      dynamicFunction(0), "Scalar function doesn't exist: dynamic_2.");
+
+  // Dynamically load the library.
+  std::string libraryPath = VELOX_TEST_DYNAMIC_LIBRARY_PATH;


Write in a single line

std::string libraryPath = VELOX_TEST_DYNAMIC_LIBRARY_PATH .append("/libvelox_function_same_twice_my_dynamic") .append(VELOX_TEST_DYNAMIC_LIBRARY_PATH_SUFFIX);

or use fmt::format function for the same.

aditi-pandit · 2025-01-07T21:40:39Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+  const auto dynamicFunction = [&](std::optional<double> a) {
+    return evaluateOnce<int64_t>("dynamic_2()", a);
+  };
+  auto signaturesBefore = getFunctionSignatures().size();


Curious, what is the value of signaturesBefore ? I presume since you have derived from FunctionsTestBase this is non-zero ?

It might be good to have a CHECK where you are evaluating an expression that involves an existing function along with the new function you have registered.

@soumiiow : This comment is not addressed. Lets ensure that we can execute an expression that involves the function you have registered with something that was registered already.

aditi-pandit · 2025-01-07T21:43:28Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+      dynamicFunction(0), "Scalar function doesn't exist: dynamic_1.");
+
+  // Dynamically load the library.
+  std::string libraryPath = VELOX_TEST_DYNAMIC_LIBRARY_PATH;


Since this pattern of appending VELOX_TEST_DYNAMIC_LIBRARY_PATH (and SUFFIX) is repeated, we could abstract a function for it.

aditi-pandit · 2025-01-07T21:50:11Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+    "Expression evaluation result is not of expected type: dynamic_3() -> CONSTANT vector of type VARCHAR");
+
+  // Dynamically load the library.
+  std::string libraryPathInt = VELOX_TEST_DYNAMIC_LIBRARY_PATH;


Its interesting that the first function got over-ridden.. Is that because the we used structs with the same name "Dynamic123Function" for the function implementations ? If you used a different name for the structures are the functions retained ? From a system perspective, we wouldn't want the overriding to happen.

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

mohsaka · 2025-01-23T01:03:23Z

@soumiiow Addressed all of the unaddressed comments and made all of the changes for compilation on mac in this PR
#12111

Specifically this commit,
ac18115

aditi-pandit

Thanks @soumiiow. Have another round of comments.

aditi-pandit · 2025-01-30T19:48:52Z

velox/common/dynamic_registry/DynamicFunctionRegistrar.h

+
+namespace facebook::velox::common {
+template <template <class> class T, typename TReturn, typename... TArgs>
+void registerFunctionWrapper(


Its probably better to add this wrapper at the Prestissimo layer. We needn't have something in Velox.

aditi-pandit · 2025-01-30T19:51:45Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+  const auto dynamicFunction = [&](std::optional<double> a) {
+    return evaluateOnce<int64_t>("dynamic_2()", a);
+  };
+  auto signaturesBefore = getFunctionSignatures().size();


@soumiiow : This comment is not addressed. Lets ensure that we can execute an expression that involves the function you have registered with something that was registered already.

aditi-pandit · 2025-01-30T19:53:09Z

velox/common/dynamic_registry/tests/MyDynamicSameTwiceFunction.cpp

@@ -0,0 +1,45 @@
+/*


This file is not needed. You could use MyDynamicFunction twice for the same effect.

aditi-pandit · 2025-01-30T19:54:30Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+      dynamicFunction(0), "Scalar function doesn't exist: dynamic_2.");
+
+  std::string libraryPath =
+      getLibraryPath("libvelox_function_same_twice_my_dynamic");


You can use the same function as the previous test. A new one is not needed.

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

mohsaka · 2025-01-31T04:22:16Z

@aditi-pandit PR Opened here to address the comments.
Removal of duplicate dynamic library, Removal of the wrapper, rename of the dynamic libraries to represent what they are testing.

soumiiow#1

Should be merged into this branch soon.

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

velox/common/dynamic_registry/README.md

velox/common/dynamic_registry/tests/CMakeLists.txt

majetideepak · 2025-01-31T21:08:33Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+else()
+  set(CMAKE_DYLIB_TEST_LINK_LIBRARIES fmt::fmt xsimd)
+
+  target_link_libraries(velox_function_my_dynamic


target_link_libraries is the same for APPLE and Linux? Can we make this common?

I see because of fmt wasn't needed on mac due to it being part of folly

@soumiiow Can you check if we can use folly:folly instead of fmt::fmt on mac so we can standardize it?

@majetideepak looks like adding folly to linux gets us this error:

ERROR: something wrong with flag 'folly_hazptr_use_executor' in file '/root/dylib/presto/presto-native-execution/scripts/deps-download/folly/folly/synchronization/Hazptr.cpp'. One possibility: file '/root/dylib/presto/presto-native-execution/scripts/deps-download/folly/folly/synchronization/Hazptr.cpp' is being linked both statically and dynamically into this executable.

Can we include only fmt in both?

Unfortunately that leads to

[build] /Users/michaelohsaka/michael/velox/./velox/common/base/VeloxException.h:26:10: fatal error: 'glog/logging.h' file not found [build] 26 | #include <glog/logging.h> [build] | ^~~~~~~~~~~~~~~~ [build] 1 error generated.

I can include glog::glog though and that works fine. Maybe we can try that on linux? @soumiiow

Let's include glog::glog. Folly is known to cause this is being linked both statically and dynamically into this executable when built dynamically because of gflags.

@majetideepak @aditi-pandit adding glog::glog works. pushed up the latest changes without the if(apple) bit. please take a look

glog header can be removed. I opened a PR here
#12248

majetideepak · 2025-01-31T21:09:11Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+                      "-Wl,-undefined,dynamic_lookup")
+
+else()
+  set(CMAKE_DYLIB_TEST_LINK_LIBRARIES fmt::fmt xsimd)


Why is this different for APPLE?

velox/common/dynamic_registry/README.md

majetideepak · 2025-02-03T15:00:15Z

@soumiiow, @mohsaka Let's unify APPLE and LINUX as much as possible to simplify the user experience.

aditi-pandit · 2025-02-03T16:41:52Z

@soumiiow, @mohsaka Let's unify APPLE and LINUX as much as possible to simplify the user experience.

Second that. Lets have any dependencies at a minimal.

czentgr · 2025-02-03T19:28:52Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+      getLibraryPath("libvelox_overload_int_function_my_dynamic");
+  loadDynamicLibrary(libraryPathInt.data());
+
+  // The first function loaded should NOT be rewritten.


Nit: overwritten instead of overwritten and we end up with two functions with name dynamic_overload having varchar and bigint respectively?

Yep. The function should be overloaded.

changed the comment @czentgr!

aditi-pandit

This is the last round from me. Thanks @soumiiow and @mohsaka

aditi-pandit · 2025-02-03T22:28:31Z

velox/functions/DynamicUdf.h

@@ -0,0 +1,25 @@
+/*


Any reason this is added here instead of velox/common folder ?

The main reason is because Udf.h is there. I can move it though.

Moved to common/dynamic_registry

aditi-pandit · 2025-02-03T22:29:21Z

velox/functions/DynamicUdf.h

+
+#include "velox/functions/Udf.h"
+
+extern "C" {


Please add a header comment to this file explaining why it is needed.

aditi-pandit · 2025-02-04T01:14:51Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+
+  std::string libraryPath = getLibraryPath("libvelox_function_my_dynamic");
+
+  loadDynamicLibrary(libraryPath.data());


This presumes that the first test has executed before this one. gtest doesn't gives this guarantee. Each test should be self-contained. We can load the same library twice within the test itself for the effect.

Fixed by loading once before with no checks except for the function existing.

majetideepak · 2025-02-04T18:17:29Z

velox/common/dynamic_registry/CMakeLists.txt

+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)


Should we name the library similar to naming everywhere?
velox_dynamic_function_loader -> velox_dynamic_library_loader

majetideepak · 2025-02-04T18:20:40Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+/// locating the function it executes the registration bringing the UDFs in the
+/// scope of the Velox runtime.
+///
+/// If the same library is loaded twice then a no-op scenerio will happen.


Can you explain why this is a no-op? Isn't registering a Velox component twice an error?

Added explanation. Velox functions overwrite by default. So it will overwrite its function with itself.

This behavior is not true for all Velox modules. Connectors will fail for example. Let's say something like
Loading a library twice can cause a module to be registered twice. This can fail for certain Velox modules.

Fixed and addressed all comments. Waiting for @soumiiow to merge in the branch.

https://github.com/soumiiow/velox/pull/4/files

Thanks. Let's use component instead of module since we used component on top.

merged the branch and pushed the component vs module change out

majetideepak · 2025-02-04T18:21:38Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+
+/// The function uses dlopen to load the shared library.
+/// It then searches for the "void registry()" C function which typically
+/// contains all the registration code for the UDFs defined in library. After


Maybe replace UDFs -> user-defined Velox components such as functions

Highlighting this because hasn't been changed yet.

majetideepak · 2025-02-04T18:26:33Z

velox/common/dynamic_registry/DynamicUdf.h

+
+extern "C" {
+void registry();
+void check() {


Let's comment here that The "registry()" declaration and "check()" function ensure the correct registry signature function is defined by the user.

majetideepak · 2025-02-04T18:31:00Z

velox/common/dynamic_registry/README.md

+    add_library(name_of_dynamic_fn SHARED TestFunction.cpp)
+    target_link_libraries(name_of_dynamic_fn PRIVATE xsimd fmt::fmt)
+    ```
+    Above, the xsimd and fmt::fmt libraries are required for all necessary symbols to be defined when loading the TestFunction.cpp dynamically


This has to be updated.

Updated with only one example, and a short reasoning on what can be skipped on linux.

Summary: This is causing the xsimd header leak. See #11439 (comment) Pull Request resolved: #12231 Reviewed By: bikramSingh91 Differential Revision: D68982274 Pulled By: kgpai fbshipit-source-id: 55624bc0161460fc8eff317329af6f009aeb0f94

czentgr · 2025-02-05T15:59:31Z

velox/common/dynamic_registry/tests/CMakeLists.txt

+  PRIVATE ${CMAKE_DYLIB_TEST_LINK_LIBRARIES})
+
+target_link_options(velox_function_my_dynamic PRIVATE
+                    "-Wl,-undefined,dynamic_lookup")


I'm thinking this could be a variable like

set(COMMON_LIBRARY_LINK_OPTIONS "-Wl,-undefined,dynamic_lookup")

and then use

target_link_options(velox_function_my_dynamic PRIVATE ${COMMON_LIBRARY_LINK_OPTIONS})

czentgr · 2025-02-05T16:11:56Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+}
+
+TEST_F(DynamicLinkTest, dynamicLoadOverwriteFunc) {
+  const auto dynamicIntFunction = [&](std::optional<double> a) {


This should be have no argument at all.
evaluateOnce should allow this

return evaluateOnce<int64_t>("dynamic_overwrite()");

The overwrite only occurs because the signature (aka no input args) are the same for both functions. The difference is the return type. Am I correct?
Same for the varchar return type function.

Removed the arguments, but had to put something in for evaluateOnce for the row or else it will be a nullptr result. Followed example from pi() and e().

Changed to

const auto dynamicIntFunction = [&]() { return evaluateOnce<int64_t>("dynamic_overwrite()", makeRowVector(ROW({}), 1)); };

Correct, the only difference is the return type. The rest of the signature is the same.

czentgr · 2025-02-05T16:13:47Z

velox/common/dynamic_registry/README.md

@@ -0,0 +1,26 @@
+# Dynamic Loading of Velox Extensions


Can we please add a note to the overwrite vs overload topic at the bottom to explain to users when either case occurs?

aditi-pandit

Thanks @mohsaka and @soumilw

velox/common/dynamic_registry/DynamicLibraryLoader.cpp

aditi-pandit · 2025-02-05T16:24:11Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+///
+/// void registry();
+///
+/// The registration function needs to be defined in the root-level namespace,


Is root-level namespace the global namespace ?

Should be the same thing. Root, ex the base level. I can rename it to global as well.

aditi-pandit · 2025-02-05T16:28:21Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+
+class DynamicLinkTest : public FunctionBaseTest {};
+
+std::string getLibraryPath(const std::string& filename) {


Wrap this in an anonymous namespace.

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

aditi-pandit · 2025-02-05T19:01:54Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+}
+
+TEST_F(DynamicLinkTest, dynamicLoadSameFuncTwice) {
+  const auto dynamicFunction = [&](std::optional<double> a) {


Like @czentgr's comment is the argument for "a" needed here as well ?

Removed the other ones as well.

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

czentgr · 2025-02-05T19:09:32Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+
+/// The function uses dlopen to load the shared library.
+/// It then searches for the "void registry()" C function which typically
+/// contains all the registration code for the UDFs defined in library. After


Highlighting this because hasn't been changed yet.

czentgr · 2025-02-05T19:13:02Z

velox/common/dynamic_registry/DynamicLibraryLoader.h

+/// Loading a library twice can cause a components to be registered twice.
+/// This can fail for certain Velox components.
+
+void loadDynamicLibrary(const char* fileName);


Why are we using char* here instead of const std::string&? Is it because we don't want to include ?

This function is only used in velox itself so this wouldn't be a problem if we are concerned about externalizing it.

czentgr · 2025-02-05T19:14:55Z

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp

+      dynamicFunction(0), "Scalar function doesn't exist: dynamic.");
+
+  std::string libraryPath = getLibraryPath("libvelox_function_my_dynamic");
+  loadDynamicLibrary(libraryPath.data());


If we use a const char* as input then we need to use libraryPath().c_str() and not data(). There is no guarantee data is null terminated.

This occurs multiple times.

switched to str

…uccessful mac & linux compilation

aditi-pandit · 2025-02-06T07:12:41Z

velox/common/dynamic_registry/tests/MyDynamicErrFunction.cpp

+namespace facebook::velox::common::dynamicRegistry {
+
+template <typename T>
+struct Dynamic123Function {


@mohsaka, @soumiiow : Nit : Its better to avoid using notation like 123 in the function name. Use something like DynamicTestFunction or just DynamicFunction for the struct names everywhere.

Also remove the usage of "My" in the filenames... They can be DynamicTestFunction or just DynamicFunction to be consistent with the struct name.

soumiiow requested a review from majetideepak as a code owner November 5, 2024 13:11

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Nov 5, 2024

Yuhta requested a review from pedroerp November 5, 2024 15:40

pedroerp reviewed Nov 5, 2024

View reviewed changes

majetideepak reviewed Nov 5, 2024

View reviewed changes

soumiiow force-pushed the velox-dylib branch from 789da39 to c5cbea2 Compare November 12, 2024 15:42

soumiiow requested a review from assignUser as a code owner November 12, 2024 15:42

soumiiow force-pushed the velox-dylib branch from c5cbea2 to f8afbc0 Compare November 12, 2024 15:51

soumiiow requested review from pedroerp and majetideepak November 12, 2024 16:59

assignUser reviewed Nov 12, 2024

View reviewed changes

majetideepak reviewed Nov 12, 2024

View reviewed changes

aditi-pandit reviewed Nov 13, 2024

View reviewed changes

soumiiow force-pushed the velox-dylib branch 2 times, most recently from 45cf7af to 4e82b71 Compare December 4, 2024 07:04

soumiiow force-pushed the velox-dylib branch from 4e82b71 to ae2ad2e Compare December 17, 2024 13:51

soumiiow force-pushed the velox-dylib branch 2 times, most recently from b6660ad to eb20996 Compare January 7, 2025 16:55

soumiiow mentioned this pull request Jan 7, 2025

Dynamically Linked Library in Presto CPP prestodb/presto#24330

Open

6 tasks

aditi-pandit reviewed Jan 7, 2025

View reviewed changes

soumiiow mentioned this pull request Jan 14, 2025

presto requires namespaces when using function registration prestodb/presto#24363

Open

mohsaka mentioned this pull request Jan 17, 2025

[WIP] MacOS CMake compatibility for dynamic functions #12111

Closed

soumiiow force-pushed the velox-dylib branch from eb20996 to d0b34fc Compare January 25, 2025 02:12

soumiiow changed the title ~~Dynamically Linked Library in CPP~~ feat: Dynamically Linked Library in CPP Jan 27, 2025

soumiiow force-pushed the velox-dylib branch from d0b34fc to ff3bd9c Compare January 27, 2025 22:11

aditi-pandit reviewed Jan 30, 2025

View reviewed changes

velox/common/dynamic_registry/tests/DynamicLinkTest.cpp Show resolved Hide resolved

majetideepak reviewed Jan 31, 2025

View reviewed changes

soumiiow force-pushed the velox-dylib branch from 5ccc106 to 1a34be6 Compare January 31, 2025 19:30

majetideepak reviewed Jan 31, 2025

View reviewed changes

majetideepak mentioned this pull request Jan 31, 2025

refactor: Remove unused header in functions/Registerer.h #12231

Closed

soumiiow force-pushed the velox-dylib branch from 1a34be6 to ad51fb2 Compare February 1, 2025 00:52

soumiiow force-pushed the velox-dylib branch 2 times, most recently from ff3c929 to a9b49ab Compare February 3, 2025 17:34

czentgr reviewed Feb 3, 2025

View reviewed changes

soumiiow force-pushed the velox-dylib branch 2 times, most recently from 3afb947 to 480e169 Compare February 3, 2025 20:11

majetideepak mentioned this pull request Feb 3, 2025

refactor: Remove unused headers and gflags #12248

Open

aditi-pandit reviewed Feb 4, 2025

View reviewed changes

soumiiow force-pushed the velox-dylib branch from 480e169 to d7ed141 Compare February 4, 2025 17:51

majetideepak reviewed Feb 4, 2025

View reviewed changes

soumiiow force-pushed the velox-dylib branch 2 times, most recently from bda5649 to 020aeda Compare February 5, 2025 16:14

czentgr reviewed Feb 5, 2025

View reviewed changes

mohsaka force-pushed the velox-dylib branch from 020aeda to 76fd0a0 Compare February 5, 2025 18:54

aditi-pandit reviewed Feb 5, 2025

View reviewed changes

czentgr reviewed Feb 5, 2025

View reviewed changes

dylib functionality + namespace wrapper and other review comments + s…

21dae50

…uccessful mac & linux compilation

mohsaka force-pushed the velox-dylib branch from 76fd0a0 to 21dae50 Compare February 6, 2025 00:35

aditi-pandit mentioned this pull request Feb 6, 2025

feat(functions): Add support for REST based remote functions #10911

Open

aditi-pandit reviewed Feb 6, 2025

View reviewed changes


		static constexpr const char* kSymbolName = "registry";

		void loadDynamicLibraryFunctions(const char* fileName) {

	velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)
	velox_add_library(velox_dynamic_function_loader DynamicLibraryLoader.cpp)
	velox_link_libraries(velox_dynamic_function_loader PRIVATE velox_exception)

		@@ -0,0 +1,22 @@
		# Velox: Dynamically Loading Registry Libraries in C++

		@@ -0,0 +1,22 @@
		# Velox: Dynamically Loading Registry Libraries in C++

		This library adds the ability to load User Defined Functions (UDFs), connectors, or types without having to fork and build Prestissimo, through the use of shared libraries that a Prestissimo worker can access. These are to be loaded on launch of the Presto server. The Presto server searches for any .so or .dylib files and loads them using this library.


		namespace facebook::velox {

		/// Dynamically opens and registers functions defined in a shared library (.so)


		std::string libraryPath = getLibraryPath("libvelox_function_my_dynamic");

		loadDynamicLibrary(libraryPath.data());


		class DynamicLinkTest : public FunctionBaseTest {};

		std::string getLibraryPath(const std::string& filename) {

feat: Dynamically Linked Library in CPP #11439

Are you sure you want to change the base?

feat: Dynamically Linked Library in CPP #11439

Conversation

soumiiow commented Nov 5, 2024 • edited Loading

netlify bot commented Nov 5, 2024 • edited Loading

✅ Deploy Preview for meta-velox canceled.

pedroerp commented Nov 5, 2024

Choose a reason for hiding this comment

pedroerp left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

majetideepak Nov 5, 2024 • edited Loading

Choose a reason for hiding this comment

majetideepak Nov 5, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aditi-pandit left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aditi-pandit left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mohsaka commented Jan 23, 2025

aditi-pandit left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mohsaka commented Jan 31, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

soumiiow Feb 1, 2025 • edited Loading

Choose a reason for hiding this comment

majetideepak Feb 1, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

majetideepak commented Feb 3, 2025 • edited Loading

aditi-pandit commented Feb 3, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aditi-pandit left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

soumiiow commented Nov 5, 2024 •

edited

Loading

netlify bot commented Nov 5, 2024 •

edited

Loading

majetideepak Nov 5, 2024 •

edited

Loading

majetideepak Nov 5, 2024 •

edited

Loading

aditi-pandit left a comment •

edited

Loading

soumiiow Feb 1, 2025 •

edited

Loading

majetideepak Feb 1, 2025 •

edited

Loading

majetideepak commented Feb 3, 2025 •

edited

Loading

aditi-pandit Feb 4, 2025 •

edited

Loading

mohsaka Feb 4, 2025 •

edited

Loading

majetideepak Feb 4, 2025 •

edited

Loading