Regorus extension to invoke C# code by kusha · Pull Request #391 · microsoft/regorus

kusha · 2025-04-09T15:41:53Z

See example code to invoke C# code during policy evaluation: bindings/csharp/scripts/invoke.csx. bindings/csharp/scripts also contains compilation insturctions.

Changes:

FFI changes to register and unregister callbacks, invoked with invoke("name", args...) calls from policy
C# bindings update to support new FFI callbacks and register them as delegates
new "invoke" extension might be extended to other languages (e.g. C, C++), so it is wrapped as a built-in extension
built-in extensions can be loaded by name, using FFI without providing the implementations
compilation feature flags for built in extensions
by default extensions are considered as not thread-safe and are under mutex access (new "invoke" extension is no thread-safe).

Copilot

Copilot reviewed 7 out of 8 changed files in this pull request and generated no comments.

Files not reviewed (1)

bindings/csharp/Regorus/Regorus.csproj: Language not supported

anakrish

@kusha, thank you for taking on this super useful feature has has been in the back burner for a while.

anakrish · 2025-04-10T05:48:05Z

bindings/csharp/Regorus/Regorus.cs

+    /// </summary>
+    /// <param name="payload">Deserialized JSON object containing the payload from Rego</param>
+    /// <returns>Object that will be serialized to JSON and converted to a Rego value</returns>
+    public delegate object RegoCallback(object payload);


Note: An expectation on the Rust side is that an Engine is efficient to clone. It is much faster to add policies to an engine and clone it many times than creating many instances of the engine and adding policies to them individually.

Cloning an engine will also clone the extension. And a clone could be used from another thread. This means that the extension must be stateless or must be safe to be called from multiple threads.

anakrish · 2025-04-10T06:12:09Z

bindings/csharp/Regorus/Regorus.cs


-
+        // Generic callback handler that routes to the appropriate user-provided callback
+        private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)


We could generate closures that forward to RegoCallback delegates and register the closures with Regorus. We might be able to avoid maintaining a dictionary for dispatching.

Suggested change

private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)

private static RegorusCallbackDelegate GenerateRegorusCallback(RegoCallback extension)

{

return delegate(byte* payloadPtr, void* contextPtr

{

...

var result = extension(payloadObject);

...

};

}

private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)

anakrish · 2025-04-10T06:13:04Z

bindings/ffi/src/lib.rs

 }

+// Define callback function type for orchestration to Regorus communication
+pub type RegorusCallbackFn = extern "C" fn(payload: *const c_char, context: *mut c_void) -> *mut c_char;


The context is a nice idea and can be leveraged for C and C++ extensions too.

anakrish · 2025-04-10T06:28:46Z

bindings/ffi/src/lib.rs

 pub extern "C" fn regorus_engine_new() -> *mut RegorusEngine {
-    let engine = ::regorus::Engine::new();
+    let mut engine = ::regorus::Engine::new();
+    let _ = engine.add_extension("invoke".to_string(), 2, Box::new(invoke_callback));


To enable writing function(args) instead of invoke("function", args) in the rego policy, we could change the behavior of the interpreter so that it looks at the table of foreign language extensions if no function exists for a given name.

On the other hand, invoke does make it clear to the reader of the policy that a foreign language function is being used.

I think invoke() would be a cleaner option. Although we can check that it is possible to to trigger invoke from another extension, so that user can optionally use add_extension("function"...) that will invoke("function"...)

anakrish · 2025-04-10T06:30:25Z

@mkulke I think you will find this interesting.

anakrish

Overall looks good to me.

Two comments:

It would be really great if there was a way to use the invoke builtin-extension to implement missing builtins. For example, the JWT functions and the crypto functions would ideally be implemented via invoke to route to regulation compliant libraries in the client language (e.g. C#). This doesn't have to be part of the PR.
It would be great if you can document your usecase (leaving out confidential info) as part of the PR.

anakrish · 2025-05-08T20:07:33Z

bindings/ffi/src/lib.rs


+// Store a callback function and its context
+#[no_mangle]
+pub extern "C" fn regorus_register_callback(


Suggested change

pub extern "C" fn regorus_register_callback(

pub extern "C" fn regorus_register_invoke_callback(

anakrish · 2025-05-08T20:07:56Z

bindings/ffi/src/lib.rs

+
+// Remove a callback function
+#[no_mangle]
+pub extern "C" fn regorus_unregister_callback(name: *const c_char) -> RegorusStatus {


Suggested change

pub extern "C" fn regorus_unregister_callback(name: *const c_char) -> RegorusStatus {

pub extern "C" fn regorus_unregister_invoke_callback(name: *const c_char) -> RegorusStatus {

anakrish · 2025-05-08T20:36:15Z

bindings/csharp/scripts/invoke.csx

+var engine = new Regorus.Engine();
+
+// Enable the invoke extension
+bool enableResult = engine.EnableInvoke();


To catch regressions, I'd prefer a test over a script.

anakrish requested a review from Copilot April 9, 2025 19:30

Copilot AI reviewed Apr 9, 2025

View reviewed changes

anakrish reviewed Apr 10, 2025

View reviewed changes

anakrish requested a review from mkulke April 10, 2025 06:29

kusha added 2 commits April 25, 2025 13:03

Initial draft of Regorus extension to invoke C# code

319265e

Builtin extensions model & thread safety

3a5e251

kusha force-pushed the kusha/csharp-invoke branch from 55318c9 to 3a5e251 Compare May 5, 2025 09:24

kusha marked this pull request as ready for review May 5, 2025 09:26

kusha changed the title ~~Initial draft of Regorus extension to invoke C# code~~ Regorus extension to invoke C# code May 5, 2025

anakrish reviewed May 8, 2025

View reviewed changes

anakrish requested review from anthony-c-martin and cemheren May 8, 2025 20:36

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Regorus extension to invoke C# code#391

Regorus extension to invoke C# code#391
kusha wants to merge 2 commits intomicrosoft:mainfrom
kusha:kusha/csharp-invoke

kusha commented Apr 9, 2025 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

anakrish left a comment

Uh oh!

anakrish Apr 10, 2025

Uh oh!

anakrish Apr 10, 2025

Uh oh!

anakrish Apr 10, 2025

Uh oh!

anakrish Apr 10, 2025

Uh oh!

kusha Apr 10, 2025

Uh oh!

anakrish commented Apr 10, 2025

Uh oh!

anakrish left a comment

Uh oh!

anakrish May 8, 2025

Uh oh!

anakrish May 8, 2025

Uh oh!

anakrish May 8, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants



		// Generic callback handler that routes to the appropriate user-provided callback
		private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)

-        private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)
+       private static RegorusCallbackDelegate GenerateRegorusCallback(RegoCallback extension)
+       {
+           return delegate(byte* payloadPtr, void* contextPtr
+           {
+               ...
+               var result = extension(payloadObject);
+               ...
+           };
+       }
+        private static unsafe byte* CallbackHandler(byte* payloadPtr, void* contextPtr)

	pub extern "C" fn regorus_register_callback(
	pub extern "C" fn regorus_register_invoke_callback(

	pub extern "C" fn regorus_unregister_callback(name: *const c_char) -> RegorusStatus {
	pub extern "C" fn regorus_unregister_invoke_callback(name: *const c_char) -> RegorusStatus {

Conversation

kusha commented Apr 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

anakrish left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anakrish commented Apr 10, 2025

Uh oh!

anakrish left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kusha commented Apr 9, 2025 •

edited

Loading