refactor: new optimized op-layer using serde_v8 (#9843)

- Improves op performance.
- Handle op-metadata (errors, promise IDs) explicitly in the op-layer vs
  per op-encoding (aka: out-of-payload).
- Remove shared queue & custom "asyncHandlers", all async values are
  returned in batches via js_recv_cb.
- The op-layer should be thought of as simple function calls with little
  indirection or translation besides the conceptually straightforward
  serde_v8 bijections.
- Preserve concepts of json/bin/min as semantic groups of their
  inputs/outputs instead of their op-encoding strategy, preserving these
  groups will also facilitate partial transitions over to v8 Fast API for the
  "min" and "bin" groups

1 files changed, 24 insertions, 29 deletions

diff --git a/core/examples/hello_world.rs b/core/examples/hello_world.rs
index c46fc1d98..3b63d2bda 100644
--- a/core/examples/hello_world.rs
+++ b/core/examples/hello_world.rs
@@ -2,11 +2,8 @@
 //!  This example shows you how to define ops in Rust and then call them from
 //!  JavaScript.
 
-use anyhow::anyhow;
 use deno_core::json_op_sync;
 use deno_core::JsRuntime;
-use deno_core::Op;
-use serde_json::Value;
 use std::io::Write;
 
 fn main() {
@@ -21,55 +18,52 @@ fn main() {
   //
   // The second one just transforms some input and returns it to JavaScript.
 
-  // Register the op for outputting bytes to stdout.
+  // Register the op for outputting a string to stdout.
   // It can be invoked with Deno.core.dispatch and the id this method returns
   // or Deno.core.dispatchByName and the name provided.
   runtime.register_op(
     "op_print",
-    // The op_fn callback takes a state object OpState
-    // and a vector of ZeroCopyBuf's, which are mutable references
-    // to ArrayBuffer's in JavaScript.
-    |_state, zero_copy| {
+    // The op_fn callback takes a state object OpState,
+    // a structured arg of type `T` and an optional ZeroCopyBuf,
+    // a mutable reference to a JavaScript ArrayBuffer
+    json_op_sync(|_state, msg: Option<String>, zero_copy| {
       let mut out = std::io::stdout();
 
+      // Write msg to stdout
+      if let Some(msg) = msg {
+        out.write_all(msg.as_bytes()).unwrap();
+      }
+
       // Write the contents of every buffer to stdout
       for buf in zero_copy {
         out.write_all(&buf).unwrap();
       }
 
-      Op::Sync(Box::new([])) // No meaningful result
-    },
+      Ok(()) // No meaningful result
+    }),
   );
 
   // Register the JSON op for summing a number array.
-  // A JSON op is just an op where the first ZeroCopyBuf is a serialized JSON
-  // value, the return value is also a serialized JSON value.  It can be invoked
-  // with Deno.core.jsonOpSync and the name.
   runtime.register_op(
     "op_sum",
     // The json_op_sync function automatically deserializes
     // the first ZeroCopyBuf and serializes the return value
     // to reduce boilerplate
-    json_op_sync(|_state, json: Vec<f64>, zero_copy| {
-      // We check that we only got the JSON value.
-      if !zero_copy.is_empty() {
-        Err(anyhow!("Expected exactly one argument"))
-      } else {
-        // And if we did, do our actual task
-        let sum = json.iter().fold(0.0, |a, v| a + v);
-
-        // Finally we return a JSON value
-        Ok(Value::from(sum))
-      }
+    json_op_sync(|_state, nums: Vec<f64>, _| {
+      // Sum inputs
+      let sum = nums.iter().fold(0.0, |a, v| a + v);
+      // return as a Result<f64, AnyError>
+      Ok(sum)
     }),
   );
 
   // Now we see how to invoke the ops we just defined. The runtime automatically
   // contains a Deno.core object with several functions for interacting with it.
   // You can find its definition in core.js.
-  runtime.execute(
-    "<init>",
-    r#"
+  runtime
+    .execute(
+      "<init>",
+      r#"
 // First we initialize the ops cache.
 // This maps op names to their id's.
 Deno.core.ops();
@@ -78,14 +72,15 @@ Deno.core.ops();
 // our op_print op to display the stringified argument.
 const _newline = new Uint8Array([10]);
 function print(value) {
-  Deno.core.dispatchByName('op_print', Deno.core.encode(value.toString()), _newline);
+  Deno.core.dispatchByName('op_print', 0, value.toString(), _newline);
 }
 
 // Finally we register the error class used by op_sum
 // so that it throws the correct class.
 Deno.core.registerErrorClass('Error', Error);
 "#,
-  ).unwrap();
+    )
+    .unwrap();
 
   // Now we can finally use this in an example.
   runtime