Expression VTable (#5191)

Experiment with new style of vtable. 

Overview:
* No more vtable macro.
* "Encoding" struct is now combined with the vtable struct (we had an
empty vtable struct before for no reason).
* Instead of Arc<dyn Expr>, we have an Expression struct with vtable,
opaque instance data, and children.
  * We may want this struct to be Arc'd also?
* Because private data can be anything, simple expressions do not need a
separate instance struct (e.g. Binary simply stores Operator enum as
private data)
* We retain benefits of vtables in that we can run pre/post logic around
typed impl functions. We can also expose a smaller API surface on
Expression than we do on VTable.

---------

Signed-off-by: Nicholas Gates <[email protected]>

Author: gatesn

Cargo.lock

@@ -42,7 +42,7 @@ fuzz_target!(|fuzz: FuzzFileAction| -> Corpus {
         let filtered = filter(&array_data, &mask).vortex_unwrap();
         projection_expr
             .clone()
-            .unwrap_or_else(|| root())
+            .unwrap_or_else(root)
             .evaluate(&Scope::new(filtered))
             .vortex_unwrap()
     };
@@ -69,7 +69,7 @@ fuzz_target!(|fuzz: FuzzFileAction| -> Corpus {
         .vortex_unwrap()
         .scan()
         .vortex_unwrap()
-        .with_projection(projection_expr.unwrap_or_else(|| root()))
+        .with_projection(projection_expr.unwrap_or_else(root))
         .with_some_filter(filter_expr)
         .into_array_iter(&*RUNTIME)
         .vortex_unwrap()

fuzz/fuzz_targets/file_io.rs

@@ -4,16 +4,16 @@
 use libfuzzer_sys::arbitrary::{Arbitrary, Unstructured};
 use vortex_array::ArrayRef;
 use vortex_array::arrays::arbitrary::ArbitraryArray;
-use vortex_expr::ExprRef;
+use vortex_expr::Expression;
 use vortex_expr::arbitrary::{filter_expr, projection_expr};
 
 use crate::array::CompressorStrategy;
 
 #[derive(Debug)]
 pub struct FuzzFileAction {
     pub array: ArrayRef,
-    pub projection_expr: Option<ExprRef>,
-    pub filter_expr: Option<ExprRef>,
+    pub projection_expr: Option<Expression>,
+    pub filter_expr: Option<Expression>,
     pub compressor_strategy: CompressorStrategy,
 }
 

fuzz/src/file/mod.rs

@@ -172,7 +172,7 @@ public static Binary ltEq(Expression left, Expression right) {
 
     @Override
     public String id() {
-        return "binary";
+        return "vortex.binary";
     }
 
     @Override

java/testfiles/Cargo.lock

@@ -78,7 +78,7 @@ public String getPath() {
 
     @Override
     public String id() {
-        return "get_item";
+        return "vortex.get_item";
     }
 
     @Override

java/vortex-jni/src/main/java/dev/vortex/api/expressions/Binary.java

@@ -72,7 +72,7 @@ public T getValue() {
 
     @Override
     public String id() {
-        return "literal";
+        return "vortex.literal";
     }
 
     @Override

java/vortex-jni/src/main/java/dev/vortex/api/expressions/GetItem.java

@@ -63,7 +63,7 @@ public int hashCode() {
 
     @Override
     public String id() {
-        return "not";
+        return "vortex.not";
     }
 
     @Override

java/vortex-jni/src/main/java/dev/vortex/api/expressions/Literal.java

@@ -39,7 +39,7 @@ public static Root parse(byte[] _metadata, List<Expression> children) {
 
     @Override
     public String id() {
-        return "root";
+        return "vortex.root";
     }
 
     @Override

java/vortex-jni/src/main/java/dev/vortex/api/expressions/Not.java

@@ -46,15 +46,15 @@ public static Expression deserialize(ExprProtos.Expr expr) {
                 expr.getChildrenList().stream().map(Expressions::deserialize).collect(Collectors.toList());
 
         switch (expr.getId()) {
-            case "binary":
+            case "vortex.binary":
                 return Binary.parse(metadata, children);
-            case "get_item":
+            case "vortex.get_item":
                 return GetItem.parse(metadata, children);
-            case "root":
+            case "vortex.root":
                 return Root.parse(metadata, children);
-            case "literal":
+            case "vortex.literal":
                 return Literal.parse(metadata, children);
-            case "not":
+            case "vortex.not":
                 return Not.parse(metadata, children);
             default:
                 return new Unknown(expr.getId(), children, expr.getMetadata().toByteArray());