perf: intern identifiers (array, layout, etc). (#7412)

We use id for scalar_fns, array and layouts. This we used to match rules
among other things.

This PR replace Arc<str> ID with interned str (u32) id.

---------

Signed-off-by: Nicholas Gates <nick@nickgates.com>
Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Author: joseph-isaacs

Cargo.lock

@@ -163,6 +163,7 @@ itertools = "0.14.0"
 jetscii = "0.5.3"
 jiff = "0.2.0"
 kanal = "0.1.1"
+lasso = { version = "0.7", features = ["multi-threaded"] }
 lending-iterator = "0.1.7"
 libfuzzer-sys = "0.4"
 libloading = "0.8"

Cargo.toml

@@ -6,10 +6,6 @@ pub struct vortex_alp::ALP
 
 impl vortex_alp::ALP
 
-pub const vortex_alp::ALP::ID: vortex_array::array::ArrayId
-
-impl vortex_alp::ALP
-
 pub fn vortex_alp::ALP::new(encoded: vortex_array::array::erased::ArrayRef, exponents: vortex_alp::Exponents, patches: core::option::Option<vortex_array::patches::Patches>) -> vortex_alp::ALPArray
 
 pub unsafe fn vortex_alp::ALP::new_unchecked(encoded: vortex_array::array::erased::ArrayRef, exponents: vortex_alp::Exponents, patches: core::option::Option<vortex_array::patches::Patches>) -> vortex_alp::ALPArray
@@ -148,8 +144,6 @@ pub struct vortex_alp::ALPRD
 
 impl vortex_alp::ALPRD
 
-pub const vortex_alp::ALPRD::ID: vortex_array::array::ArrayId
-
 pub unsafe fn vortex_alp::ALPRD::new_unchecked(dtype: vortex_array::dtype::DType, left_parts: vortex_array::array::erased::ArrayRef, left_parts_dictionary: vortex_buffer::buffer::Buffer<u16>, right_parts: vortex_array::array::erased::ArrayRef, right_bit_width: u8, left_parts_patches: core::option::Option<vortex_array::patches::Patches>) -> vortex_alp::ALPRDArray
 
 pub fn vortex_alp::ALPRD::try_new(dtype: vortex_array::dtype::DType, left_parts: vortex_array::array::erased::ArrayRef, left_parts_dictionary: vortex_buffer::buffer::Buffer<u16>, right_parts: vortex_array::array::erased::ArrayRef, right_bit_width: u8, left_parts_patches: core::option::Option<vortex_array::patches::Patches>) -> vortex_error::VortexResult<vortex_alp::ALPRDArray>

encodings/alp/public-api.lock

@@ -39,6 +39,7 @@ use vortex_error::vortex_bail;
 use vortex_error::vortex_ensure;
 use vortex_error::vortex_panic;
 use vortex_session::VortexSession;
+use vortex_session::registry::CachedId;
 
 use crate::ALPFloat;
 use crate::alp::Exponents;
@@ -72,7 +73,8 @@ impl VTable for ALP {
     type ValidityVTable = ValidityVTableFromChild;
 
     fn id(&self) -> ArrayId {
-        Self::ID
+        static ID: CachedId = CachedId::new("vortex.alp");
+        *ID
     }
 
     fn validate(
@@ -237,10 +239,6 @@ impl Display for ALPData {
 #[derive(Clone, Debug)]
 pub struct ALP;
 
-impl ALP {
-    pub const ID: ArrayId = ArrayId::new_ref("vortex.alp");
-}
-
 #[derive(Clone, prost::Message)]
 pub struct ALPMetadata {
     #[prost(uint32, tag = "1")]

encodings/alp/src/alp/array.rs

@@ -6,9 +6,11 @@
 //!
 //! This enables zero-cost backward compatibility with previously written datasets.
 
+use vortex_array::Array;
 use vortex_array::ArrayId;
 use vortex_array::ArrayPlugin;
 use vortex_array::ArrayRef;
+use vortex_array::ArrayVTable;
 use vortex_array::IntoArray;
 use vortex_array::VortexSessionExecute;
 use vortex_array::arrays::Patched;
@@ -32,7 +34,8 @@ impl ArrayPlugin for ALPPatchedPlugin {
     fn id(&self) -> ArrayId {
         // We reuse the existing `ALP` ID so that we can take over its
         // deserialization pathway.
-        ALP::ID
+        // TODO(joe): dedup method name
+        ArrayVTable::id(&ALP)
     }
 
     fn serialize(
@@ -53,10 +56,10 @@ impl ArrayPlugin for ALPPatchedPlugin {
         children: &dyn ArrayChildren,
         session: &VortexSession,
     ) -> VortexResult<ArrayRef> {
-        let alp_array = ALP
-            .deserialize(dtype, len, metadata, buffers, children, session)?
-            .try_downcast::<ALP>()
-            .map_err(|_| vortex_err!("ALP plugin should only deserialize vortex.alp"))?;
+        let alp_array = Array::<ALP>::try_from_parts(ArrayVTable::deserialize(
+            &ALP, dtype, len, metadata, buffers, children, session,
+        )?)
+        .map_err(|_| vortex_err!("ALP plugin should only deserialize vortex.alp"))?;
 
         // Check if there are interior patches to externalize.
         let Some(patches) = alp_array.patches() else {
@@ -78,7 +81,8 @@ impl ArrayPlugin for ALPPatchedPlugin {
     }
 
     fn is_supported_encoding(&self, id: &ArrayId) -> bool {
-        id == &Patched.id() || id == &ALP.id()
+        // TODO(joe): dedup method name
+        id == ArrayVTable::id(&Patched) || id == ArrayVTable::id(&ALP)
     }
 }
 

encodings/alp/src/alp/plugin.rs

@@ -44,6 +44,7 @@ use vortex_error::vortex_ensure;
 use vortex_error::vortex_err;
 use vortex_error::vortex_panic;
 use vortex_session::VortexSession;
+use vortex_session::registry::CachedId;
 
 use crate::alp_rd::kernel::PARENT_KERNELS;
 use crate::alp_rd::rules::RULES;
@@ -92,7 +93,8 @@ impl VTable for ALPRD {
     type ValidityVTable = ValidityVTableFromChild;
 
     fn id(&self) -> ArrayId {
-        Self::ID
+        static ID: CachedId = CachedId::new("vortex.alprd");
+        *ID
     }
 
     fn validate(
@@ -357,8 +359,6 @@ pub struct ALPRDDataParts {
 pub struct ALPRD;
 
 impl ALPRD {
-    pub const ID: ArrayId = ArrayId::new_ref("vortex.alprd");
-
     pub fn try_new(
         dtype: DType,
         left_parts: ArrayRef,

encodings/alp/src/alp_rd/array.rs

@@ -18,6 +18,7 @@
 
 pub use alp::*;
 pub use alp_rd::*;
+use vortex_array::ArrayVTable;
 use vortex_array::aggregate_fn::AggregateFnVTable;
 use vortex_array::aggregate_fn::fns::nan_count::NanCount;
 use vortex_array::aggregate_fn::session::AggregateFnSessionExt;
@@ -41,7 +42,7 @@ pub fn initialize(session: &VortexSession) {
 
     // Register the ALP-specific NaN count aggregate kernel.
     session.aggregate_fns().register_aggregate_kernel(
-        ALP::ID,
+        ALP.id(),
         Some(NanCount.id()),
         &compute::nan_count::ALPNanCountKernel,
     );

encodings/alp/src/lib.rs

@@ -4,8 +4,6 @@ pub struct vortex_bytebool::ByteBool
 
 impl vortex_bytebool::ByteBool
 
-pub const vortex_bytebool::ByteBool::ID: vortex_array::array::ArrayId
-
 pub fn vortex_bytebool::ByteBool::from_option_vec(data: alloc::vec::Vec<core::option::Option<bool>>) -> vortex_bytebool::ByteBoolArray
 
 pub fn vortex_bytebool::ByteBool::from_vec<V: core::convert::Into<vortex_array::validity::Validity>>(data: alloc::vec::Vec<bool>, validity: V) -> vortex_bytebool::ByteBoolArray

encodings/bytebool/public-api.lock

@@ -37,6 +37,7 @@ use vortex_error::vortex_ensure;
 use vortex_error::vortex_panic;
 use vortex_mask::Mask;
 use vortex_session::VortexSession;
+use vortex_session::registry::CachedId;
 
 use crate::kernel::PARENT_KERNELS;
 
@@ -62,7 +63,8 @@ impl VTable for ByteBool {
     type ValidityVTable = Self;
 
     fn id(&self) -> ArrayId {
-        Self::ID
+        static ID: CachedId = CachedId::new("vortex.bytebool");
+        *ID
     }
 
     fn validate(
@@ -200,8 +202,6 @@ impl<T: TypedArrayRef<ByteBool>> ByteBoolArrayExt for T {}
 pub struct ByteBool;
 
 impl ByteBool {
-    pub const ID: ArrayId = ArrayId::new_ref("vortex.bytebool");
-
     pub fn new(buffer: BufferHandle, validity: Validity) -> ByteBoolArray {
         let dtype = DType::Bool(validity.nullability());
         let slots = ByteBoolData::make_slots(&validity, buffer.len());

encodings/bytebool/src/array.rs

@@ -4,8 +4,6 @@ pub struct vortex_datetime_parts::DateTimeParts
 
 impl vortex_datetime_parts::DateTimeParts
 
-pub const vortex_datetime_parts::DateTimeParts::ID: vortex_array::array::ArrayId
-
 pub fn vortex_datetime_parts::DateTimeParts::try_from_temporal(temporal: vortex_array::arrays::datetime::TemporalArray) -> vortex_error::VortexResult<vortex_datetime_parts::DateTimePartsArray>
 
 pub fn vortex_datetime_parts::DateTimeParts::try_new(dtype: vortex_array::dtype::DType, days: vortex_array::array::erased::ArrayRef, seconds: vortex_array::array::erased::ArrayRef, subseconds: vortex_array::array::erased::ArrayRef) -> vortex_error::VortexResult<vortex_datetime_parts::DateTimePartsArray>