hyperstreamdb 0.5.0__tar.gz → 0.5.2__tar.gz

{hyperstreamdb-0.5.0 → hyperstreamdb-0.5.2}/Cargo.lock

@@ -3392,7 +3392,7 @@ dependencies = [
 
 [[package]]
 name = "hyperstreamdb"
-version = "0.5.0"
+version = "0.5.2"
 dependencies = [
  "ahash",
  "anyhow",

{hyperstreamdb-0.5.0 → hyperstreamdb-0.5.2}/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "hyperstreamdb"
-version = "0.5.0"
+version = "0.5.2"
 edition = "2021"
 license = "MIT AND Apache-2.0"
 description = "HyperStreamDB - Serverless Index-Streaming Database with Overlay Indexing and Vector Search"

{hyperstreamdb-0.5.0 → hyperstreamdb-0.5.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hyperstreamdb
-Version: 0.5.0
+Version: 0.5.2
 Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Developers
 Classifier: Programming Language :: Rust
@@ -383,7 +383,7 @@ session.register("users", table)
 
 # Optional: Enable GPU acceleration for SQL queries
 ctx = hdb.GPUContext.auto_detect()
-hdb.set_global_gpu_context(ctx)
+hdb.set_thread_gpu_context(ctx)
 
 # Simple SQL
 results = table.sql("SELECT * FROM t WHERE id > 100")
@@ -440,7 +440,7 @@ python tests/integration/test_nyc_taxi.py
 
   **Performance Targets:**
   - **Scalar Ingest**: >10K rows/sec ✅
-  - **Vector Ingest (768D)**: >4,000 rows/sec ✅ (April 2026)
+  - **Vector Ingest (768D)**: >22,000 rows/sec ✅ (v0.5.0)
   - **Query (indexed)**: <100ms p99 ⏱️
   - **Vector search**: <50ms for k=10 on 10M vectors ⏱️
   - **Compaction**: <5min for 10GB ⏱️
@@ -456,7 +456,7 @@ python tests/integration/test_nyc_taxi.py
   - **Memory**: 128GB Unified Memory
   - **OS**: macOS (Arm64)
   - **Optimizations**: `target-cpu=native` (NEON SIMD)
-  - **Results (100K vectors, 768D)**:
+  - **Results (100K vectors, 768D) [OUT OF DATE - Pre-v0.5.0]**:
     - **Vector Ingest**: 16,707 rows/sec (CPU) ✅
     - **Vector Search (k=10)**: 819ms (CPU / NEON) ✅
     - **Vector Search (k=10)**: 860ms (MPS GPU) ⏱️
@@ -685,3 +685,5 @@ This project contains modified source code from various upstream open-source pro
 
 **Built with ❤️ in Rust**
 
+
+

{hyperstreamdb-0.5.0 → hyperstreamdb-0.5.2}/README.md

@@ -345,7 +345,7 @@ session.register("users", table)
 
 # Optional: Enable GPU acceleration for SQL queries
 ctx = hdb.GPUContext.auto_detect()
-hdb.set_global_gpu_context(ctx)
+hdb.set_thread_gpu_context(ctx)
 
 # Simple SQL
 results = table.sql("SELECT * FROM t WHERE id > 100")
@@ -402,7 +402,7 @@ python tests/integration/test_nyc_taxi.py
 
   **Performance Targets:**
   - **Scalar Ingest**: >10K rows/sec ✅
-  - **Vector Ingest (768D)**: >4,000 rows/sec ✅ (April 2026)
+  - **Vector Ingest (768D)**: >22,000 rows/sec ✅ (v0.5.0)
   - **Query (indexed)**: <100ms p99 ⏱️
   - **Vector search**: <50ms for k=10 on 10M vectors ⏱️
   - **Compaction**: <5min for 10GB ⏱️
@@ -418,7 +418,7 @@ python tests/integration/test_nyc_taxi.py
   - **Memory**: 128GB Unified Memory
   - **OS**: macOS (Arm64)
   - **Optimizations**: `target-cpu=native` (NEON SIMD)
-  - **Results (100K vectors, 768D)**:
+  - **Results (100K vectors, 768D) [OUT OF DATE - Pre-v0.5.0]**:
     - **Vector Ingest**: 16,707 rows/sec (CPU) ✅
     - **Vector Search (k=10)**: 819ms (CPU / NEON) ✅
     - **Vector Search (k=10)**: 860ms (MPS GPU) ⏱️
@@ -646,3 +646,5 @@ This project contains modified source code from various upstream open-source pro
 ---
 
 **Built with ❤️ in Rust**
+
+

{hyperstreamdb-0.5.0 → hyperstreamdb-0.5.2}/benches/bench_table.rs

@@ -1,8 +1,8 @@
+use arrow::array::{FixedSizeListArray, Float32Array, Float64Array, Int32Array, Int64Array};
+use arrow::datatypes::{DataType, Field, Schema};
+use arrow::record_batch::RecordBatch;
 use criterion::{black_box, criterion_group, criterion_main, Criterion};
 use hyperstreamdb::{Table, VectorSearchParams};
-use arrow::record_batch::RecordBatch;
-use arrow::datatypes::{Schema, Field, DataType};
-use arrow::array::{Int64Array, Float64Array, Int32Array, Float32Array, FixedSizeListArray};
 use std::sync::Arc;
 use tempfile::TempDir;
 use tokio::runtime::Runtime;
@@ -21,12 +21,9 @@ fn create_test_batch(num_rows: usize) -> RecordBatch {
 
     RecordBatch::try_new(
         schema,
-        vec![
-            Arc::new(ids),
-            Arc::new(values),
-            Arc::new(categories),
-        ],
-    ).unwrap()
+        vec![Arc::new(ids), Arc::new(values), Arc::new(categories)],
+    )
+    .unwrap()
 }
 
 /// Utility to create a vector batch for HNSW benchmarks
@@ -34,31 +31,25 @@ fn create_vector_batch(num_rows: usize, dim: usize) -> RecordBatch {
     let inner_field = Arc::new(Field::new("item", DataType::Float32, true));
     let schema = Arc::new(Schema::new(vec![
         Field::new("id", DataType::Int64, false),
-        Field::new("embedding", DataType::FixedSizeList(inner_field.clone(), dim as i32), false),
+        Field::new(
+            "embedding",
+            DataType::FixedSizeList(inner_field.clone(), dim as i32),
+            false,
+        ),
     ]));
 
     let ids = Int64Array::from_iter_values(0..num_rows as i64);
-    
+
     let mut values = Vec::with_capacity(num_rows * dim);
     for i in 0..num_rows * dim {
         values.push(i as f32 / 1000.0);
     }
     let values_array = Float32Array::from(values);
-    
-    let embedding = FixedSizeListArray::try_new(
-        inner_field,
-        dim as i32,
-        Arc::new(values_array),
-        None,
-    ).unwrap();
 
-    RecordBatch::try_new(
-        schema,
-        vec![
-            Arc::new(ids),
-            Arc::new(embedding),
-        ],
-    ).unwrap()
+    let embedding =
+        FixedSizeListArray::try_new(inner_field, dim as i32, Arc::new(values_array), None).unwrap();
+
+    RecordBatch::try_new(schema, vec![Arc::new(ids), Arc::new(embedding)]).unwrap()
 }
 
 /// Benchmark: Ingest throughput (In-memory)
@@ -77,7 +68,7 @@ fn bench_ingest_in_memory(c: &mut Criterion) {
                     let tmp_dir = TempDir::new().unwrap();
                     let table_uri = format!("file://{}", tmp_dir.path().to_str().unwrap());
                     let table = Table::new_async(table_uri).await.unwrap();
-                    
+
                     let start = std::time::Instant::now();
                     for _ in 0..(num_rows / batch_size) {
                         table.write_async(vec![batch.clone()]).await.unwrap();
@@ -96,14 +87,16 @@ fn bench_query_indexed(c: &mut Criterion) {
     let rt = Runtime::new().unwrap();
     let num_rows = 100_000;
     let batch = create_test_batch(num_rows);
-    
+
     // Setup table
     let tmp_dir = TempDir::new().unwrap();
     let table_uri = format!("file://{}", tmp_dir.path().to_str().unwrap());
-    
+
     let table = rt.block_on(async {
         let mut t = Table::new_async(table_uri).await.unwrap();
-        t.add_index_columns_async(vec!["id".to_string()], None).await.unwrap(); 
+        t.add_index_columns_async(vec!["id".to_string()], None)
+            .await
+            .unwrap();
         t.write_async(vec![batch]).await.unwrap();
         t.commit_async().await.unwrap();
         t.wait_for_background_tasks_async().await.unwrap();
@@ -114,7 +107,12 @@ fn bench_query_indexed(c: &mut Criterion) {
         b.to_async(&rt).iter(|| {
             let table = table.clone();
             async move {
-                black_box(table.read_async(Some("id > 50000"), None, None).await.unwrap())
+                black_box(
+                    table
+                        .read_async(Some("id > 50000"), None, None)
+                        .await
+                        .unwrap(),
+                )
             }
         });
     });
@@ -126,14 +124,16 @@ fn bench_vector_search_in_memory(c: &mut Criterion) {
     let num_vectors = 10_000;
     let dim = 128;
     let batch = create_vector_batch(num_vectors, dim);
-    
+
     // Setup table
     let tmp_dir = TempDir::new().unwrap();
     let table_uri = format!("file://{}", tmp_dir.path().to_str().unwrap());
-    
+
     let table = rt.block_on(async {
         let mut t = Table::new_async(table_uri).await.unwrap();
-        t.add_index_columns_async(vec!["embedding".to_string()], None).await.unwrap();
+        t.add_index_columns_async(vec!["embedding".to_string()], None)
+            .await
+            .unwrap();
         t.write_async(vec![batch]).await.unwrap();
         t.commit_async().await.unwrap();
         t.wait_for_background_tasks_async().await.unwrap();
@@ -142,7 +142,9 @@ fn bench_vector_search_in_memory(c: &mut Criterion) {
 
     let vs_params = VectorSearchParams::new(
         "embedding",
-        hyperstreamdb::core::index::VectorValue::Float32((0..dim).map(|i| i as f32 / 100.0).collect()),
+        hyperstreamdb::core::index::VectorValue::Float32(
+            (0..dim).map(|i| i as f32 / 100.0).collect(),
+        ),
         10,
     );
 
@@ -151,9 +153,7 @@ fn bench_vector_search_in_memory(c: &mut Criterion) {
         b.to_async(&rt).iter(|| {
             let vs_params = vs_params.clone();
             let table = table.clone();
-            async move {
-                black_box(table.read_async(None, Some(vs_params), None).await.unwrap())
-            }
+            async move { black_box(table.read_async(None, Some(vs_params), None).await.unwrap()) }
         });
     });
 }
@@ -174,12 +174,12 @@ fn bench_compaction_in_memory(c: &mut Criterion) {
                     let tmp_dir = TempDir::new().unwrap();
                     let table_uri = format!("file://{}", tmp_dir.path().to_str().unwrap());
                     let table = Table::new_async(table_uri).await.unwrap();
-                    
+
                     for _ in 0..num_segments {
                         table.write_async(vec![batch.clone()]).await.unwrap();
                         table.commit_async().await.unwrap();
                     }
-                    
+
                     let start = std::time::Instant::now();
                     table.rewrite_data_files_async(None).await.unwrap();
                     total_duration += start.elapsed();