Cache miniblock metadata

rust/lance-core/src/cache.rs

@@ -6,14 +6,15 @@
 use std::any::{Any, TypeId};
 use std::sync::Arc;
 
-use deepsize::{Context, DeepSizeOf};
 use futures::Future;
 use moka::sync::Cache;
 use object_store::path::Path;
 
 use crate::utils::path::LancePathExt;
 use crate::Result;
 
+pub use deepsize::{Context, DeepSizeOf};
+
 type ArcAny = Arc<dyn Any + Send + Sync>;
 
 #[derive(Clone)]
@@ -121,6 +122,12 @@ impl FileMetadataCache {
         }
     }
 
+    /// Fetch an item from the cache, using a str as the key
+    pub fn get_by_str<T: Send + Sync + 'static>(&self, path: &str) -> Option<Arc<T>> {
+        self.get(&Path::parse(path).unwrap())
+    }
+
+    /// Fetch an item from the cache
     pub fn get<T: Send + Sync + 'static>(&self, path: &Path) -> Option<Arc<T>> {
         let cache = self.cache.as_ref()?;
         let temp: Path;
@@ -135,6 +142,7 @@ impl FileMetadataCache {
             .map(|metadata| metadata.record.clone().downcast::<T>().unwrap())
     }
 
+    /// Insert an item into the cache
     pub fn insert<T: DeepSizeOf + Send + Sync + 'static>(&self, path: Path, metadata: Arc<T>) {
         let Some(cache) = self.cache.as_ref() else {
             return;
@@ -147,6 +155,15 @@ impl FileMetadataCache {
         cache.insert((path, TypeId::of::<T>()), SizedRecord::new(metadata));
     }
 
+    /// Insert an item into the cache, using a str as the key
+    pub fn insert_by_str<T: DeepSizeOf + Send + Sync + 'static>(
+        &self,
+        key: &str,
+        metadata: Arc<T>,
+    ) {
+        self.insert(Path::parse(key).unwrap(), metadata);
+    }
+
     /// Get an item
     ///
     /// If it exists in the cache return that

rust/lance-core/src/datatypes.rs

@@ -29,6 +29,8 @@ pub const COMPRESSION_LEVEL_META_KEY: &str = "lance-encoding:compression-level";
 pub const BLOB_META_KEY: &str = "lance-encoding:blob";
 pub const PACKED_STRUCT_LEGACY_META_KEY: &str = "packed";
 pub const PACKED_STRUCT_META_KEY: &str = "lance-encoding:packed";
+pub const STRUCTURAL_ENCODING_META_KEY: &str = "lance-encoding:structural-encoding";
+pub const STRUCTURAL_ENCODING_MINIBLOCK: &str = "miniblock";
 
 lazy_static::lazy_static! {
     pub static ref BLOB_DESC_FIELDS: Fields =

rust/lance-core/src/utils/path.rs

@@ -11,7 +11,7 @@ impl LancePathExt for Path {
     fn child_path(&self, path: &Path) -> Path {
         let mut new_path = self.clone();
         for part in path.parts() {
-            new_path = path.child(part);
+            new_path = new_path.child(part);
         }
         new_path
     }

rust/lance-datagen/src/generator.rs

@@ -1307,6 +1307,38 @@ impl BatchGeneratorBuilder {
     }
 }
 
+/// Factory for creating a single random array
+pub struct ArrayGeneratorBuilder {
+    generator: Box<dyn ArrayGenerator>,
+    seed: Option<Seed>,
+}
+
+impl ArrayGeneratorBuilder {
+    fn new(generator: Box<dyn ArrayGenerator>) -> Self {
+        Self {
+            generator,
+            seed: None,
+        }
+    }
+
+    /// Use the given seed for the generator
+    pub fn with_seed(mut self, seed: Seed) -> Self {
+        self.seed = Some(seed);
+        self
+    }
+
+    /// Generate a single array with the given length
+    pub fn into_array_rows(
+        mut self,
+        length: RowCount,
+    ) -> Result<Arc<dyn arrow_array::Array>, ArrowError> {
+        let mut rng = rand_xoshiro::Xoshiro256PlusPlus::seed_from_u64(
+            self.seed.map(|s| s.0).unwrap_or(DEFAULT_SEED.0),
+        );
+        self.generator.generate(length, &mut rng)
+    }
+}
+
 const MS_PER_DAY: i64 = 86400000;
 
 pub mod array {
@@ -1858,11 +1890,16 @@ pub mod array {
     }
 }
 
-/// Create a BatchGeneratorBuilder to start generating data
+/// Create a BatchGeneratorBuilder to start generating batch data
 pub fn gen() -> BatchGeneratorBuilder {
     BatchGeneratorBuilder::default()
 }
 
+/// Create an ArrayGeneratorBuilder to start generating array data
+pub fn gen_array(gen: Box<dyn ArrayGenerator>) -> ArrayGeneratorBuilder {
+    ArrayGeneratorBuilder::new(gen)
+}
+
 /// Create a BatchGeneratorBuilder with the given schema
 ///
 /// You can add more columns or convert this into a reader immediately

rust/lance-encoding/src/data.rs

@@ -928,7 +928,7 @@ impl DataBlock {
             Self::Empty() => Self::Empty(),
             Self::Constant(inner) => Self::Constant(inner),
             Self::AllNull(_) => panic!("Cannot remove validity on all-null data"),
-            Self::Nullable(inner) => *inner.data,
+            Self::Nullable(inner) => inner.data.remove_validity(),
             Self::FixedWidth(inner) => Self::FixedWidth(inner),
             Self::FixedSizeList(inner) => Self::FixedSizeList(inner.remove_validity()),
             Self::VariableWidth(inner) => Self::VariableWidth(inner),