fix: attempt to workaround problems reading actions as empty rather than null

Cargo.toml

@@ -26,7 +26,7 @@ debug = true
 debug = "line-tables-only"
 
 [workspace.dependencies]
-delta_kernel = { version = "0.4.1", features = ["sync-engine"] }
+delta_kernel = { version = "0.4.1", features = ["default-engine"] }
 #delta_kernel = { path = "../delta-kernel-rs/kernel", features = ["sync-engine"]  }
 
 # arrow

crates/core/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "deltalake-core"
-version = "0.22.2"
+version = "0.22.3"
 authors.workspace = true
 keywords.workspace = true
 readme.workspace = true

crates/core/src/kernel/snapshot/log_data.rs

@@ -245,23 +245,14 @@ impl LogicalFile<'_> {
 
     /// Defines a deletion vector
     pub fn deletion_vector(&self) -> Option<DeletionVectorView<'_>> {
-        if let Some(arr) = self.deletion_vector.as_ref() {
-            // With v0.22 and the upgrade to a more recent arrow. Reading nullable structs with
-            // non-nullable entries back out of parquet is resulting in the DeletionVector having
-            // an empty string rather than a null. The addition check on the value ensures that a
-            // [DeletionVectorView] is not created in this scenario
-            //
-            // <https://github.com/delta-io/delta-rs/issues/3030>
-            if arr.storage_type.is_valid(self.index)
-                && !arr.storage_type.value(self.index).is_empty()
-            {
-                return Some(DeletionVectorView {
+        self.deletion_vector.as_ref().and_then(|arr| {
+            arr.storage_type
+                .is_valid(self.index)
+                .then_some(DeletionVectorView {
                     data: arr,
                     index: self.index,
-                });
-            }
-        }
-        None
+                })
+        })
     }
 
     /// The number of records stored in the data file.
@@ -380,18 +371,23 @@ impl<'a> FileStatsAccessor<'a> {
         );
         let deletion_vector = extract_and_cast_opt::<StructArray>(data, "add.deletionVector");
         let deletion_vector = deletion_vector.and_then(|dv| {
-            let storage_type = extract_and_cast::<StringArray>(dv, "storageType").ok()?;
-            let path_or_inline_dv = extract_and_cast::<StringArray>(dv, "pathOrInlineDv").ok()?;
-            let size_in_bytes = extract_and_cast::<Int32Array>(dv, "sizeInBytes").ok()?;
-            let cardinality = extract_and_cast::<Int64Array>(dv, "cardinality").ok()?;
-            let offset = extract_and_cast_opt::<Int32Array>(dv, "offset");
-            Some(DeletionVector {
-                storage_type,
-                path_or_inline_dv,
-                size_in_bytes,
-                cardinality,
-                offset,
-            })
+            if dv.null_count() == dv.len() {
+                None
+            } else {
+                let storage_type = extract_and_cast::<StringArray>(dv, "storageType").ok()?;
+                let path_or_inline_dv =
+                    extract_and_cast::<StringArray>(dv, "pathOrInlineDv").ok()?;
+                let size_in_bytes = extract_and_cast::<Int32Array>(dv, "sizeInBytes").ok()?;
+                let cardinality = extract_and_cast::<Int64Array>(dv, "cardinality").ok()?;
+                let offset = extract_and_cast_opt::<Int32Array>(dv, "offset");
+                Some(DeletionVector {
+                    storage_type,
+                    path_or_inline_dv,
+                    size_in_bytes,
+                    cardinality,
+                    offset,
+                })
+            }
         });
 
         Ok(Self {

crates/core/src/kernel/snapshot/parse.rs

@@ -78,6 +78,10 @@ pub(super) fn read_adds(array: &dyn ProvidesColumnByName) -> DeltaResult<Vec<Add
     let mut result = Vec::new();
 
     if let Some(arr) = ex::extract_and_cast_opt::<StructArray>(array, "add") {
+        // Stop early if all values are null
+        if arr.null_count() == arr.len() {
+            return Ok(vec![]);
+        }
         let path = ex::extract_and_cast::<StringArray>(arr, "path")?;
         let pvs = ex::extract_and_cast_opt::<MapArray>(arr, "partitionValues");
         let size = ex::extract_and_cast::<Int64Array>(arr, "size")?;
@@ -94,22 +98,33 @@ pub(super) fn read_adds(array: &dyn ProvidesColumnByName) -> DeltaResult<Vec<Add
             let size_in_bytes = ex::extract_and_cast::<Int32Array>(d, "sizeInBytes")?;
             let cardinality = ex::extract_and_cast::<Int64Array>(d, "cardinality")?;
 
-            Box::new(|idx: usize| {
-                if ex::read_str(storage_type, idx).is_ok() {
-                    Some(DeletionVectorDescriptor {
-                        storage_type: std::str::FromStr::from_str(
-                            ex::read_str(storage_type, idx).ok()?,
-                        )
-                        .ok()?,
-                        path_or_inline_dv: ex::read_str(path_or_inline_dv, idx).ok()?.to_string(),
-                        offset: ex::read_primitive_opt(offset, idx),
-                        size_in_bytes: ex::read_primitive(size_in_bytes, idx).ok()?,
-                        cardinality: ex::read_primitive(cardinality, idx).ok()?,
-                    })
-                } else {
-                    None
-                }
-            })
+            // Column might exist but have nullability set for the whole array, so we just return Nones
+            if d.null_count() == d.len() {
+                Box::new(|_| None)
+            } else {
+                Box::new(|idx: usize| {
+                    d.is_valid(idx)
+                        .then(|| {
+                            if ex::read_str(storage_type, idx).is_ok() {
+                                Some(DeletionVectorDescriptor {
+                                    storage_type: std::str::FromStr::from_str(
+                                        ex::read_str(storage_type, idx).ok()?,
+                                    )
+                                    .ok()?,
+                                    path_or_inline_dv: ex::read_str(path_or_inline_dv, idx)
+                                        .ok()?
+                                        .to_string(),
+                                    offset: ex::read_primitive_opt(offset, idx),
+                                    size_in_bytes: ex::read_primitive(size_in_bytes, idx).ok()?,
+                                    cardinality: ex::read_primitive(cardinality, idx).ok()?,
+                                })
+                            } else {
+                                None
+                            }
+                        })
+                        .flatten()
+                })
+            }
         } else {
             Box::new(|_| None)
         };
@@ -210,22 +225,33 @@ pub(super) fn read_removes(array: &dyn ProvidesColumnByName) -> DeltaResult<Vec<
             let size_in_bytes = ex::extract_and_cast::<Int32Array>(d, "sizeInBytes")?;
             let cardinality = ex::extract_and_cast::<Int64Array>(d, "cardinality")?;
 
-            Box::new(|idx: usize| {
-                if ex::read_str(storage_type, idx).is_ok() {
-                    Some(DeletionVectorDescriptor {
-                        storage_type: std::str::FromStr::from_str(
-                            ex::read_str(storage_type, idx).ok()?,
-                        )
-                        .ok()?,
-                        path_or_inline_dv: ex::read_str(path_or_inline_dv, idx).ok()?.to_string(),
-                        offset: ex::read_primitive_opt(offset, idx),
-                        size_in_bytes: ex::read_primitive(size_in_bytes, idx).ok()?,
-                        cardinality: ex::read_primitive(cardinality, idx).ok()?,
-                    })
-                } else {
-                    None
-                }
-            })
+            // Column might exist but have nullability set for the whole array, so we just return Nones
+            if d.null_count() == d.len() {
+                Box::new(|_| None)
+            } else {
+                Box::new(|idx: usize| {
+                    d.is_valid(idx)
+                        .then(|| {
+                            if ex::read_str(storage_type, idx).is_ok() {
+                                Some(DeletionVectorDescriptor {
+                                    storage_type: std::str::FromStr::from_str(
+                                        ex::read_str(storage_type, idx).ok()?,
+                                    )
+                                    .ok()?,
+                                    path_or_inline_dv: ex::read_str(path_or_inline_dv, idx)
+                                        .ok()?
+                                        .to_string(),
+                                    offset: ex::read_primitive_opt(offset, idx),
+                                    size_in_bytes: ex::read_primitive(size_in_bytes, idx).ok()?,
+                                    cardinality: ex::read_primitive(cardinality, idx).ok()?,
+                                })
+                            } else {
+                                None
+                            }
+                        })
+                        .flatten()
+                })
+            }
         } else {
             Box::new(|_| None)
         };

crates/core/src/kernel/snapshot/replay.rs

@@ -20,7 +20,7 @@ use hashbrown::HashSet;
 use itertools::Itertools;
 use percent_encoding::percent_decode_str;
 use pin_project_lite::pin_project;
-use tracing::debug;
+use tracing::log::*;
 
 use super::parse::collect_map;
 use super::ReplayVisitor;
@@ -440,6 +440,14 @@ pub(super) struct DVInfo<'a> {
 fn seen_key(info: &FileInfo<'_>) -> String {
     let path = percent_decode_str(info.path).decode_utf8_lossy();
     if let Some(dv) = &info.dv {
+        // If storage_type is empty then delta-rs has somehow gotten an empty rather than a null
+        // deletion vector, oooof
+        //
+        // See #3030
+        if dv.storage_type.is_empty() {
+            warn!("An empty but not nullable deletionVector was seen for {info:?}");
+            return path.to_string();
+        }
         if let Some(offset) = &dv.offset {
             format!(
                 "{}::{}{}@{offset}",
@@ -551,22 +559,32 @@ fn read_file_info<'a>(arr: &'a dyn ProvidesColumnByName) -> DeltaResult<Vec<Opti
         let path_or_inline_dv = ex::extract_and_cast::<StringArray>(d, "pathOrInlineDv")?;
         let offset = ex::extract_and_cast::<Int32Array>(d, "offset")?;
 
-        Box::new(|idx: usize| {
-            if ex::read_str(storage_type, idx).is_ok() {
-                Ok(Some(DVInfo {
-                    storage_type: ex::read_str(storage_type, idx)?,
-                    path_or_inline_dv: ex::read_str(path_or_inline_dv, idx)?,
-                    offset: ex::read_primitive_opt(offset, idx),
-                }))
-            } else {
-                Ok(None)
-            }
-        })
+        // Column might exist but have nullability set for the whole array, so we just return Nones
+        if d.null_count() == d.len() {
+            Box::new(|_| Ok(None))
+        } else {
+            Box::new(|idx: usize| {
+                if d.is_valid(idx) {
+                    if ex::read_str(storage_type, idx).is_ok() {
+                        Ok(Some(DVInfo {
+                            storage_type: ex::read_str(storage_type, idx)?,
+                            path_or_inline_dv: ex::read_str(path_or_inline_dv, idx)?,
+                            offset: ex::read_primitive_opt(offset, idx),
+                        }))
+                    } else {
+                        Ok(None)
+                    }
+                } else {
+                    Ok(None)
+                }
+            })
+        }
     } else {
         Box::new(|_| Ok(None))
     };
 
     let mut adds = Vec::with_capacity(path.len());
+
     for idx in 0..path.len() {
         let value = path
             .is_valid(idx)
@@ -579,6 +597,7 @@ fn read_file_info<'a>(arr: &'a dyn ProvidesColumnByName) -> DeltaResult<Vec<Opti
             .transpose()?;
         adds.push(value);
     }
+
     Ok(adds)
 }
 

crates/core/src/protocol/checkpoints.rs

@@ -1163,10 +1163,16 @@ mod tests {
     }
 
     /// <https://github.com/delta-io/delta-rs/issues/3030>
+    #[cfg(feature = "datafusion")]
     #[tokio::test]
     async fn test_create_checkpoint_overwrite() -> DeltaResult<()> {
         use crate::protocol::SaveMode;
+        use crate::writer::test_utils::datafusion::get_data_sorted;
         use crate::writer::test_utils::get_arrow_schema;
+        use datafusion::assert_batches_sorted_eq;
+
+        let tmp_dir = tempfile::tempdir().unwrap();
+        let tmp_path = std::fs::canonicalize(tmp_dir.path()).unwrap();
 
         let batch = RecordBatch::try_new(
             Arc::clone(&get_arrow_schema(&None)),
@@ -1177,13 +1183,15 @@ mod tests {
             ],
         )
         .unwrap();
-        let table = DeltaOps::try_from_uri_with_storage_options("memory://", HashMap::default())
+
+        let mut table = DeltaOps::try_from_uri(tmp_path.as_os_str().to_str().unwrap())
             .await?
             .write(vec![batch])
             .await?;
+        table.load().await?;
         assert_eq!(table.version(), 0);
 
-        create_checkpoint_for(0, table.snapshot().unwrap(), table.log_store.as_ref()).await?;
+        create_checkpoint(&table).await?;
 
         let batch = RecordBatch::try_new(
             Arc::clone(&get_arrow_schema(&None)),
@@ -1194,11 +1202,23 @@ mod tests {
             ],
         )
         .unwrap();
-        let table = DeltaOps(table)
+
+        let table = DeltaOps::try_from_uri(tmp_path.as_os_str().to_str().unwrap())
+            .await?
             .write(vec![batch])
             .with_save_mode(SaveMode::Overwrite)
             .await?;
         assert_eq!(table.version(), 1);
+
+        let expected = [
+            "+----+-------+------------+",
+            "| id | value | modified   |",
+            "+----+-------+------------+",
+            "| A  | 0     | 2021-02-02 |",
+            "+----+-------+------------+",
+        ];
+        let actual = get_data_sorted(&table, "id,value,modified").await;
+        assert_batches_sorted_eq!(&expected, &actual);
         Ok(())
     }
 }

crates/deltalake/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "deltalake"
-version = "0.22.2"
+version = "0.22.3"
 authors.workspace = true
 keywords.workspace = true
 readme.workspace = true

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "deltalake-python"
-version = "0.22.2"
+version = "0.22.3"
 authors = ["Qingping Hou <[email protected]>", "Will Jones <[email protected]>"]
 homepage = "https://github.com/delta-io/delta-rs"
 license = "Apache-2.0"

python/tests/test_checkpoint.py

@@ -483,18 +483,19 @@ def test_checkpoint_with_multiple_writes(tmp_path: pathlib.Path):
             }
         ),
     )
-    DeltaTable(tmp_path).create_checkpoint()
+    dt = DeltaTable(tmp_path)
+    dt.create_checkpoint()
+    assert dt.version() == 0
+    df = pd.DataFrame(
+        {
+            "a": ["a"],
+            "b": [100],
+        }
+    )
+    write_deltalake(tmp_path, df, mode="overwrite")
 
     dt = DeltaTable(tmp_path)
+    assert dt.version() == 1
+    new_df = dt.to_pandas()
     print(dt.to_pandas())
-
-    write_deltalake(
-        tmp_path,
-        pd.DataFrame(
-            {
-                "a": ["a"],
-                "b": [100],
-            }
-        ),
-        mode="overwrite",
-    )
+    assert len(new_df) == 1, "We overwrote! there should only be one row"