Read from Arrow C Data interface

vegafusion-common/src/data/ffi.rs

@@ -0,0 +1,62 @@
+use arrow::array::{RecordBatch, RecordBatchReader};
+use arrow::datatypes::SchemaRef;
+use arrow::ffi_stream::ArrowArrayStreamReader;
+use arrow::ffi_stream::FFI_ArrowArrayStream;
+use pyo3::exceptions::{PyTypeError, PyValueError};
+use pyo3::prelude::*;
+use pyo3::types::PyCapsule;
+
+/// Validate PyCapsule has provided name
+fn validate_pycapsule_name(capsule: &PyCapsule, expected_name: &str) -> PyResult<()> {
+    let capsule_name = capsule.name()?;
+    if let Some(capsule_name) = capsule_name {
+        let capsule_name = capsule_name.to_str()?;
+        if capsule_name != expected_name {
+            return Err(PyValueError::new_err(format!(
+                "Expected name '{}' in PyCapsule, instead got '{}'",
+                expected_name, capsule_name
+            )));
+        }
+    } else {
+        return Err(PyValueError::new_err(
+            "Expected schema PyCapsule to have name set.",
+        ));
+    }
+
+    Ok(())
+}
+
+/// Import `__arrow_c_stream__` across Python boundary.
+fn call_arrow_c_stream(ob: &'_ PyAny) -> PyResult<&'_ PyCapsule> {
+    if !ob.hasattr("__arrow_c_stream__")? {
+        return Err(PyValueError::new_err(
+            "Expected an object with dunder __arrow_c_stream__",
+        ));
+    }
+
+    let capsule = ob.getattr("__arrow_c_stream__")?.call0()?.downcast()?;
+    Ok(capsule)
+}
+
+fn import_stream_pycapsule(capsule: &PyCapsule) -> PyResult<FFI_ArrowArrayStream> {
+    validate_pycapsule_name(capsule, "arrow_array_stream")?;
+
+    let stream = unsafe { FFI_ArrowArrayStream::from_raw(capsule.pointer() as _) };
+    Ok(stream)
+}
+
+pub(crate) fn import_arrow_c_stream(ob: &'_ PyAny) -> PyResult<(Vec<RecordBatch>, SchemaRef)> {
+    let capsule = call_arrow_c_stream(ob)?;
+    let stream = import_stream_pycapsule(capsule)?;
+    let stream_reader = ArrowArrayStreamReader::try_new(stream)
+        .map_err(|err| PyValueError::new_err(err.to_string()))?;
+    let schema = stream_reader.schema();
+
+    let mut batches = vec![];
+    for batch in stream_reader {
+        let batch = batch.map_err(|err| PyTypeError::new_err(err.to_string()))?;
+        batches.push(batch);
+    }
+
+    Ok((batches, schema))
+}

vegafusion-common/src/data/mod.rs

@@ -1,5 +1,8 @@
 pub mod table;
 
+#[cfg(feature = "pyarrow")]
+mod ffi;
+
 #[cfg(feature = "json")]
 pub mod json_writer;
 pub mod scalar;

vegafusion-common/src/data/table.rs

@@ -8,6 +8,8 @@ use arrow::{
     record_batch::RecordBatch,
 };
 
+#[cfg(feature = "pyarrow")]
+use crate::data::ffi::import_arrow_c_stream;
 use crate::{
     data::{ORDER_COL, ORDER_COL_DTYPE},
     error::{Result, ResultWithContext, VegaFusionError},
@@ -36,7 +38,7 @@ use {
     pyo3::{
         prelude::PyModule,
         types::{PyList, PyTuple},
-        PyAny, PyErr, PyObject, Python,
+        PyAny, PyErr, PyObject, PyResult, Python,
     },
 };
 
@@ -267,6 +269,12 @@ impl VegaFusionTable {
         }
     }
 
+    #[cfg(feature = "pyarrow")]
+    pub fn from_arrow_c_stream(ob: &PyAny) -> PyResult<Self> {
+        let (batches, schema) = import_arrow_c_stream(ob)?;
+        Ok(VegaFusionTable::try_new(schema, batches)?)
+    }
+
     #[cfg(feature = "pyarrow")]
     pub fn from_pyarrow(py: Python, pyarrow_table: &PyAny) -> std::result::Result<Self, PyErr> {
         // Extract table.schema as a Rust Schema

vegafusion-python-embed/src/lib.rs

@@ -186,6 +186,10 @@ impl PyVegaFusionRuntime {
                                     .scan_py_datasource(inline_dataset.to_object(py)),
                             )?;
                             VegaFusionDataset::DataFrame(df)
+                        } else if inline_dataset.hasattr("__arrow_c_stream__")? {
+                            // Import via Arrow PyCapsule Interface
+                            let table = VegaFusionTable::from_arrow_c_stream(inline_dataset)?;
+                            VegaFusionDataset::from_table_ipc_bytes(&table.to_ipc_bytes()?)?
                         } else {
                             // Assume PyArrow Table
                             // We convert to ipc bytes for two reasons: