Use arrow's string array to place string columns in vineyard

.github/workflows/build-test-graph.yml

@@ -105,12 +105,12 @@ jobs:
           wget https://apache.jfrog.io/artifactory/arrow/$(lsb_release --id --short | tr 'A-Z' 'a-z')/apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
           sudo apt install -y -V ./apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
           sudo apt update
-          sudo apt install -y libarrow-dev=14.0.0-1 \
-                              libarrow-dataset-dev=14.0.0-1 \
-                              libarrow-acero-dev=14.0.0-1 \
-                              libarrow-flight-dev=14.0.0-1 \
-                              libgandiva-dev=14.0.0-1 \
-                              libparquet-dev=14.0.0-1
+          sudo apt install -y libarrow-dev=14.0.1-1 \
+                              libarrow-dataset-dev=14.0.1-1 \
+                              libarrow-acero-dev=14.0.1-1 \
+                              libarrow-flight-dev=14.0.1-1 \
+                              libgandiva-dev=14.0.1-1 \
+                              libparquet-dev=14.0.1-1
 
           # install clang-format
           sudo curl -L https://github.com/muttleyxd/clang-tools-static-binaries/releases/download/master-1d7ec53d/clang-format-11_linux-amd64 --output /usr/bin/clang-format

.github/workflows/build-test.yml

@@ -124,12 +124,12 @@ jobs:
           wget https://apache.jfrog.io/artifactory/arrow/$(lsb_release --id --short | tr 'A-Z' 'a-z')/apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
           sudo apt install -y -V ./apache-arrow-apt-source-latest-$(lsb_release --codename --short).deb
           sudo apt update
-          sudo apt install -y libarrow-dev=14.0.0-1 \
-                              libarrow-dataset-dev=14.0.0-1 \
-                              libarrow-acero-dev=14.0.0-1 \
-                              libarrow-flight-dev=14.0.0-1 \
-                              libgandiva-dev=14.0.0-1 \
-                              libparquet-dev=14.0.0-1
+          sudo apt install -y libarrow-dev=14.0.1-1 \
+                              libarrow-dataset-dev=14.0.1-1 \
+                              libarrow-acero-dev=14.0.1-1 \
+                              libarrow-flight-dev=14.0.1-1 \
+                              libgandiva-dev=14.0.1-1 \
+                              libparquet-dev=14.0.1-1
 
           # install deps for java
           sudo apt install -y default-jdk-headless maven

python/vineyard/data/tensor.py

@@ -19,6 +19,7 @@
 import pickle
 
 import numpy as np
+import pyarrow as pa
 
 from vineyard._C import Object
 from vineyard._C import ObjectID
@@ -54,14 +55,39 @@
 
 def numpy_ndarray_builder(client, value, **kw):
     meta = ObjectMeta()
-    meta['typename'] = 'vineyard::Tensor<%s>' % normalize_cpptype(value.dtype)
-    meta['value_type_'] = value.dtype.name
-    meta['value_type_meta_'] = value.dtype.str
     meta['shape_'] = to_json(value.shape)
     meta['partition_index_'] = to_json(kw.get('partition_index', []))
     meta['nbytes'] = value.nbytes
     meta['order_'] = to_json(('C' if value.flags['C_CONTIGUOUS'] else 'F'))
-    meta.add_member('buffer_', build_numpy_buffer(client, value))
+
+    if value.dtype.name == 'object' or value.dtype.name.startswith('str'):
+        # check if it can be used as a string array
+        try:
+            from vineyard.core.builder import get_current_builders
+            from vineyard.data.arrow import string_array_builder
+
+            # string tensors in numpy like np.array(['a', 'b']) cannot be
+            # converted to pa.large_string_array directly.
+            try:
+                array = pa.array(value, type=pa.large_string())
+            except:  # noqa: E722, pylint: disable=bare-except
+                array = pa.array(value, type=pa.string())
+            meta['typename'] = 'vineyard::Tensor<std::string>'
+            meta['value_type_'] = 'string'
+            meta['value_type_meta_'] = 'str'
+            meta.add_member(
+                'buffer_', string_array_builder(client, array, get_current_builders())
+            )
+        except:  # noqa: E722, pylint: disable=bare-except
+            meta['typename'] = 'vineyard::Tensor<%s>' % normalize_cpptype(value.dtype)
+            meta['value_type_'] = value.dtype.name
+            meta['value_type_meta_'] = value.dtype.str
+            meta.add_member('buffer_', build_numpy_buffer(client, value))
+    else:
+        meta['typename'] = 'vineyard::Tensor<%s>' % normalize_cpptype(value.dtype)
+        meta['value_type_'] = value.dtype.name
+        meta['value_type_meta_'] = value.dtype.str
+        meta.add_member('buffer_', build_numpy_buffer(client, value))
     return client.create_metadata(meta)
 
 

python/vineyard/data/tests/test_dataframe.py

@@ -33,6 +33,13 @@ def test_pandas_dataframe(vineyard_client):
     pd.testing.assert_frame_equal(df, vineyard_client.get(object_id))
 
 
+def test_pandas_dataframe_string(vineyard_client):
+    # see gh#533
+    df = pd.DataFrame({'a': ['1', '2', '3', '4'], 'b': ['5', '6', '7', '8']})
+    object_id = vineyard_client.put(df)
+    pd.testing.assert_frame_equal(df, vineyard_client.get(object_id))
+
+
 def test_pandas_dataframe_complex_columns(vineyard_client):
     # see gh#533
     df = pd.DataFrame([1, 2, 3, 4], columns=[['x']])

rust/vineyard-integration-testing/src/ds/numpy_test.rs

@@ -60,7 +60,6 @@ mod tests {
         return Ok(());
     }
 
-    #[ignore = "ndarray with string type in python side needs to be fixed"]
     #[test]
     fn test_numpy_string() -> Result<()> {
         use arrow_array::array::Array;

rust/vineyard-integration-testing/src/ds/pandas_test.rs

@@ -46,7 +46,6 @@ mod tests {
         return Ok(());
     }
 
-    #[ignore = "ndarray with string type in python side needs to be fixed"]
     #[test]
     fn test_pandas_string() -> Result<()> {
         let ctx = Context::new();