fix: properly encode metadata field values when creating actions

Fixes #3174

Signed-off-by: R. Tyler Croy <[email protected]>

Author: rtyler

python/src/schema.rs

@@ -437,18 +437,18 @@ impl Field {
         inner = inner.with_metadata(metadata.iter().map(|(k, v)| {
             (
                 k,
-                if let serde_json::Value::Number(n) = v {
-                    n.as_i64().map_or_else(
+                match v {
+                    serde_json::Value::Number(n) => n.as_i64().map_or_else(
                         || MetadataValue::String(v.to_string()),
                         |i| {
                             i32::try_from(i)
                                 .ok()
                                 .map(MetadataValue::Number)
                                 .unwrap_or_else(|| MetadataValue::String(v.to_string()))
                         },
-                    )
-                } else {
-                    MetadataValue::String(v.to_string())
+                    ),
+                    serde_json::Value::String(s) => MetadataValue::String(s.to_string()),
+                    other => MetadataValue::String(other.to_string()),
                 },
             )
         }));

python/tests/test_generated_columns.py

@@ -14,7 +14,7 @@ def gc_schema() -> Schema:
             Field(
                 name="gc",
                 type=PrimitiveType("integer"),
-                metadata={"delta.generationExpression": "'5'"},
+                metadata={"delta.generationExpression": "5"},
             ),
         ]
     )
@@ -118,7 +118,7 @@ def test_write_with_invalid_gc_to_table(table_with_gc, invalid_gc_data):
     with pytest.raises(
         DeltaError,
         match=re.escape(
-            "Invariant violations: [\"Check or Invariant (gc = '5' OR (gc IS NULL AND '5' IS NULL)) violated by value in row: [10]\"]"
+            'Invariant violations: ["Check or Invariant (gc = 5 OR (gc IS NULL AND 5 IS NULL)) violated by value in row: [10]"]'
         ),
     ):
         write_deltalake(table_with_gc, mode="append", data=invalid_gc_data)
@@ -177,7 +177,7 @@ def test_raise_when_gc_passed_during_adding_new_columns(tmp_path, data_without_g
                 Field(
                     name="gc",
                     type=PrimitiveType("integer"),
-                    metadata={"delta.generationExpression": "'5'"},
+                    metadata={"delta.generationExpression": "5"},
                 )
             ]
         )
@@ -257,7 +257,7 @@ def test_merge_with_gc_invalid(table_with_gc: DeltaTable, invalid_gc_data):
     with pytest.raises(
         DeltaError,
         match=re.escape(
-            "Invariant violations: [\"Check or Invariant (gc = '5' OR (gc IS NULL AND '5' IS NULL)) violated by value in row: [10]\"]"
+            'Invariant violations: ["Check or Invariant (gc = 5 OR (gc IS NULL AND 5 IS NULL)) violated by value in row: [10]"]'
         ),
     ):
         (

python/tests/test_schema.py

@@ -553,3 +553,12 @@ def test_schema_conversions(schema, expected_schema, conversion_mode):
     result_schema = _convert_pa_schema_to_delta(schema, conversion_mode)
 
     assert result_schema == expected_schema
+
+
+# <https://github.com/delta-io/delta-rs/issues/3174>
+def test_field_serialization():
+    from deltalake import Field
+
+    f = Field("fieldname", "binary", metadata={"key": "value"})
+    assert f.name == "fieldname"
+    assert f.metadata == {"key": "value"}