Improved handling of empty cells in CSV YQ-2727

ydb/library/yql/providers/s3/actors/yql_s3_read_actor.cpp

@@ -3103,6 +3103,10 @@ std::pair<NYql::NDq::IDqComputeActorAsyncInput*, IActor*> CreateS3ReadActor(
 
         readSpec->Format = params.GetFormat();
 
+        if (readSpec->Format == "csv_with_names") {
+            readSpec->Settings.csv.empty_as_default = true;
+        }
+
         if (const auto it = settings.find("compression"); settings.cend() != it)
             readSpec->Compression = it->second;
 

ydb/library/yql/udfs/common/clickhouse/client/src/Processors/Formats/Impl/CSVRowInputFormat.cpp

@@ -429,6 +429,9 @@ bool CSVRowInputFormat::readField(IColumn & column, const DataTypePtr & type, co
     if (format_settings.csv.empty_as_default
         && (at_delimiter || at_last_column_line_end))
     {
+        if (!type->isNullable() && type->getName() != "String" && !type->getName().starts_with("FixedString")) {
+            throw ParsingException("Invalid data format", NDB::ErrorCodes::INCORRECT_DATA);
+        }
         /// Treat empty unquoted column value as default value, if
         /// specified in the settings. Tuple columns might seem
         /// problematic, because they are never quoted but still contain

ydb/tests/fq/s3/test_format_setting.py

@@ -15,6 +15,7 @@
 
 import ydb.tests.fq.s3.s3_helpers as s3_helpers
 from ydb.tests.tools.fq_runner.kikimr_utils import yq_all
+from google.protobuf import struct_pb2
 
 
 class TestS3(TestYdsBase):
@@ -753,3 +754,122 @@ def test_date_time_completeness_iso(self, kikimr, s3, client, filename, type_for
         assert result_set.columns[3].type.type_id == ydb.Type.INT32
 
         assert len(result_set.rows) == 6
+
+    @yq_all
+    @pytest.mark.parametrize("filename", [
+        ("date_null/as_default/test.csv"),
+        ("date_null/parse_error/test.csv")
+    ])
+    def test_date_null(self, kikimr, s3, client, filename):
+        self.create_bucket_and_upload_file(filename, s3, kikimr)
+        client.create_storage_connection("hcpp", "fbucket")
+
+        sql = '''
+            SELECT
+                `put`
+            FROM
+                `hcpp`.`{name}`
+            WITH (FORMAT="csv_with_names",
+                csv_delimiter=",",
+                SCHEMA=(
+                `put` Date
+                ))
+            LIMIT 10;
+            '''.format(name="/" + filename)
+
+        query_id = client.create_query("simple", sql, type=fq.QueryContent.QueryType.ANALYTICS).result.query_id
+        client.wait_query_status(query_id, fq.QueryMeta.COMPLETED)
+        data = client.get_result_data(query_id, limit=50)
+        assert data.result.result_set.rows[0].items[0].null_flag_value == struct_pb2.NULL_VALUE, str(data.result.result_set)
+
+    @yq_all
+    @pytest.mark.parametrize("filename", [
+        ("date_null/as_default/test.csv"),
+        ("date_null/parse_error/test.csv")
+    ])
+    def test_date_null_with_not_null_type(self, kikimr, s3, client, filename):
+        self.create_bucket_and_upload_file(filename, s3, kikimr)
+        client.create_storage_connection("hcpp", "fbucket")
+
+        sql = '''
+            SELECT
+                `put`
+            FROM
+                `hcpp`.`{name}`
+            WITH (FORMAT="csv_with_names",
+                csv_delimiter=",",
+                SCHEMA=(
+                `put` Date NOT NULL
+                ))
+            LIMIT 10;
+            '''.format(name="/" + filename)
+
+        query_id = client.create_query("simple", sql, type=fq.QueryContent.QueryType.ANALYTICS).result.query_id
+        client.wait_query_status(query_id, fq.QueryMeta.FAILED)
+        describe_result = client.describe_query(query_id).result
+        issues = describe_result.query.issue[0].issues
+        assert "Invalid data format" in str(issues), str(describe_result)
+        assert "name: put, type: Date, ERROR: text " in str(issues), str(describe_result)
+        assert "is not like Date" in str(issues), str(describe_result)
+
+    @yq_all
+    @pytest.mark.parametrize("filename", [
+        ("date_null/as_default/multi_null.csv"),
+        ("date_null/parse_error/multi_null.csv")
+    ])
+    def test_date_null_multi(self, kikimr, s3, client, filename):
+        self.create_bucket_and_upload_file(filename, s3, kikimr)
+        client.create_storage_connection("hcpp", "fbucket")
+
+        sql = '''
+            SELECT
+                `put`, `a`, `t`
+            FROM
+                `hcpp`.`{name}`
+            WITH (FORMAT="csv_with_names",
+                csv_delimiter=",",
+                SCHEMA=(
+                `put` Date,
+                `a` Date,
+                `t` Date
+                ))
+            LIMIT 10;
+            '''.format(name="/" + filename)
+
+        query_id = client.create_query("simple", sql, type=fq.QueryContent.QueryType.ANALYTICS).result.query_id
+        client.wait_query_status(query_id, fq.QueryMeta.COMPLETED)
+        data = client.get_result_data(query_id, limit=50)
+        assert data.result.result_set.rows[0].items[0].null_flag_value == struct_pb2.NULL_VALUE, str(data.result.result_set)
+        assert data.result.result_set.rows[0].items[1].null_flag_value == struct_pb2.NULL_VALUE, str(data.result.result_set)
+        assert data.result.result_set.rows[0].items[2].null_flag_value == struct_pb2.NULL_VALUE, str(data.result.result_set)
+
+    @yq_all
+    @pytest.mark.parametrize("filename", [
+        ("date_null/as_default/multi_null.csv"),
+        ("date_null/parse_error/multi_null.csv")
+    ])
+    def test_string_not_null_multi(self, kikimr, s3, client, filename):
+        self.create_bucket_and_upload_file(filename, s3, kikimr)
+        client.create_storage_connection("hcpp", "fbucket")
+
+        sql = '''
+            SELECT
+                `put`, `a`, `t`
+            FROM
+                `hcpp`.`{name}`
+            WITH (FORMAT="csv_with_names",
+                csv_delimiter=",",
+                SCHEMA=(
+                `put` String NOT NULL,
+                `a` Utf8 NOT NULL,
+                `t` String NOT NULL
+                ))
+            LIMIT 10;
+            '''.format(name="/" + filename)
+
+        query_id = client.create_query("simple", sql, type=fq.QueryContent.QueryType.ANALYTICS).result.query_id
+        client.wait_query_status(query_id, fq.QueryMeta.COMPLETED)
+        data = client.get_result_data(query_id, limit=50)
+        assert data.result.result_set.rows[0].items[0].bytes_value == b"", str(data.result.result_set)
+        assert data.result.result_set.rows[0].items[1].bytes_value == b"", str(data.result.result_set)
+        assert data.result.result_set.rows[0].items[2].bytes_value == b"", str(data.result.result_set)

ydb/tests/fq/s3/test_format_settings/date_null/as_default/multi_null.csv

@@ -0,0 +1,2 @@
+id,put,call,a,t
+id1,,2022-12-22,,

ydb/tests/fq/s3/test_format_settings/date_null/as_default/test.csv

@@ -0,0 +1,2 @@
+id,put,call
+id1,,2022-12-22

ydb/tests/fq/s3/test_format_settings/date_null/parse_error/multi_null.csv

@@ -0,0 +1,2 @@
+id,put,call,a,t
+id1,,hello2,,

ydb/tests/fq/s3/test_format_settings/date_null/parse_error/test.csv

@@ -0,0 +1,2 @@
+id,put,call
+id1,,hello2