Merge pull request #66782 from a-a-f/input_format_json_empty_as_default

Add support for the `input_format_json_empty_as_default` setting

docs/en/interfaces/formats.md

@@ -1396,6 +1396,7 @@ SELECT * FROM json_each_row_nested
 - [input_format_json_ignore_unknown_keys_in_named_tuple](/docs/en/operations/settings/settings-formats.md/#input_format_json_ignore_unknown_keys_in_named_tuple) - ignore unknown keys in json object for named tuples. Default value - `false`.
 - [input_format_json_compact_allow_variable_number_of_columns](/docs/en/operations/settings/settings-formats.md/#input_format_json_compact_allow_variable_number_of_columns) - allow variable number of columns in JSONCompact/JSONCompactEachRow format, ignore extra columns and use default values on missing columns. Default value - `false`.
 - [input_format_json_throw_on_bad_escape_sequence](/docs/en/operations/settings/settings-formats.md/#input_format_json_throw_on_bad_escape_sequence) - throw an exception if JSON string contains bad escape sequence. If disabled, bad escape sequences will remain as is in the data. Default value - `true`.
+- [input_format_json_empty_as_default](/docs/en/operations/settings/settings-formats.md/#input_format_json_empty_as_default) - treat empty fields in JSON input as default values. Default value - `false`. For complex default expressions [input_format_defaults_for_omitted_fields](/docs/en/operations/settings/settings-formats.md/#input_format_defaults_for_omitted_fields) must be enabled too.
 - [output_format_json_quote_64bit_integers](/docs/en/operations/settings/settings-formats.md/#output_format_json_quote_64bit_integers) - controls quoting of 64-bit integers in JSON output format. Default value - `true`.
 - [output_format_json_quote_64bit_floats](/docs/en/operations/settings/settings-formats.md/#output_format_json_quote_64bit_floats) - controls quoting of 64-bit floats in JSON output format. Default value - `false`.
 - [output_format_json_quote_denormals](/docs/en/operations/settings/settings-formats.md/#output_format_json_quote_denormals) - enables '+nan', '-nan', '+inf', '-inf' outputs in JSON output format. Default value - `false`.

docs/en/operations/settings/settings-formats.md

@@ -752,6 +752,17 @@ Possible values:
 
 Default value: 0.
 
+### input_format_json_empty_as_default {#input_format_json_empty_as_default}
+
+When enabled, replace empty input fields in JSON with default values. For complex default expressions `input_format_defaults_for_omitted_fields` must be enabled too.
+
+Possible values:
+
++ 0 — Disable.
++ 1 — Enable.
+
+Default value: 0.
+
 ## TSV format settings {#tsv-format-settings}
 
 ### input_format_tsv_empty_as_default {#input_format_tsv_empty_as_default}

src/Core/Settings.h

@@ -1144,6 +1144,7 @@ class IColumn;
     M(Bool, input_format_try_infer_variants, false, "Try to infer the Variant type in text formats when there is more than one possible type for column/array elements", 0) \
     M(Bool, type_json_skip_duplicated_paths, false, "When enabled, during parsing JSON object into JSON type duplicated paths will be ignored and only the first one will be inserted instead of an exception", 0) \
     M(UInt64, input_format_json_max_depth, 1000, "Maximum depth of a field in JSON. This is not a strict limit, it does not have to be applied precisely.", 0) \
+    M(Bool, input_format_json_empty_as_default, false, "Treat empty fields in JSON input as default values.", 0) \
     M(Bool, input_format_try_infer_integers, true, "Try to infer integers instead of floats while schema inference in text formats", 0) \
     M(Bool, input_format_try_infer_dates, true, "Try to infer dates from string fields while schema inference in text formats", 0) \
     M(Bool, input_format_try_infer_datetimes, true, "Try to infer datetimes from string fields while schema inference in text formats", 0) \

src/Core/SettingsChangesHistory.cpp

@@ -71,6 +71,7 @@ static std::initializer_list<std::pair<ClickHouseVersion, SettingsChangesHistory
     },
     {"24.9",
         {
+            {"input_format_json_empty_as_default", false, false, "Added new setting to allow to treat empty fields in JSON input as default values."},
             {"input_format_try_infer_variants", false, false, "Try to infer Variant type in text formats when there is more than one possible type for column/array elements"},
             {"join_output_by_rowlist_perkey_rows_threshold", 0, 5, "The lower limit of per-key average rows in the right table to determine whether to output by row list in hash join."},
             {"create_if_not_exists", false, false, "New setting."},

src/DataTypes/Serializations/SerializationArray.cpp

@@ -11,6 +11,7 @@
 #include <IO/WriteBufferFromString.h>
 
 #include <Formats/FormatSettings.h>
+#include <Formats/JSONUtils.h>
 
 namespace DB
 {
@@ -615,28 +616,49 @@ void SerializationArray::serializeTextJSONPretty(const IColumn & column, size_t
 }
 
 
-void SerializationArray::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+template <typename ReturnType>
+ReturnType SerializationArray::deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    deserializeTextImpl(column, istr,
-        [&](IColumn & nested_column)
+    auto deserialize_nested = [&settings, this](IColumn & nested_column, ReadBuffer & buf) -> ReturnType
+    {
+        if constexpr (std::is_same_v<ReturnType, void>)
         {
             if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(nested_column))
-                SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(nested_column, istr, settings, nested);
+                SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(nested_column, buf, settings, nested);
             else
-                nested->deserializeTextJSON(nested_column, istr, settings);
-        }, false);
+                nested->deserializeTextJSON(nested_column, buf, settings);
+        }
+        else
+        {
+            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(nested_column))
+                return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(nested_column, buf, settings, nested);
+            return nested->tryDeserializeTextJSON(nested_column, buf, settings);
+        }
+    };
+
+    if (settings.json.empty_as_default)
+        return deserializeTextImpl<ReturnType>(column, istr,
+            [&deserialize_nested, &istr](IColumn & nested_column) -> ReturnType
+            {
+                return JSONUtils::deserializeEmpyStringAsDefaultOrNested<ReturnType>(nested_column, istr, deserialize_nested);
+            }, false);
+    else
+        return deserializeTextImpl<ReturnType>(column, istr,
+            [&deserialize_nested, &istr](IColumn & nested_column) -> ReturnType
+            {
+                return deserialize_nested(nested_column, istr);
+            }, false);
+}
+
+
+void SerializationArray::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    deserializeTextJSONImpl<void>(column, istr, settings);
 }
 
 bool SerializationArray::tryDeserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    auto read_nested = [&](IColumn & nested_column)
-    {
-        if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(nested_column))
-            return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(nested_column, istr, settings, nested);
-        return nested->tryDeserializeTextJSON(nested_column, istr, settings);
-    };
-
-    return deserializeTextImpl<bool>(column, istr, std::move(read_nested), false);
+    return deserializeTextJSONImpl<bool>(column, istr, settings);
 }
 
 

src/DataTypes/Serializations/SerializationArray.h

@@ -82,6 +82,10 @@ public:
         SerializationPtr create(const SerializationPtr & prev) const override;
         ColumnPtr create(const ColumnPtr & prev) const override;
     };
+
+private:
+    template <typename ReturnType>
+    ReturnType deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const;
 };
 
 }

src/DataTypes/Serializations/SerializationMap.cpp

@@ -6,6 +6,7 @@
 #include <Columns/ColumnMap.h>
 #include <Core/Field.h>
 #include <Formats/FormatSettings.h>
+#include <Formats/JSONUtils.h>
 #include <Common/assert_cast.h>
 #include <Common/quoteString.h>
 #include <IO/WriteHelpers.h>
@@ -316,28 +317,52 @@ void SerializationMap::serializeTextJSONPretty(const IColumn & column, size_t ro
 }
 
 
-void SerializationMap::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+template <typename ReturnType>
+ReturnType SerializationMap::deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    deserializeTextImpl(column, istr,
-        [&settings](ReadBuffer & buf, const SerializationPtr & subcolumn_serialization, IColumn & subcolumn)
+    auto deserialize_nested = [&settings](IColumn & subcolumn, ReadBuffer & buf, const SerializationPtr & subcolumn_serialization) -> ReturnType
+    {
+        if constexpr (std::is_same_v<ReturnType, void>)
         {
             if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(subcolumn))
                 SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(subcolumn, buf, settings, subcolumn_serialization);
             else
                 subcolumn_serialization->deserializeTextJSON(subcolumn, buf, settings);
-        });
+        }
+        else
+        {
+            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(subcolumn))
+                return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(subcolumn, buf, settings, subcolumn_serialization);
+            return subcolumn_serialization->tryDeserializeTextJSON(subcolumn, buf, settings);
+        }
+    };
+
+    if (settings.json.empty_as_default)
+        return deserializeTextImpl<ReturnType>(column, istr,
+            [&deserialize_nested](ReadBuffer & buf, const SerializationPtr & subcolumn_serialization, IColumn & subcolumn) -> ReturnType
+            {
+                return JSONUtils::deserializeEmpyStringAsDefaultOrNested<ReturnType>(subcolumn, buf,
+                    [&deserialize_nested, &subcolumn_serialization](IColumn & subcolumn_, ReadBuffer & buf_) -> ReturnType
+                    {
+                        return deserialize_nested(subcolumn_, buf_, subcolumn_serialization);
+                    });
+            });
+    else
+        return deserializeTextImpl<ReturnType>(column, istr,
+            [&deserialize_nested](ReadBuffer & buf, const SerializationPtr & subcolumn_serialization, IColumn & subcolumn) -> ReturnType
+            {
+                return deserialize_nested(subcolumn, buf, subcolumn_serialization);
+            });
+}
+
+void SerializationMap::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    deserializeTextJSONImpl<void>(column, istr, settings);
 }
 
 bool SerializationMap::tryDeserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    auto reader = [&settings](ReadBuffer & buf, const SerializationPtr & subcolumn_serialization, IColumn & subcolumn)
-    {
-        if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(subcolumn))
-            return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(subcolumn, buf, settings, subcolumn_serialization);
-        return subcolumn_serialization->tryDeserializeTextJSON(subcolumn, buf, settings);
-    };
-
-    return deserializeTextImpl<bool>(column, istr, reader);
+    return deserializeTextJSONImpl<bool>(column, istr, settings);
 }
 
 void SerializationMap::serializeTextXML(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const

src/DataTypes/Serializations/SerializationMap.h

@@ -74,6 +74,9 @@ private:
 
     template <typename ReturnType = void, typename Reader>
     ReturnType deserializeTextImpl(IColumn & column, ReadBuffer & istr, Reader && reader) const;
+
+    template <typename ReturnType>
+    ReturnType deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const;
 };
 
 }

src/DataTypes/Serializations/SerializationTuple.cpp

@@ -5,6 +5,7 @@
 #include <Core/Field.h>
 #include <Columns/ColumnTuple.h>
 #include <Common/assert_cast.h>
+#include <Formats/JSONUtils.h>
 #include <IO/WriteHelpers.h>
 #include <IO/ReadHelpers.h>
 #include <IO/ReadBufferFromString.h>
@@ -313,27 +314,9 @@ void SerializationTuple::serializeTextJSONPretty(const IColumn & column, size_t
 }
 
 template <typename ReturnType>
-ReturnType SerializationTuple::deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+ReturnType SerializationTuple::deserializeTupleJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings, auto && deserialize_element) const
 {
-    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
-
-    auto deserialize_element = [&](IColumn & element_column, size_t element_pos)
-    {
-        if constexpr (throw_exception)
-        {
-            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(element_column))
-                SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(element_column, istr, settings, elems[element_pos]);
-            else
-                elems[element_pos]->deserializeTextJSON(element_column, istr, settings);
-            return true;
-        }
-        else
-        {
-            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(element_column))
-                return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(element_column, istr, settings, elems[element_pos]);
-            return elems[element_pos]->tryDeserializeTextJSON(element_column, istr, settings);
-        }
-    };
+    static constexpr auto throw_exception = std::is_same_v<ReturnType, void>;
 
     if (settings.json.read_named_tuples_as_objects
         && have_explicit_names)
@@ -506,12 +489,51 @@ ReturnType SerializationTuple::deserializeTextJSONImpl(IColumn & column, ReadBuf
     }
 }
 
-void SerializationTuple::deserializeTextJSON(DB::IColumn & column, DB::ReadBuffer & istr, const DB::FormatSettings & settings) const
+template <typename ReturnType>
+ReturnType SerializationTuple::deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    deserializeTextJSONImpl(column, istr, settings);
+    auto deserialize_nested = [&settings](IColumn & nested_column, ReadBuffer & buf, const SerializationPtr & nested_column_serialization) -> ReturnType
+    {
+        if constexpr (std::is_same_v<ReturnType, void>)
+        {
+            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(nested_column))
+                SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(nested_column, buf, settings, nested_column_serialization);
+            else
+                nested_column_serialization->deserializeTextJSON(nested_column, buf, settings);
+        }
+        else
+        {
+            if (settings.null_as_default && !isColumnNullableOrLowCardinalityNullable(nested_column))
+                return SerializationNullable::tryDeserializeNullAsDefaultOrNestedTextJSON(nested_column, buf, settings, nested_column_serialization);
+            else
+                return nested_column_serialization->tryDeserializeTextJSON(nested_column, buf, settings);
+        }
+    };
+
+    if (settings.json.empty_as_default)
+        return deserializeTupleJSONImpl<ReturnType>(column, istr, settings,
+            [&deserialize_nested, &istr, this](IColumn & nested_column, size_t element_pos) -> ReturnType
+            {
+                return JSONUtils::deserializeEmpyStringAsDefaultOrNested<ReturnType>(nested_column, istr,
+                    [&deserialize_nested, element_pos, this](IColumn & nested_column_, ReadBuffer & buf) -> ReturnType
+                    {
+                        return deserialize_nested(nested_column_, buf, elems[element_pos]);
+                    });
+            });
+    else
+        return deserializeTupleJSONImpl<ReturnType>(column, istr, settings,
+            [&deserialize_nested, &istr, this](IColumn & nested_column, size_t element_pos) -> ReturnType
+            {
+                return deserialize_nested(nested_column, istr, elems[element_pos]);
+            });
 }
 
-bool SerializationTuple::tryDeserializeTextJSON(DB::IColumn & column, DB::ReadBuffer & istr, const DB::FormatSettings & settings) const
+void SerializationTuple::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    deserializeTextJSONImpl<void>(column, istr, settings);
+}
+
+bool SerializationTuple::tryDeserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
     return deserializeTextJSONImpl<bool>(column, istr, settings);
 }

src/DataTypes/Serializations/SerializationTuple.h

@@ -81,7 +81,10 @@ private:
     template <typename ReturnType = void>
     ReturnType deserializeTextImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings, bool whole) const;
 
-    template <typename ReturnType = void>
+    template <typename ReturnType>
+    ReturnType deserializeTupleJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings, auto && deserialize_element) const;
+
+    template <typename ReturnType>
     ReturnType deserializeTextJSONImpl(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const;
 
     template <typename ReturnType = void>

src/Formats/FormatFactory.cpp

@@ -152,6 +152,7 @@ FormatSettings getFormatSettings(const ContextPtr & context, const Settings & se
     format_settings.json.try_infer_objects_as_tuples = settings.input_format_json_try_infer_named_tuples_from_objects;
     format_settings.json.throw_on_bad_escape_sequence = settings.input_format_json_throw_on_bad_escape_sequence;
     format_settings.json.ignore_unnecessary_fields = settings.input_format_json_ignore_unnecessary_fields;
+    format_settings.json.empty_as_default = settings.input_format_json_empty_as_default;
     format_settings.json.type_json_skip_duplicated_paths = settings.type_json_skip_duplicated_paths;
     format_settings.null_as_default = settings.input_format_null_as_default;
     format_settings.force_null_for_omitted_fields = settings.input_format_force_null_for_omitted_fields;

src/Formats/FormatSettings.h

@@ -237,6 +237,7 @@ struct FormatSettings
         bool infer_incomplete_types_as_strings = true;
         bool throw_on_bad_escape_sequence = true;
         bool ignore_unnecessary_fields = true;
+        bool empty_as_default = false;
         bool type_json_skip_duplicated_paths = false;
     } json{};
 

src/Formats/JSONUtils.cpp

@@ -2,12 +2,14 @@
 #include <Formats/JSONUtils.h>
 #include <Formats/ReadSchemaUtils.h>
 #include <Formats/EscapingRuleUtils.h>
+#include <IO/PeekableReadBuffer.h>
 #include <IO/ReadBufferFromString.h>
 #include <IO/WriteBufferValidUTF8.h>
 #include <DataTypes/Serializations/SerializationNullable.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeObjectDeprecated.h>
 #include <DataTypes/DataTypeFactory.h>
+#include <Common/assert_cast.h>
 
 #include <base/find_symbols.h>
 
@@ -286,11 +288,19 @@ namespace JSONUtils
                 return true;
             }
 
-            if (as_nullable)
-                return SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(column, in, format_settings, serialization);
+            auto deserialize = [as_nullable, &format_settings, &serialization](IColumn & column_, ReadBuffer & buf) -> bool
+            {
+                if (as_nullable)
+                    return SerializationNullable::deserializeNullAsDefaultOrNestedTextJSON(column_, buf, format_settings, serialization);
 
-            serialization->deserializeTextJSON(column, in, format_settings);
-            return true;
+                serialization->deserializeTextJSON(column_, buf, format_settings);
+                return true;
+            };
+
+            if (format_settings.json.empty_as_default)
+                return JSONUtils::deserializeEmpyStringAsDefaultOrNested<bool, false>(column, in, deserialize);
+            else
+                return deserialize(column, in);
         }
         catch (Exception & e)
         {
@@ -920,6 +930,78 @@ namespace JSONUtils
         }
     }
 
+    template <typename ReturnType, bool default_column_return_value>
+    ReturnType deserializeEmpyStringAsDefaultOrNested(IColumn & column, ReadBuffer & istr, const NestedDeserialize<ReturnType> & deserialize_nested)
+    {
+        static constexpr auto throw_exception = std::is_same_v<ReturnType, void>;
+
+        static constexpr auto EMPTY_STRING = "\"\"";
+        static constexpr auto EMPTY_STRING_LENGTH = std::string_view(EMPTY_STRING).length();
+
+        if (istr.eof() || *istr.position() != EMPTY_STRING[0])
+            return deserialize_nested(column, istr);
+
+        auto do_deserialize = [](IColumn & column_, ReadBuffer & buf, auto && check_for_empty_string, auto && deserialize) -> ReturnType
+        {
+            if (check_for_empty_string(buf))
+            {
+                column_.insertDefault();
+                return ReturnType(default_column_return_value);
+            }
+            return deserialize(column_, buf);
+        };
+
+        if (istr.available() >= EMPTY_STRING_LENGTH)
+        {
+            /// We have enough data in buffer to check if we have an empty string.
+            auto check_for_empty_string = [](ReadBuffer & buf) -> bool
+            {
+                auto * pos = buf.position();
+                if (checkString(EMPTY_STRING, buf))
+                    return true;
+                buf.position() = pos;
+                return false;
+            };
+
+            return do_deserialize(column, istr, check_for_empty_string, deserialize_nested);
+        }
+
+        /// We don't have enough data in buffer to check if we have an empty string.
+        /// Use PeekableReadBuffer to make a checkpoint before checking for an
+        /// empty string and rollback if check was failed.
+
+        auto check_for_empty_string = [](ReadBuffer & buf) -> bool
+        {
+            auto & peekable_buf = assert_cast<PeekableReadBuffer &>(buf);
+            peekable_buf.setCheckpoint();
+            SCOPE_EXIT(peekable_buf.dropCheckpoint());
+            if (checkString(EMPTY_STRING, peekable_buf))
+                return true;
+            peekable_buf.rollbackToCheckpoint();
+            return false;
+        };
+
+        auto deserialize_nested_with_check = [&deserialize_nested](IColumn & column_, ReadBuffer & buf) -> ReturnType
+        {
+            auto & peekable_buf = assert_cast<PeekableReadBuffer &>(buf);
+            if constexpr (throw_exception)
+                deserialize_nested(column_, peekable_buf);
+            else if (!deserialize_nested(column_, peekable_buf))
+                return ReturnType(false);
+
+            if (unlikely(peekable_buf.hasUnreadData()))
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Incorrect state while parsing JSON: PeekableReadBuffer has unread data in own memory: {}", String(peekable_buf.position(), peekable_buf.available()));
+
+            return ReturnType(true);
+        };
+
+        PeekableReadBuffer peekable_buf(istr, true);
+        return do_deserialize(column, peekable_buf, check_for_empty_string, deserialize_nested_with_check);
+    }
+
+    template void deserializeEmpyStringAsDefaultOrNested<void, true>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<void> & deserialize_nested);
+    template bool deserializeEmpyStringAsDefaultOrNested<bool, true>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<bool> & deserialize_nested);
+    template bool deserializeEmpyStringAsDefaultOrNested<bool, false>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<bool> & deserialize_nested);
 }
 
 }

src/Formats/JSONUtils.h

@@ -8,6 +8,7 @@
 #include <IO/Progress.h>
 #include <Core/NamesAndTypes.h>
 #include <Common/Stopwatch.h>
+#include <functional>
 #include <utility>
 
 namespace DB
@@ -146,6 +147,16 @@ namespace JSONUtils
 
     bool skipUntilFieldInObject(ReadBuffer & in, const String & desired_field_name, const FormatSettings::JSON & settings);
     void skipTheRestOfObject(ReadBuffer & in, const FormatSettings::JSON & settings);
+
+    template <typename ReturnType>
+    using NestedDeserialize = std::function<ReturnType(IColumn &, ReadBuffer &)>;
+
+    template <typename ReturnType, bool default_column_return_value = true>
+    ReturnType deserializeEmpyStringAsDefaultOrNested(IColumn & column, ReadBuffer & istr, const NestedDeserialize<ReturnType> & deserialize_nested);
+
+    extern template void deserializeEmpyStringAsDefaultOrNested<void, true>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<void> & deserialize_nested);
+    extern template bool deserializeEmpyStringAsDefaultOrNested<bool, true>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<bool> & deserialize_nested);
+    extern template bool deserializeEmpyStringAsDefaultOrNested<bool, false>(IColumn & column, ReadBuffer & istr, const NestedDeserialize<bool> & deserialize_nested);
 }
 
 }

tests/queries/0_stateless/03222_json_empty_as_default.reference

@@ -0,0 +1,47 @@
+-- Simple types
+-- { echoOn }
+SELECT x FROM format(JSONEachRow, 'x Date', '{"x":""}');
+1970-01-01
+SELECT x FROM format(JSONEachRow, 'x Date32', '{"x":""}');
+1970-01-01
+SELECT toTimeZone(x, 'UTC') FROM format(JSONEachRow, 'x DateTime', '{"x":""}');
+1970-01-01 00:00:00
+SELECT toTimeZone(x, 'UTC') FROM format(JSONEachRow, 'x DateTime64', '{"x":""}');
+1970-01-01 00:00:00.000
+SELECT x FROM format(JSONEachRow, 'x IPv4', '{"x":""}');
+0.0.0.0
+SELECT x FROM format(JSONEachRow, 'x IPv6', '{"x":""}');
+::
+SELECT x FROM format(JSONEachRow, 'x UUID', '{"x":""}');
+00000000-0000-0000-0000-000000000000
+-- { echoOn }
+SELECT COUNT(DISTINCT col) FROM table1;
+1
+-- { echoOn }
+SELECT * FROM table1 ORDER BY address ASC;
+::
+2001:db8:3333:4444:5555:6666:7777:8888
+-- Nullable
+-- { echoOn }
+SELECT x FROM format(JSONEachRow, 'x Nullable(IPv6)', '{"x":""}');
+\N
+-- Compound types
+SELECT x FROM format(JSONEachRow, 'x Array(UUID)', '{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e",""]}');
+['00000000-0000-0000-0000-000000000000','b15f852c-c41a-4fd6-9247-1929c841715e','00000000-0000-0000-0000-000000000000']
+SELECT x FROM format(JSONEachRow, 'x Array(Nullable(IPv6))', '{"x":["",""]}');
+[NULL,NULL]
+SELECT x FROM format(JSONEachRow, 'x Tuple(Date, IPv4, String)', '{"x":["", "", "abc"]}');
+('1970-01-01','0.0.0.0','abc')
+SELECT x FROM format(JSONEachRow, 'x Map(String, IPv6)', '{"x":{"abc": ""}}');
+{'abc':'::'}
+SELECT x FROM format(JSONEachRow, 'x Variant(Date, UUID)', '{"x":""}');
+\N
+-- Deep composition
+SELECT x FROM format(JSONEachRow, 'x Array(Array(IPv6))', '{"x":[["2001:db8:3333:4444:CCCC:DDDD:EEEE:FFFF", ""], ["", "2001:db8:3333:4444:5555:6666:7777:8888"]]}');
+[['2001:db8:3333:4444:cccc:dddd:eeee:ffff','::'],['::','2001:db8:3333:4444:5555:6666:7777:8888']]
+SELECT x FROM format(JSONEachRow, 'x Variant(Date, Array(UUID))', '{"x":["", "b15f852c-c41a-4fd6-9247-1929c841715e"]}');
+['00000000-0000-0000-0000-000000000000','b15f852c-c41a-4fd6-9247-1929c841715e']
+SELECT x FROM format(JSONEachRow, 'x Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))', '{"x":[[""], ["",{"abc":""}]]}');
+(['00000000-0000-0000-0000-000000000000'],('00000000-0000-0000-0000-000000000000',{'abc':'::'}))
+SELECT x FROM format(JSONEachRow, 'x Map(Tuple(Date,IPv4), Variant(UUID,IPv6))', '{"x":{["",""]:""}}');
+{('1970-01-01','0.0.0.0'):NULL}

tests/queries/0_stateless/03222_json_empty_as_default.sql

@@ -0,0 +1,60 @@
+SET input_format_json_empty_as_default = 1, allow_experimental_variant_type = 1;
+
+-- Simple types
+-- { echoOn }
+SELECT x FROM format(JSONEachRow, 'x Date', '{"x":""}');
+SELECT x FROM format(JSONEachRow, 'x Date32', '{"x":""}');
+SELECT toTimeZone(x, 'UTC') FROM format(JSONEachRow, 'x DateTime', '{"x":""}');
+SELECT toTimeZone(x, 'UTC') FROM format(JSONEachRow, 'x DateTime64', '{"x":""}');
+SELECT x FROM format(JSONEachRow, 'x IPv4', '{"x":""}');
+SELECT x FROM format(JSONEachRow, 'x IPv6', '{"x":""}');
+SELECT x FROM format(JSONEachRow, 'x UUID', '{"x":""}');
+-- { echoOff }
+
+-- Simple type AggregateFunction
+DROP TABLE IF EXISTS table1;
+CREATE TABLE table1(col AggregateFunction(uniq, UInt64)) ENGINE=Memory();
+DROP TABLE IF EXISTS table2;
+CREATE TABLE table2(UserID UInt64) ENGINE=Memory();
+
+INSERT INTO table1 SELECT uniqState(UserID) FROM table2;
+INSERT INTO table1 SELECT x FROM format(JSONEachRow, 'x AggregateFunction(uniq, UInt64)' AS T, '{"x":""}');
+
+-- { echoOn }
+SELECT COUNT(DISTINCT col) FROM table1;
+-- { echoOff }
+
+DROP TABLE table1;
+DROP TABLE table2;
+
+-- The setting input_format_defaults_for_omitted_fields determines the default value if enabled.
+CREATE TABLE table1(address IPv6 DEFAULT toIPv6('2001:db8:3333:4444:5555:6666:7777:8888')) ENGINE=Memory();
+
+SET input_format_defaults_for_omitted_fields = 0;
+INSERT INTO table1 FORMAT JSONEachRow {"address":""};
+
+SET input_format_defaults_for_omitted_fields = 1;
+INSERT INTO table1 FORMAT JSONEachRow {"address":""};
+
+-- { echoOn }
+SELECT * FROM table1 ORDER BY address ASC;
+-- { echoOff }
+
+DROP TABLE table1;
+
+-- Nullable
+-- { echoOn }
+SELECT x FROM format(JSONEachRow, 'x Nullable(IPv6)', '{"x":""}');
+
+-- Compound types
+SELECT x FROM format(JSONEachRow, 'x Array(UUID)', '{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e",""]}');
+SELECT x FROM format(JSONEachRow, 'x Array(Nullable(IPv6))', '{"x":["",""]}');
+SELECT x FROM format(JSONEachRow, 'x Tuple(Date, IPv4, String)', '{"x":["", "", "abc"]}');
+SELECT x FROM format(JSONEachRow, 'x Map(String, IPv6)', '{"x":{"abc": ""}}');
+SELECT x FROM format(JSONEachRow, 'x Variant(Date, UUID)', '{"x":""}');
+
+-- Deep composition
+SELECT x FROM format(JSONEachRow, 'x Array(Array(IPv6))', '{"x":[["2001:db8:3333:4444:CCCC:DDDD:EEEE:FFFF", ""], ["", "2001:db8:3333:4444:5555:6666:7777:8888"]]}');
+SELECT x FROM format(JSONEachRow, 'x Variant(Date, Array(UUID))', '{"x":["", "b15f852c-c41a-4fd6-9247-1929c841715e"]}');
+SELECT x FROM format(JSONEachRow, 'x Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))', '{"x":[[""], ["",{"abc":""}]]}');
+SELECT x FROM format(JSONEachRow, 'x Map(Tuple(Date,IPv4), Variant(UUID,IPv6))', '{"x":{["",""]:""}}');

tests/queries/0_stateless/03222_json_empty_as_default_small_read_buffer.reference

@@ -0,0 +1,8 @@
+Array(UUID)
+{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e","00000000-0000-0000-0000-000000000000"]}
+{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e","00000000-0000-0000-0000-000000000000"]}
+{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e","00000000-0000-0000-0000-000000000000"]}
+Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))
+{"x":[["00000000-0000-0000-0000-000000000000"],["00000000-0000-0000-0000-000000000000",{"abc":"::"}]]}
+{"x":[["00000000-0000-0000-0000-000000000000"],["00000000-0000-0000-0000-000000000000",{"abc":"::"}]]}
+{"x":[["00000000-0000-0000-0000-000000000000"],["00000000-0000-0000-0000-000000000000",{"abc":"::"}]]}

tests/queries/0_stateless/03222_json_empty_as_default_small_read_buffer.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+# Tags: no-parallel
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CURDIR"/../shell_config.sh
+
+DATA_FILE=$CLICKHOUSE_TEST_UNIQUE_NAME.json
+
+# Wrapper for clickhouse-client to always output in JSONEachRow format, that
+# way format settings will not affect output.
+function clickhouse_local()
+{
+    $CLICKHOUSE_LOCAL --output-format JSONEachRow "$@"
+}
+
+echo 'Array(UUID)'
+echo '{"x":["00000000-0000-0000-0000-000000000000","b15f852c-c41a-4fd6-9247-1929c841715e",""]}' > $DATA_FILE
+# Use increasingly smaller read buffers.
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Array(UUID)') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=4"
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Array(UUID)') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=2"
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Array(UUID)') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=1"
+
+echo 'Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))'
+echo '{"x":[[""], ["",{"abc":""}]]}' > $DATA_FILE
+# Use increasingly smaller read buffers.
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=16"
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=8"
+clickhouse_local -q "SELECT x FROM file('$DATA_FILE', 'JSONEachRow', 'x Tuple(Array(UUID), Tuple(UUID, Map(String, IPv6)))') SETTINGS input_format_json_empty_as_default=1, input_format_parallel_parsing=0, storage_file_read_method='read', max_read_buffer_size=1"
+
+rm $DATA_FILE