Add settings for better control of indexes type in Arrow dictionary

docker/test/stateless/Dockerfile

@@ -46,7 +46,7 @@ RUN apt-get update -y \
             p7zip-full \
     && apt-get clean
 
-RUN pip3 install numpy scipy pandas Jinja2
+RUN pip3 install numpy scipy pandas Jinja2 pyarrow
 
 RUN mkdir -p /tmp/clickhouse-odbc-tmp \
    && wget -nv -O - ${odbc_driver_url} | tar --strip-components=1 -xz -C /tmp/clickhouse-odbc-tmp \

docs/en/interfaces/formats.md

@@ -2356,6 +2356,8 @@ $ clickhouse-client --query="SELECT * FROM {some_table} FORMAT Arrow" > {filenam
 ### Arrow format settings {#parquet-format-settings}
 
 - [output_format_arrow_low_cardinality_as_dictionary](/docs/en/operations/settings/settings-formats.md/#output_format_arrow_low_cardinality_as_dictionary) - enable output ClickHouse LowCardinality type as Dictionary Arrow type. Default value - `false`.
+- [output_format_arrow_use_64_bit_indexes_for_dictionary](/docs/en/operations/settings/settings-formats.md/#output_format_arrow_use_64_bit_indexes_for_dictionary) - use 64-bit integer type for Dictionary indexes. Default value - `false`.
+- [output_format_arrow_use_signed_indexes_for_dictionary](/docs/en/operations/settings/settings-formats.md/#output_format_arrow_use_signed_indexes_for_dictionary) - use signed integer type for Dictionary indexes. Default value - `true`.
 - [output_format_arrow_string_as_string](/docs/en/operations/settings/settings-formats.md/#output_format_arrow_string_as_string) - use Arrow String type instead of Binary for String columns. Default value - `false`.
 - [input_format_arrow_case_insensitive_column_matching](/docs/en/operations/settings/settings-formats.md/#input_format_arrow_case_insensitive_column_matching) - ignore case when matching Arrow columns with ClickHouse columns. Default value - `false`.
 - [input_format_arrow_allow_missing_columns](/docs/en/operations/settings/settings-formats.md/#input_format_arrow_allow_missing_columns) - allow missing columns while reading Arrow data. Default value - `false`.

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

@@ -1269,6 +1269,28 @@ Possible values:
 
 Default value: `0`.
 
+### output_format_arrow_use_signed_indexes_for_dictionary {#output_format_arrow_use_signed_indexes_for_dictionary}
+
+Use signed integer types instead of unsigned in `DICTIONARY` type of the [Arrow](../../interfaces/formats.md/#data-format-arrow) format during  [LowCardinality](../../sql-reference/data-types/lowcardinality.md) output when `output_format_arrow_low_cardinality_as_dictionary` is enabled.
+
+Possible values:
+
+- 0 — Unsigned integer types are used for indexes in `DICTIONARY` type.
+- 1 — Signed integer types are used for indexes in `DICTIONARY` type.
+
+Default value: `1`.
+
+### output_format_arrow_use_64_bit_indexes_for_dictionary {#output_format_arrow_use_64_bit_indexes_for_dictionary}
+
+Use 64-bit integer type in `DICTIONARY` type of the [Arrow](../../interfaces/formats.md/#data-format-arrow) format during  [LowCardinality](../../sql-reference/data-types/lowcardinality.md) output when `output_format_arrow_low_cardinality_as_dictionary` is enabled.
+
+Possible values:
+
+- 0 — Type for indexes in `DICTIONARY` type is determined automatically.
+- 1 — 64-bit integer type is used for indexes in `DICTIONARY` type.
+
+Default value: `0`.
+
 ### output_format_arrow_string_as_string {#output_format_arrow_string_as_string}
 
 Use Arrow String type instead of Binary for String columns.

src/Core/Settings.h

@@ -1100,6 +1100,8 @@ class IColumn;
     M(UInt64, cross_to_inner_join_rewrite, 1, "Use inner join instead of comma/cross join if there're joining expressions in the WHERE section. Values: 0 - no rewrite, 1 - apply if possible for comma/cross, 2 - force rewrite all comma joins, cross - if possible", 0) \
     \
     M(Bool, output_format_arrow_low_cardinality_as_dictionary, false, "Enable output LowCardinality type as Dictionary Arrow type", 0) \
+    M(Bool, output_format_arrow_use_signed_indexes_for_dictionary, true, "Use signed integers for dictionary indexes in Arrow format", 0) \
+    M(Bool, output_format_arrow_use_64_bit_indexes_for_dictionary, false, "Always use 64 bit integers for dictionary indexes in Arrow format", 0) \
     M(Bool, output_format_arrow_string_as_string, false, "Use Arrow String type instead of Binary for String columns", 0) \
     M(Bool, output_format_arrow_fixed_string_as_fixed_byte_array, true, "Use Arrow FIXED_SIZE_BINARY type instead of Binary for FixedString columns.", 0) \
     M(ArrowCompression, output_format_arrow_compression_method, "lz4_frame", "Compression method for Arrow output format. Supported codecs: lz4_frame, zstd, none (uncompressed)", 0) \

src/Core/SettingsChangesHistory.h

@@ -82,7 +82,8 @@ namespace SettingsChangesHistory
 static std::map<ClickHouseVersion, SettingsChangesHistory::SettingsChanges> settings_changes_history =
 {
     {"24.1", {{"print_pretty_type_names", false, true, "Better user experience."},
-              {"input_format_json_read_bools_as_strings", false, true, "Allow to read bools as strings in JSON formats by default"}}},
+              {"input_format_json_read_bools_as_strings", false, true, "Allow to read bools as strings in JSON formats by default"},
+              {"output_format_arrow_use_signed_indexes_for_dictionary", false, true, "Use signed indexes type for Arrow dictionaries by default as it's recommended"}}},
     {"23.12", {{"allow_suspicious_ttl_expressions", true, false, "It is a new setting, and in previous versions the behavior was equivalent to allowing."},
               {"input_format_parquet_allow_missing_columns", false, true, "Allow missing columns in Parquet files by default"},
               {"input_format_orc_allow_missing_columns", false, true, "Allow missing columns in ORC files by default"},

src/Formats/FormatFactory.cpp

@@ -182,6 +182,8 @@ FormatSettings getFormatSettings(ContextPtr context, const Settings & settings)
     format_settings.with_types_use_header = settings.input_format_with_types_use_header;
     format_settings.write_statistics = settings.output_format_write_statistics;
     format_settings.arrow.low_cardinality_as_dictionary = settings.output_format_arrow_low_cardinality_as_dictionary;
+    format_settings.arrow.use_signed_indexes_for_dictionary = settings.output_format_arrow_use_signed_indexes_for_dictionary;
+    format_settings.arrow.use_64_bit_indexes_for_dictionary = settings.output_format_arrow_use_64_bit_indexes_for_dictionary;
     format_settings.arrow.allow_missing_columns = settings.input_format_arrow_allow_missing_columns;
     format_settings.arrow.skip_columns_with_unsupported_types_in_schema_inference = settings.input_format_arrow_skip_columns_with_unsupported_types_in_schema_inference;
     format_settings.arrow.skip_columns_with_unsupported_types_in_schema_inference = settings.input_format_arrow_skip_columns_with_unsupported_types_in_schema_inference;

src/Formats/FormatSettings.h

@@ -122,6 +122,8 @@ struct FormatSettings
     {
         UInt64 row_group_size = 1000000;
         bool low_cardinality_as_dictionary = false;
+        bool use_signed_indexes_for_dictionary = false;
+        bool use_64_bit_indexes_for_dictionary = false;
         bool allow_missing_columns = false;
         bool skip_columns_with_unsupported_types_in_schema_inference = false;
         bool case_insensitive_column_matching = false;

src/Processors/Formats/Impl/ArrowBlockOutputFormat.cpp

@@ -53,9 +53,14 @@ void ArrowBlockOutputFormat::consume(Chunk chunk)
         ch_column_to_arrow_column = std::make_unique<CHColumnToArrowColumn>(
             header,
             "Arrow",
-            format_settings.arrow.low_cardinality_as_dictionary,
-            format_settings.arrow.output_string_as_string,
-            format_settings.arrow.output_fixed_string_as_fixed_byte_array);
+            CHColumnToArrowColumn::Settings
+            {
+                format_settings.arrow.output_string_as_string,
+                format_settings.arrow.output_fixed_string_as_fixed_byte_array,
+                format_settings.arrow.low_cardinality_as_dictionary,
+                format_settings.arrow.use_signed_indexes_for_dictionary,
+                format_settings.arrow.use_64_bit_indexes_for_dictionary
+            });
     }
 
     auto chunks = std::vector<Chunk>();