Merge bc866e64ea into e0f8b8d351

Fix ephemeral column comment

docker/server/README.md

@@ -16,16 +16,18 @@ ClickHouse works 100-1000x faster than traditional database management systems,
 
 For more information and documentation see https://clickhouse.com/.
 
-<!-- This is not related to the docker official library, remove it before commit to https://github.com/docker-library/docs -->
 ## Versions
 
 -	The `latest` tag points to the latest release of the latest stable branch.
 -	Branch tags like `22.2` point to the latest release of the corresponding branch.
--	Full version tags like `22.2.3.5` point to the corresponding release.
+-	Full version tags like `22.2.3` and `22.2.3.5` point to the corresponding release.
+<!-- docker-official-library:off -->
+<!-- This is not related to the docker official library, remove it before commit to https://github.com/docker-library/docs -->
 -	The tag `head` is built from the latest commit to the default branch.
 -	Each tag has optional `-alpine` suffix to reflect that it's built on top of `alpine`.
 
 <!-- REMOVE UNTIL HERE -->
+<!-- docker-official-library:on -->
 ### Compatibility
 
 -	The amd64 image requires support for [SSE3 instructions](https://en.wikipedia.org/wiki/SSE3). Virtually all x86 CPUs after 2005 support SSE3.

docker/server/README.src/content.md

@@ -10,16 +10,18 @@ ClickHouse works 100-1000x faster than traditional database management systems,
 
 For more information and documentation see https://clickhouse.com/.
 
-<!-- This is not related to the docker official library, remove it before commit to https://github.com/docker-library/docs -->
 ## Versions
 
 -	The `latest` tag points to the latest release of the latest stable branch.
 -	Branch tags like `22.2` point to the latest release of the corresponding branch.
--	Full version tags like `22.2.3.5` point to the corresponding release.
+-	Full version tags like `22.2.3` and `22.2.3.5` point to the corresponding release.
+<!-- docker-official-library:off -->
+<!-- This is not related to the docker official library, remove it before commit to https://github.com/docker-library/docs -->
 -	The tag `head` is built from the latest commit to the default branch.
 -	Each tag has optional `-alpine` suffix to reflect that it's built on top of `alpine`.
 
 <!-- REMOVE UNTIL HERE -->
+<!-- docker-official-library:on -->
 ### Compatibility
 
 -	The amd64 image requires support for [SSE3 instructions](https://en.wikipedia.org/wiki/SSE3). Virtually all x86 CPUs after 2005 support SSE3.

docs/changelogs/v24.8.1.2684-lts.md

@@ -522,4 +522,3 @@ sidebar_label: 2024
 * Backported in [#68518](https://github.com/ClickHouse/ClickHouse/issues/68518): Minor update in Dynamic/JSON serializations. [#68459](https://github.com/ClickHouse/ClickHouse/pull/68459) ([Kruglov Pavel](https://github.com/Avogar)).
 * Backported in [#68558](https://github.com/ClickHouse/ClickHouse/issues/68558): CI: Minor release workflow fix. [#68536](https://github.com/ClickHouse/ClickHouse/pull/68536) ([Max K.](https://github.com/maxknv)).
 * Backported in [#68576](https://github.com/ClickHouse/ClickHouse/issues/68576): CI: Tidy build timeout from 2h to 3h. [#68567](https://github.com/ClickHouse/ClickHouse/pull/68567) ([Max K.](https://github.com/maxknv)).
-

docs/changelogs/v24.9.1.3278-stable.md

@@ -497,4 +497,3 @@ sidebar_label: 2024
 * Backported in [#69899](https://github.com/ClickHouse/ClickHouse/issues/69899): Revert "Merge pull request [#69032](https://github.com/ClickHouse/ClickHouse/issues/69032) from alexon1234/include_real_time_execution_in_http_header". [#69885](https://github.com/ClickHouse/ClickHouse/pull/69885) ([Alexey Milovidov](https://github.com/alexey-milovidov)).
 * Backported in [#69931](https://github.com/ClickHouse/ClickHouse/issues/69931): RIPE is an acronym and thus should be capital. RIPE stands for **R**ACE **I**ntegrity **P**rimitives **E**valuation and RACE stands for **R**esearch and Development in **A**dvanced **C**ommunications **T**echnologies in **E**urope. [#69901](https://github.com/ClickHouse/ClickHouse/pull/69901) ([Nikita Mikhaylov](https://github.com/nikitamikhaylov)).
 * Backported in [#70034](https://github.com/ClickHouse/ClickHouse/issues/70034): Revert "Add RIPEMD160 function". [#70005](https://github.com/ClickHouse/ClickHouse/pull/70005) ([Robert Schulze](https://github.com/rschu1ze)).
-

programs/server/dashboard.html

@@ -476,7 +476,7 @@
             <input id="edit" type="button" value="✎" style="display: none;">
             <input id="add" type="button" value="Add chart" style="display: none;">
             <input id="reload" type="button" value="Reload">
-            <span id="search-span" class="nowrap" style="display: none;"><input id="search" type="button" value="🔎" title="Run query to obtain list of charts from ClickHouse"><input id="search-query" name="search" type="text" spellcheck="false"></span>
+            <span id="search-span" class="nowrap" style="display: none;"><input id="search" type="button" value="🔎" title="Run query to obtain list of charts from ClickHouse. Either select dashboard name or write your own query"><input id="search-query" name="search" list="search-options" type="text" spellcheck="false"><datalist id="search-options"></datalist></span>
             <div id="chart-params"></div>
         </div>
     </form>
@@ -532,9 +532,15 @@ const errorMessages = [
     }
 ]
 
+/// Dashboard selector
+const dashboardSearchQuery = (dashboard_name) => `SELECT title, query FROM system.dashboards WHERE dashboard = '${dashboard_name}'`;
+let dashboard_queries = {
+    "Overview": dashboardSearchQuery("Overview"),
+};
+const default_dashboard = 'Overview';
 
 /// Query to fill `queries` list for the dashboard
-let search_query = `SELECT title, query FROM system.dashboards WHERE dashboard = 'Overview'`;
+let search_query = dashboardSearchQuery(default_dashboard);
 let customized = false;
 let queries = [];
 
@@ -1439,7 +1445,7 @@ async function reloadAll(do_search) {
     try {
         updateParams();
         if (do_search) {
-            search_query = document.getElementById('search-query').value;
+            search_query = toSearchQuery(document.getElementById('search-query').value);
             queries = [];
             refreshCustomized(false);
         }
@@ -1504,7 +1510,7 @@ function updateFromState() {
     document.getElementById('url').value = host;
     document.getElementById('user').value = user;
     document.getElementById('password').value = password;
-    document.getElementById('search-query').value = search_query;
+    document.getElementById('search-query').value = fromSearchQuery(search_query);
     refreshCustomized();
 }
 
@@ -1543,6 +1549,44 @@ if (window.location.hash) {
     } catch {}
 }
 
+function fromSearchQuery(query) {
+    for (const dashboard_name in dashboard_queries) {
+        if (query == dashboard_queries[dashboard_name])
+            return dashboard_name;
+    }
+    return query;
+}
+
+function toSearchQuery(value) {
+    if (value in dashboard_queries)
+        return dashboard_queries[value];
+    else
+        return value;
+}
+
+async function populateSearchOptions() {
+    let {reply, error} = await doFetch("SELECT dashboard FROM system.dashboards GROUP BY dashboard ORDER BY ALL");
+    if (error) {
+        throw new Error(error);
+    }
+    let data = reply.data;
+    if (data.dashboard.length == 0) {
+        console.log("Unable to fetch dashboards list");
+        return;
+    }
+    dashboard_queries = {};
+    for (let i = 0; i < data.dashboard.length; i++) {
+        const dashboard = data.dashboard[i];
+        dashboard_queries[dashboard] = dashboardSearchQuery(dashboard);
+    }
+    const searchOptions = document.getElementById('search-options');
+    for (const dashboard in dashboard_queries) {
+        const opt = document.createElement('option');
+        opt.value = dashboard;
+        searchOptions.appendChild(opt);
+    }
+}
+
 async function start() {
     try {
         updateFromState();
@@ -1558,6 +1602,7 @@ async function start() {
         } else {
             drawAll();
         }
+        await populateSearchOptions();
     } catch (e) {
         showError(e.message);
     }

src/Analyzer/Resolve/IdentifierResolver.cpp

@@ -528,7 +528,7 @@ QueryTreeNodePtr IdentifierResolver::tryResolveIdentifierFromCompoundExpression(
   *
   * Resolve strategy:
   * 1. Try to bind identifier to scope argument name to node map.
-  * 2. If identifier is binded but expression context and node type are incompatible return nullptr.
+  * 2. If identifier is bound but expression context and node type are incompatible return nullptr.
   *
   * It is important to support edge cases, where we lookup for table or function node, but argument has same name.
   * Example: WITH (x -> x + 1) AS func, (func -> func(1) + func) AS lambda SELECT lambda(1);

src/Client/ReplxxLineReader.cpp

@@ -362,7 +362,7 @@ ReplxxLineReader::ReplxxLineReader(
     if (highlighter)
         rx.set_highlighter_callback(highlighter);
 
-    /// By default C-p/C-n binded to COMPLETE_NEXT/COMPLETE_PREV,
+    /// By default C-p/C-n bound to COMPLETE_NEXT/COMPLETE_PREV,
     /// bind C-p/C-n to history-previous/history-next like readline.
     rx.bind_key(Replxx::KEY::control('N'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::HISTORY_NEXT, code); });
     rx.bind_key(Replxx::KEY::control('P'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::HISTORY_PREVIOUS, code); });
@@ -384,9 +384,9 @@ ReplxxLineReader::ReplxxLineReader(
     rx.bind_key(Replxx::KEY::control('J'), commit_action);
     rx.bind_key(Replxx::KEY::ENTER, commit_action);
 
-    /// By default COMPLETE_NEXT/COMPLETE_PREV was binded to C-p/C-n, re-bind
+    /// By default COMPLETE_NEXT/COMPLETE_PREV was bound to C-p/C-n, re-bind
     /// to M-P/M-N (that was used for HISTORY_COMMON_PREFIX_SEARCH before, but
-    /// it also binded to M-p/M-n).
+    /// it also bound to M-p/M-n).
     rx.bind_key(Replxx::KEY::meta('N'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::COMPLETE_NEXT, code); });
     rx.bind_key(Replxx::KEY::meta('P'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::COMPLETE_PREVIOUS, code); });
     /// By default M-BACKSPACE is KILL_TO_WHITESPACE_ON_LEFT, while in readline it is backward-kill-word

src/Parsers/ParserCreateQuery.h

@@ -237,6 +237,7 @@ bool IParserColumnDeclaration<NameParser>::parseImpl(Pos & pos, ASTPtr & node, E
             null_modifier.emplace(true);
     }
 
+    bool is_comment = false;
     /// Collate is also allowed after NULL/NOT NULL
     if (!collation_expression && s_collate.ignore(pos, expected)
         && !collation_parser.parse(pos, collation_expression, expected))
@@ -254,7 +255,9 @@ bool IParserColumnDeclaration<NameParser>::parseImpl(Pos & pos, ASTPtr & node, E
     else if (s_ephemeral.ignore(pos, expected))
     {
         default_specifier = s_ephemeral.getName();
-        if (!expr_parser.parse(pos, default_expression, expected) && type)
+        if (s_comment.ignore(pos, expected))
+            is_comment = true;
+        if ((is_comment || !expr_parser.parse(pos, default_expression, expected)) && type)
         {
             ephemeral_default = true;
 
@@ -289,6 +292,8 @@ bool IParserColumnDeclaration<NameParser>::parseImpl(Pos & pos, ASTPtr & node, E
     if (require_type && !type && !default_expression)
         return false; /// reject column name without type
 
+    if (!is_comment)
+    {
         if ((type || default_expression) && allow_null_modifiers && !null_modifier.has_value())
         {
             if (s_not.ignore(pos, expected))
@@ -300,8 +305,9 @@ bool IParserColumnDeclaration<NameParser>::parseImpl(Pos & pos, ASTPtr & node, E
             else if (s_null.ignore(pos, expected))
                 null_modifier.emplace(true);
         }
+    }
 
-    if (s_comment.ignore(pos, expected))
+    if (is_comment || s_comment.ignore(pos, expected))
     {
         /// should be followed by a string literal
         if (!string_literal_parser.parse(pos, comment_expression, expected))

src/Processors/Formats/Impl/Parquet/ParquetDataBuffer.h

@@ -48,6 +48,22 @@ public:
         consume(bytes);
     }
 
+    template <typename TValue, typename ParquetType>
+    void ALWAYS_INLINE readValuesOfDifferentSize(TValue * dst, size_t count)
+    {
+        auto necessary_bytes = count * sizeof(ParquetType);
+        checkAvaible(necessary_bytes);
+
+        const ParquetType* src = reinterpret_cast<const ParquetType*>(data);
+
+        for (std::size_t i = 0; i < count; i++)
+        {
+            dst[i] = static_cast<TValue>(src[i]);
+        }
+
+        consume(necessary_bytes);
+    }
+
     void ALWAYS_INLINE readDateTime64FromInt96(DateTime64 & dst)
     {
         static const int max_scale_num = 9;

src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp

@@ -240,8 +240,8 @@ TValue * getResizedPrimitiveData(TColumn & column, size_t size)
 } // anoynomous namespace
 
 
-template <>
-void ParquetPlainValuesReader<ColumnString>::readBatch(
+template <typename TColumn>
+void ParquetPlainByteArrayValuesReader<TColumn>::readBatch(
     MutableColumnPtr & col_ptr, LazyNullMap & null_map, UInt32 num_values)
 {
     auto & column = *assert_cast<ColumnString *>(col_ptr.get());
@@ -322,8 +322,8 @@ void ParquetBitPlainReader<TColumn>::readBatch(
 }
 
 
-template <>
-void ParquetPlainValuesReader<ColumnDecimal<DateTime64>, ParquetReaderTypes::TimestampInt96>::readBatch(
+template <typename TColumn>
+void ParquetPlainInt96ValuesReader<TColumn>::readBatch(
     MutableColumnPtr & col_ptr, LazyNullMap & null_map, UInt32 num_values)
 {
     auto cursor = col_ptr->size();
@@ -350,8 +350,8 @@ void ParquetPlainValuesReader<ColumnDecimal<DateTime64>, ParquetReaderTypes::Tim
     );
 }
 
-template <typename TColumn, ParquetReaderTypes reader_type>
-void ParquetPlainValuesReader<TColumn, reader_type>::readBatch(
+template <typename TColumn, typename ParquetType>
+void ParquetPlainValuesReader<TColumn, ParquetType>::readBatch(
     MutableColumnPtr & col_ptr, LazyNullMap & null_map, UInt32 num_values)
 {
     auto cursor = col_ptr->size();
@@ -365,11 +365,11 @@ void ParquetPlainValuesReader<TColumn, reader_type>::readBatch(
         null_map,
         /* individual_visitor */ [&](size_t nest_cursor)
         {
-            plain_data_buffer.readValue(column_data[nest_cursor]);
+            plain_data_buffer.readValuesOfDifferentSize<TValue, ParquetType>(column_data + nest_cursor, 1);
         },
         /* repeated_visitor */ [&](size_t nest_cursor, UInt32 count)
         {
-            plain_data_buffer.readBytes(column_data + nest_cursor, count * sizeof(TValue));
+            plain_data_buffer.readValuesOfDifferentSize<TValue, ParquetType>(column_data + nest_cursor, count);
         }
     );
 }
@@ -576,18 +576,19 @@ void ParquetRleDictReader<TColumnVector>::readBatch(
 }
 
 
-template class ParquetPlainValuesReader<ColumnInt32>;
-template class ParquetPlainValuesReader<ColumnUInt32>;
-template class ParquetPlainValuesReader<ColumnInt64>;
-template class ParquetPlainValuesReader<ColumnUInt64>;
-template class ParquetPlainValuesReader<ColumnBFloat16>;
-template class ParquetPlainValuesReader<ColumnFloat32>;
-template class ParquetPlainValuesReader<ColumnFloat64>;
-template class ParquetPlainValuesReader<ColumnDecimal<Decimal32>>;
-template class ParquetPlainValuesReader<ColumnDecimal<Decimal64>>;
-template class ParquetPlainValuesReader<ColumnDecimal<DateTime64>>;
-template class ParquetPlainValuesReader<ColumnString>;
-template class ParquetPlainValuesReader<ColumnUInt8>;
+template class ParquetPlainValuesReader<ColumnUInt8, int32_t>;
+template class ParquetPlainValuesReader<ColumnInt8, int32_t>;
+template class ParquetPlainValuesReader<ColumnUInt16, int32_t>;
+template class ParquetPlainValuesReader<ColumnInt16, int32_t>;
+template class ParquetPlainValuesReader<ColumnUInt32, int32_t>;
+template class ParquetPlainValuesReader<ColumnInt32, int32_t>;
+template class ParquetPlainValuesReader<ColumnUInt64, int64_t>;
+template class ParquetPlainValuesReader<ColumnInt64, int64_t>;
+template class ParquetPlainValuesReader<ColumnFloat32, float>;
+template class ParquetPlainValuesReader<ColumnFloat64, double>;
+template class ParquetPlainValuesReader<ColumnDecimal<Decimal32>, int32_t>;
+template class ParquetPlainValuesReader<ColumnDecimal<Decimal64>, int64_t>;
+template class ParquetPlainValuesReader<ColumnDecimal<DateTime64>, int64_t>;
 
 template class ParquetBitPlainReader<ColumnUInt8>;
 
@@ -598,12 +599,10 @@ template class ParquetRleLCReader<ColumnUInt8>;
 template class ParquetRleLCReader<ColumnUInt16>;
 template class ParquetRleLCReader<ColumnUInt32>;
 
-template class ParquetRleDictReader<ColumnUInt8>;
 template class ParquetRleDictReader<ColumnInt32>;
 template class ParquetRleDictReader<ColumnUInt32>;
 template class ParquetRleDictReader<ColumnInt64>;
 template class ParquetRleDictReader<ColumnUInt64>;
-template class ParquetRleDictReader<ColumnBFloat16>;
 template class ParquetRleDictReader<ColumnFloat32>;
 template class ParquetRleDictReader<ColumnFloat64>;
 template class ParquetRleDictReader<ColumnDecimal<Decimal32>>;
@@ -613,4 +612,8 @@ template class ParquetRleDictReader<ColumnDecimal<Decimal256>>;
 template class ParquetRleDictReader<ColumnDecimal<DateTime64>>;
 template class ParquetRleDictReader<ColumnString>;
 
+template class ParquetPlainByteArrayValuesReader<ColumnString>;
+
+template class ParquetPlainInt96ValuesReader<ColumnDecimal<DateTime64>>;
+
 }

src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.h

@@ -150,7 +150,7 @@ enum class ParquetReaderTypes
 /**
  * The definition level is RLE or BitPacked encoding, while data is read directly
  */
-template <typename TColumn, ParquetReaderTypes reader_type = ParquetReaderTypes::Normal>
+template <typename TColumn, typename ParquetType>
 class ParquetPlainValuesReader : public ParquetDataValuesReader
 {
 public:
@@ -172,6 +172,50 @@ private:
     ParquetDataBuffer plain_data_buffer;
 };
 
+template <typename TColumn>
+class ParquetPlainInt96ValuesReader : public ParquetDataValuesReader
+{
+public:
+
+    ParquetPlainInt96ValuesReader(
+        Int32 max_def_level_,
+        std::unique_ptr<RleValuesReader> def_level_reader_,
+        ParquetDataBuffer data_buffer_)
+        : max_def_level(max_def_level_)
+        , def_level_reader(std::move(def_level_reader_))
+        , plain_data_buffer(std::move(data_buffer_))
+    {}
+
+    void readBatch(MutableColumnPtr & col_ptr, LazyNullMap & null_map, UInt32 num_values) override;
+
+private:
+    Int32 max_def_level;
+    std::unique_ptr<RleValuesReader> def_level_reader;
+    ParquetDataBuffer plain_data_buffer;
+};
+
+template <typename TColumn>
+class ParquetPlainByteArrayValuesReader : public ParquetDataValuesReader
+{
+public:
+
+    ParquetPlainByteArrayValuesReader(
+        Int32 max_def_level_,
+        std::unique_ptr<RleValuesReader> def_level_reader_,
+        ParquetDataBuffer data_buffer_)
+        : max_def_level(max_def_level_)
+        , def_level_reader(std::move(def_level_reader_))
+        , plain_data_buffer(std::move(data_buffer_))
+    {}
+
+    void readBatch(MutableColumnPtr & col_ptr, LazyNullMap & null_map, UInt32 num_values) override;
+
+private:
+    Int32 max_def_level;
+    std::unique_ptr<RleValuesReader> def_level_reader;
+    ParquetDataBuffer plain_data_buffer;
+};
+
 template <typename TColumn>
 class ParquetBitPlainReader : public ParquetDataValuesReader
 {

src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp

@@ -173,13 +173,7 @@ ColumnPtr readDictPage(
 }
 
 
-template <typename TColumn>
-std::unique_ptr<ParquetDataValuesReader> createPlainReader(
-    const parquet::ColumnDescriptor & col_des,
-    RleValuesReaderPtr def_level_reader,
-    ParquetDataBuffer buffer);
-
-template <is_col_over_big_decimal TColumnDecimal>
+template <is_col_over_big_decimal TColumnDecimal, typename ParquetType>
 std::unique_ptr<ParquetDataValuesReader> createPlainReader(
     const parquet::ColumnDescriptor & col_des,
     RleValuesReaderPtr def_level_reader,
@@ -192,25 +186,62 @@ std::unique_ptr<ParquetDataValuesReader> createPlainReader(
         std::move(buffer));
 }
 
-template <typename TColumn>
+
+template <typename TColumn, typename ParquetType>
 std::unique_ptr<ParquetDataValuesReader> createPlainReader(
     const parquet::ColumnDescriptor & col_des,
     RleValuesReaderPtr def_level_reader,
     ParquetDataBuffer buffer)
 {
-    if (std::is_same_v<TColumn, ColumnDecimal<DateTime64>> && col_des.physical_type() == parquet::Type::INT96)
-        return std::make_unique<ParquetPlainValuesReader<TColumn, ParquetReaderTypes::TimestampInt96>>(
+    if constexpr (std::is_same_v<TColumn, ColumnDecimal<DateTime64>> && std::is_same_v<ParquetType, ParquetInt96TypeStub>)
+        return std::make_unique<ParquetPlainInt96ValuesReader<TColumn>>(
             col_des.max_definition_level(), std::move(def_level_reader), std::move(buffer));
-    return std::make_unique<ParquetPlainValuesReader<TColumn>>(
+
+    if constexpr (std::is_same_v<ParquetType, ParquetByteArrayTypeStub>)
+    {
+        return std::make_unique<ParquetPlainByteArrayValuesReader<TColumn>>(
             col_des.max_definition_level(), std::move(def_level_reader), std::move(buffer));
+    }
+
+    return std::make_unique<ParquetPlainValuesReader<TColumn, ParquetType>>(
+        col_des.max_definition_level(), std::move(def_level_reader), std::move(buffer));
+}
+
+template <typename TColumn, typename ParquetType>
+std::unique_ptr<ParquetDataValuesReader> createReader(
+    const parquet::ColumnDescriptor & col_descriptor,
+    RleValuesReaderPtr def_level_reader,
+    const uint8_t * buffer,
+    std::size_t buffer_max_size,
+    const DataTypePtr & base_data_type)
+{
+    if constexpr (std::is_same_v<ParquetType, bool>)
+    {
+        auto bit_reader = std::make_unique<arrow::bit_util::BitReader>(buffer, buffer_max_size);
+        return std::make_unique<ParquetBitPlainReader<TColumn>>(
+            col_descriptor.max_definition_level(), std::move(def_level_reader), std::move(bit_reader));
+    }
+    else
+    {
+        ParquetDataBuffer parquet_buffer = [&]()
+        {
+            if constexpr (!std::is_same_v<ColumnDecimal<DateTime64>, TColumn>)
+                return ParquetDataBuffer(buffer, buffer_max_size);
+
+            auto scale = assert_cast<const DataTypeDateTime64 &>(*base_data_type).getScale();
+            return ParquetDataBuffer(buffer, buffer_max_size, scale);
+        }();
+
+        return createPlainReader<TColumn, ParquetType>(col_descriptor, std::move(def_level_reader), parquet_buffer);
+    }
 }
 
 
 } // anonymous namespace
 
 
-template <typename TColumn>
-ParquetLeafColReader<TColumn>::ParquetLeafColReader(
+template <typename TColumn, typename ParquetType>
+ParquetLeafColReader<TColumn, ParquetType>::ParquetLeafColReader(
     const parquet::ColumnDescriptor & col_descriptor_,
     DataTypePtr base_type_,
     std::unique_ptr<parquet::ColumnChunkMetaData> meta_,
@@ -223,8 +254,8 @@ ParquetLeafColReader<TColumn>::ParquetLeafColReader(
 {
 }
 
-template <typename TColumn>
-ColumnWithTypeAndName ParquetLeafColReader<TColumn>::readBatch(UInt64 rows_num, const String & name)
+template <typename TColumn, typename ParquetType>
+ColumnWithTypeAndName ParquetLeafColReader<TColumn, ParquetType>::readBatch(UInt64 rows_num, const String & name)
 {
     reading_rows_num = rows_num;
     auto readPageIfEmpty = [&]()
@@ -251,9 +282,11 @@ ColumnWithTypeAndName ParquetLeafColReader<TColumn>::readBatch(UInt64 rows_num,
     return releaseColumn(name);
 }
 
-template <>
-void ParquetLeafColReader<ColumnString>::resetColumn(UInt64 rows_num)
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::resetColumn(UInt64 rows_num)
 {
+    if constexpr (std::is_same_v<TColumn, ColumnString>)
+    {
         if (reading_low_cardinality)
         {
             assert(dictionary);
@@ -272,20 +305,19 @@ void ParquetLeafColReader<ColumnString>::resetColumn(UInt64 rows_num)
             column = ColumnString::create();
             reserveColumnStrRows(column, rows_num);
         }
-}
-
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::resetColumn(UInt64 rows_num)
-{
+    }
+    else
+    {
         assert(!reading_low_cardinality);
 
         column = base_data_type->createColumn();
         column->reserve(rows_num);
         null_map = std::make_unique<LazyNullMap>(rows_num);
+    }
 }
 
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::degradeDictionary()
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::degradeDictionary()
 {
     // if last batch read all dictionary indices, then degrade is not needed this time
     if (!column)
@@ -331,8 +363,8 @@ void ParquetLeafColReader<TColumn>::degradeDictionary()
     LOG_DEBUG(log, "degraded dictionary to normal column");
 }
 
-template <typename TColumn>
-ColumnWithTypeAndName ParquetLeafColReader<TColumn>::releaseColumn(const String & name)
+template <typename TColumn, typename ParquetType>
+ColumnWithTypeAndName ParquetLeafColReader<TColumn, ParquetType>::releaseColumn(const String & name)
 {
     DataTypePtr data_type = base_data_type;
     if (reading_low_cardinality)
@@ -365,8 +397,8 @@ ColumnWithTypeAndName ParquetLeafColReader<TColumn>::releaseColumn(const String
     return res;
 }
 
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::readPage()
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::readPage()
 {
     // refer to: ColumnReaderImplBase::ReadNewPage in column_reader.cc
     // this is where decompression happens
@@ -408,8 +440,8 @@ void ParquetLeafColReader<TColumn>::readPage()
     }
 }
 
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::initDataReader(
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::initDataReader(
     parquet::Encoding::type enconding_type,
     const uint8_t * buffer,
     std::size_t max_size,
@@ -425,29 +457,8 @@ void ParquetLeafColReader<TColumn>::initDataReader(
                 degradeDictionary();
             }
 
-            if (col_descriptor.physical_type() == parquet::Type::BOOLEAN)
-            {
-                if constexpr (std::is_same_v<TColumn, ColumnUInt8>)
-                {
-                    auto bit_reader = std::make_unique<arrow::bit_util::BitReader>(buffer, max_size);
-                    data_values_reader = std::make_unique<ParquetBitPlainReader<ColumnUInt8>>(col_descriptor.max_definition_level(),
-                                                                                              std::move(def_level_reader),
-                                                                                              std::move(bit_reader));
-                }
-            }
-            else
-            {
-                ParquetDataBuffer parquet_buffer = [&]()
-                {
-                    if constexpr (!std::is_same_v<ColumnDecimal<DateTime64>, TColumn>)
-                        return ParquetDataBuffer(buffer, max_size);
-
-                    auto scale = assert_cast<const DataTypeDateTime64 &>(*base_data_type).getScale();
-                    return ParquetDataBuffer(buffer, max_size, scale);
-                }();
-                data_values_reader = createPlainReader<TColumn>(
-                    col_descriptor, std::move(def_level_reader), std::move(parquet_buffer));
-            }
+            data_values_reader = createReader<TColumn, ParquetType>(
+                col_descriptor, std::move(def_level_reader), buffer, max_size, base_data_type);
             break;
         }
         case parquet::Encoding::RLE_DICTIONARY:
@@ -476,8 +487,8 @@ void ParquetLeafColReader<TColumn>::initDataReader(
     }
 }
 
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::readPageV1(const parquet::DataPageV1 & page)
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::readPageV1(const parquet::DataPageV1 & page)
 {
     cur_page_values = page.num_values();
 
@@ -562,8 +573,8 @@ void ParquetLeafColReader<TColumn>::readPageV1(const parquet::DataPageV1 & page)
  * The data buffer is "offset-ed" by rl bytes length and then dl decoder is built using RLE decoder. Since dl bytes length was present in the header,
  * there is no need to read it and apply an offset like in page v1.
  * */
-template <typename TColumn>
-void ParquetLeafColReader<TColumn>::readPageV2(const parquet::DataPageV2 & page)
+template <typename TColumn, typename ParquetType>
+void ParquetLeafColReader<TColumn, ParquetType>::readPageV2(const parquet::DataPageV2 & page)
 {
     cur_page_values = page.num_values();
 
@@ -609,11 +620,19 @@ void ParquetLeafColReader<TColumn>::readPageV2(const parquet::DataPageV2 & page)
     initDataReader(page.encoding(), buffer, page.size() - total_levels_length, std::move(def_level_reader));
 }
 
-template <typename TColumn>
-std::unique_ptr<ParquetDataValuesReader> ParquetLeafColReader<TColumn>::createDictReader(
+template <typename TColumn, typename ParquetType>
+std::unique_ptr<ParquetDataValuesReader> ParquetLeafColReader<TColumn, ParquetType>::createDictReader(
     std::unique_ptr<RleValuesReader> def_level_reader, std::unique_ptr<RleValuesReader> rle_data_reader)
 {
-    if (reading_low_cardinality && std::same_as<TColumn, ColumnString>)
+    if constexpr (std::is_same_v<TColumn, ColumnUInt8> || std::is_same_v<TColumn, ColumnInt8>
+        || std::is_same_v<TColumn, ColumnUInt16> || std::is_same_v<TColumn, ColumnInt16>)
+    {
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Dictionary encoding for booleans is not supported");
+    }
+
+    if (reading_low_cardinality)
+    {
+        if constexpr (std::same_as<TColumn, ColumnString>)
         {
             std::unique_ptr<ParquetDataValuesReader> res;
             visitColStrIndexType(dictionary->size(), [&]<typename TCol>(TCol *)
@@ -625,10 +644,6 @@ std::unique_ptr<ParquetDataValuesReader> ParquetLeafColReader<TColumn>::createDi
                                  });
             return res;
         }
-
-    if (col_descriptor.physical_type() == parquet::Type::type::BOOLEAN)
-    {
-        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Dictionary encoding for booleans is not supported");
     }
 
     return std::make_unique<ParquetRleDictReader<TColumn>>(
@@ -639,19 +654,23 @@ std::unique_ptr<ParquetDataValuesReader> ParquetLeafColReader<TColumn>::createDi
 }
 
 
-template class ParquetLeafColReader<ColumnUInt8>;
-template class ParquetLeafColReader<ColumnInt32>;
-template class ParquetLeafColReader<ColumnUInt32>;
-template class ParquetLeafColReader<ColumnInt64>;
-template class ParquetLeafColReader<ColumnUInt64>;
-template class ParquetLeafColReader<ColumnBFloat16>;
-template class ParquetLeafColReader<ColumnFloat32>;
-template class ParquetLeafColReader<ColumnFloat64>;
-template class ParquetLeafColReader<ColumnString>;
-template class ParquetLeafColReader<ColumnDecimal<Decimal32>>;
-template class ParquetLeafColReader<ColumnDecimal<Decimal64>>;
-template class ParquetLeafColReader<ColumnDecimal<Decimal128>>;
-template class ParquetLeafColReader<ColumnDecimal<Decimal256>>;
-template class ParquetLeafColReader<ColumnDecimal<DateTime64>>;
+template class ParquetLeafColReader<ColumnUInt8, bool>;
+template class ParquetLeafColReader<ColumnUInt8, int32_t>;
+template class ParquetLeafColReader<ColumnInt8, int32_t>;
+template class ParquetLeafColReader<ColumnUInt16, int32_t>;
+template class ParquetLeafColReader<ColumnInt16, int32_t>;
+template class ParquetLeafColReader<ColumnUInt32, int32_t>;
+template class ParquetLeafColReader<ColumnInt32, int32_t>;
+template class ParquetLeafColReader<ColumnUInt64, int64_t>;
+template class ParquetLeafColReader<ColumnInt64, int64_t>;
+template class ParquetLeafColReader<ColumnFloat32, float>;
+template class ParquetLeafColReader<ColumnFloat64, double>;
+template class ParquetLeafColReader<ColumnString, ParquetByteArrayTypeStub>;
+template class ParquetLeafColReader<ColumnDecimal<Decimal32>, int32_t>;
+template class ParquetLeafColReader<ColumnDecimal<Decimal64>, int64_t>;
+template class ParquetLeafColReader<ColumnDecimal<Decimal128>, ParquetByteArrayTypeStub>;
+template class ParquetLeafColReader<ColumnDecimal<Decimal256>, ParquetByteArrayTypeStub>;
+template class ParquetLeafColReader<ColumnDecimal<DateTime64>, ParquetInt96TypeStub>;
+template class ParquetLeafColReader<ColumnDecimal<DateTime64>, int64_t>;
 
 }

src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.h

@@ -17,7 +17,10 @@ class ColumnDescriptor;
 namespace DB
 {
 
-template <typename TColumn>
+struct ParquetByteArrayTypeStub {};
+struct ParquetInt96TypeStub {};
+
+template <typename TColumn, typename ParquetType>
 class ParquetLeafColReader : public ParquetColumnReader
 {
 public:

src/Processors/Formats/Impl/Parquet/ParquetRecordReader.cpp

@@ -93,19 +93,20 @@ private:
     std::unique_ptr<ParquetColumnReader> fromInt32INT(const parquet::IntLogicalType & int_type);
     std::unique_ptr<ParquetColumnReader> fromInt64INT(const parquet::IntLogicalType & int_type);
 
-    template<class DataType>
+    template<class ClickHouseType, typename ParquetType>
     auto makeLeafReader()
     {
-        return std::make_unique<ParquetLeafColReader<typename DataType::ColumnType>>(
-            col_descriptor, std::make_shared<DataType>(), std::move(meta), std::move(page_reader));
+        return std::make_unique<ParquetLeafColReader<typename ClickHouseType::ColumnType, ParquetType>>(
+            col_descriptor, std::make_shared<ClickHouseType>(), std::move(meta), std::move(page_reader));
     }
 
-    template<class DecimalType>
+    template<class DecimalType, typename ParquetType>
     auto makeDecimalLeafReader()
     {
         auto data_type = std::make_shared<DataTypeDecimal<DecimalType>>(
             col_descriptor.type_precision(), col_descriptor.type_scale());
-        return std::make_unique<ParquetLeafColReader<ColumnDecimal<DecimalType>>>(
+
+        return std::make_unique<ParquetLeafColReader<ColumnDecimal<DecimalType>, ParquetType>>(
             col_descriptor, std::move(data_type), std::move(meta), std::move(page_reader));
     }
 
@@ -157,11 +158,11 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromInt32()
         case parquet::LogicalType::Type::INT:
             return fromInt32INT(dynamic_cast<const parquet::IntLogicalType &>(*col_descriptor.logical_type()));
         case parquet::LogicalType::Type::NONE:
-            return makeLeafReader<DataTypeInt32>();
+            return makeLeafReader<DataTypeInt32, int32_t>();
         case parquet::LogicalType::Type::DATE:
-            return makeLeafReader<DataTypeDate32>();
+            return makeLeafReader<DataTypeDate32, int32_t>();
         case parquet::LogicalType::Type::DECIMAL:
-            return makeDecimalLeafReader<Decimal32>();
+            return makeDecimalLeafReader<Decimal32, int32_t>();
         default:
             return throwUnsupported();
     }
@@ -174,16 +175,16 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromInt64()
         case parquet::LogicalType::Type::INT:
             return fromInt64INT(dynamic_cast<const parquet::IntLogicalType &>(*col_descriptor.logical_type()));
         case parquet::LogicalType::Type::NONE:
-            return makeLeafReader<DataTypeInt64>();
+            return makeLeafReader<DataTypeInt64, int64_t>();
         case parquet::LogicalType::Type::TIMESTAMP:
         {
             const auto & tm_type = dynamic_cast<const parquet::TimestampLogicalType &>(*col_descriptor.logical_type());
             auto read_type = std::make_shared<DataTypeDateTime64>(getScaleFromLogicalTimestamp(tm_type.time_unit()));
-            return std::make_unique<ParquetLeafColReader<ColumnDecimal<DateTime64>>>(
+            return std::make_unique<ParquetLeafColReader<ColumnDecimal<DateTime64>, int64_t>>(
                 col_descriptor, std::move(read_type), std::move(meta), std::move(page_reader));
         }
         case parquet::LogicalType::Type::DECIMAL:
-            return makeDecimalLeafReader<Decimal64>();
+            return makeDecimalLeafReader<Decimal64, int64_t>();
         default:
             return throwUnsupported();
     }
@@ -195,7 +196,7 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromByteArray()
     {
         case parquet::LogicalType::Type::STRING:
         case parquet::LogicalType::Type::NONE:
-            return makeLeafReader<DataTypeString>();
+            return makeLeafReader<DataTypeString, ParquetByteArrayTypeStub>();
         default:
             return throwUnsupported();
     }
@@ -210,9 +211,9 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromFLBA()
             if (col_descriptor.type_length() > 0)
             {
                 if (col_descriptor.type_length() <= static_cast<int>(sizeof(Decimal128)))
-                    return makeDecimalLeafReader<Decimal128>();
+                    return makeDecimalLeafReader<Decimal128, ParquetByteArrayTypeStub>();
                 if (col_descriptor.type_length() <= static_cast<int>(sizeof(Decimal256)))
-                    return makeDecimalLeafReader<Decimal256>();
+                    return makeDecimalLeafReader<Decimal256, ParquetByteArrayTypeStub>();
             }
 
             return throwUnsupported(PreformattedMessage::create(
@@ -227,11 +228,23 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromInt32INT(const parque
 {
     switch (int_type.bit_width())
     {
+        case 8:
+        {
+            if (int_type.is_signed())
+                return makeLeafReader<DataTypeInt8, int32_t>();
+            return makeLeafReader<DataTypeUInt8, int32_t>();
+        }
+        case 16:
+        {
+            if (int_type.is_signed())
+                return makeLeafReader<DataTypeInt16, int32_t>();
+            return makeLeafReader<DataTypeUInt16, int32_t>();
+        }
         case 32:
         {
             if (int_type.is_signed())
-                return makeLeafReader<DataTypeInt32>();
-            return makeLeafReader<DataTypeUInt32>();
+                return makeLeafReader<DataTypeInt32, int32_t>();
+            return makeLeafReader<DataTypeUInt32, int32_t>();
         }
         default:
             return throwUnsupported(PreformattedMessage::create(", bit width: {}", int_type.bit_width()));
@@ -245,8 +258,8 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::fromInt64INT(const parque
         case 64:
         {
             if (int_type.is_signed())
-                return makeLeafReader<DataTypeInt64>();
-            return makeLeafReader<DataTypeUInt64>();
+                return makeLeafReader<DataTypeInt64, int64_t>();
+            return makeLeafReader<DataTypeUInt64, int64_t>();
         }
         default:
             return throwUnsupported(PreformattedMessage::create(", bit width: {}", int_type.bit_width()));
@@ -263,7 +276,7 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::makeReader()
     switch (col_descriptor.physical_type())
     {
         case parquet::Type::BOOLEAN:
-            return makeLeafReader<DataTypeUInt8>();
+            return makeLeafReader<DataTypeUInt8, bool>();
         case parquet::Type::INT32:
             return fromInt32();
         case parquet::Type::INT64:
@@ -276,13 +289,13 @@ std::unique_ptr<ParquetColumnReader> ColReaderFactory::makeReader()
                 auto scale = getScaleFromArrowTimeUnit(arrow_properties.coerce_int96_timestamp_unit());
                 read_type = std::make_shared<DataTypeDateTime64>(scale);
             }
-            return std::make_unique<ParquetLeafColReader<ColumnDecimal<DateTime64>>>(
+            return std::make_unique<ParquetLeafColReader<ColumnDecimal<DateTime64>, ParquetInt96TypeStub>>(
                 col_descriptor, read_type, std::move(meta), std::move(page_reader));
         }
         case parquet::Type::FLOAT:
-            return makeLeafReader<DataTypeFloat32>();
+            return makeLeafReader<DataTypeFloat32, float>();
         case parquet::Type::DOUBLE:
-            return makeLeafReader<DataTypeFloat64>();
+            return makeLeafReader<DataTypeFloat64, double>();
         case parquet::Type::BYTE_ARRAY:
             return fromByteArray();
         case parquet::Type::FIXED_LEN_BYTE_ARRAY:

tests/ci/official_docker.py

@@ -299,8 +299,6 @@ class TagAttrs:
 
     # Only one latest can exist
     latest: ClickHouseVersion
-    # Only one can be a major one (the most fresh per a year)
-    majors: Dict[int, ClickHouseVersion]
     # Only one lts version can exist
     lts: Optional[ClickHouseVersion]
 
@@ -345,14 +343,6 @@ def ldf_tags(version: ClickHouseVersion, distro: str, tag_attrs: TagAttrs) -> st
             tags.append("lts")
         tags.append(f"lts-{distro}")
 
-    # If the tag `22`, `23`, `24` etc. should be included in the tags
-    with_major = tag_attrs.majors.get(version.major) in (None, version)
-    if with_major:
-        tag_attrs.majors[version.major] = version
-        if without_distro:
-            tags.append(f"{version.major}")
-        tags.append(f"{version.major}-{distro}")
-
     # Add all normal tags
     for tag in (
         f"{version.major}.{version.minor}",
@@ -384,7 +374,7 @@ def generate_ldf(args: argparse.Namespace) -> None:
         args.directory / git_runner(f"git -C {args.directory} rev-parse --show-cdup")
     ).absolute()
     lines = ldf_header(git, directory)
-    tag_attrs = TagAttrs(versions[-1], {}, None)
+    tag_attrs = TagAttrs(versions[-1], None)
 
     # We iterate from the most recent to the oldest version
     for version in reversed(versions):

tests/integration/test_server_reload/test.py

@@ -378,7 +378,7 @@ def test_reload_via_client(cluster, zk):
                         configure_from_zk(zk)
                     break
                 except QueryRuntimeException:
-                    logging.exception("The new socket is not binded yet")
+                    logging.exception("The new socket is not bound yet")
                     time.sleep(0.1)
 
     if exception:

tests/queries/0_stateless/03250_ephemeral_comment.sql

@@ -0,0 +1,11 @@
+drop table if exists test;
+CREATE TABLE test (
+    `start_s`  UInt32 EPHEMERAL COMMENT 'start UNIX time' ,
+    `start_us` UInt16 EPHEMERAL COMMENT 'start microseconds',
+    `finish_s`  UInt32 EPHEMERAL COMMENT 'finish UNIX time',
+    `finish_us` UInt16 EPHEMERAL COMMENT 'finish microseconds',
+    `captured` DateTime MATERIALIZED fromUnixTimestamp(start_s),
+    `duration` Decimal32(6) MATERIALIZED finish_s - start_s + (finish_us - start_us)/1000000
+)
+ENGINE Null;
+drop table if exists test;

tests/queries/0_stateless/03262_test_parquet_native_reader_int_logical_type.reference

@@ -0,0 +1,40 @@
+-94	53304	17815465730223871
+57	15888	33652524900575246
+-4	14877	53832092832965652
+33	3387	86326601511136103
+104	3383	115438187156564782
+-11	37403	145056169255259589
+-72	46473	159324626361233509
+103	35510	173644182696185097
+-26	60902	185175917734318892
+70	48767	193167023342307884
+2	21648	247953090704786001
+20	2986	268127160817221407
+76	20277	290178827409195337
+61	28692	305149163504092270
+-74	65427	326871531363668398
+-15	20256	351812901947846888
+-39	65472	357371822264135234
+79	38671	371605113770958364
+-29	41706	394460710549666968
+92	25026	412913269933311543
+-94	53304	17815465730223871
+57	15888	33652524900575246
+-4	14877	53832092832965652
+33	3387	86326601511136103
+104	3383	115438187156564782
+-11	37403	145056169255259589
+-72	46473	159324626361233509
+103	35510	173644182696185097
+-26	60902	185175917734318892
+70	48767	193167023342307884
+2	21648	247953090704786001
+20	2986	268127160817221407
+76	20277	290178827409195337
+61	28692	305149163504092270
+-74	65427	326871531363668398
+-15	20256	351812901947846888
+-39	65472	357371822264135234
+79	38671	371605113770958364
+-29	41706	394460710549666968
+92	25026	412913269933311543

tests/queries/0_stateless/03262_test_parquet_native_reader_int_logical_type.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+# Tags: no-ubsan, no-fasttest
+
+CUR_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CUR_DIR"/../shell_config.sh
+
+
+USER_FILES_PATH=$($CLICKHOUSE_CLIENT_BINARY --query "select _path,_file from file('nonexist.txt', 'CSV', 'val1 char')" 2>&1 | grep Exception | awk '{gsub("/nonexist.txt","",$9); print $9}')
+
+WORKING_DIR="${USER_FILES_PATH}/${CLICKHOUSE_TEST_UNIQUE_NAME}"
+
+mkdir -p "${WORKING_DIR}"
+
+DATA_FILE="${CUR_DIR}/data_parquet/multi_column_bf.gz.parquet"
+
+DATA_FILE_USER_PATH="${WORKING_DIR}/multi_column_bf.gz.parquet"
+
+cp ${DATA_FILE} ${DATA_FILE_USER_PATH}
+
+${CLICKHOUSE_CLIENT} --query="select int8_logical, uint16_logical, uint64_logical from file('${DATA_FILE_USER_PATH}', Parquet) order by uint64_logical limit 20 SETTINGS input_format_parquet_use_native_reader=false;";
+${CLICKHOUSE_CLIENT} --query="select int8_logical, uint16_logical, uint64_logical from file('${DATA_FILE_USER_PATH}', Parquet) order by uint64_logical limit 20 SETTINGS input_format_parquet_use_native_reader=true;";