Cosmetics II

docs/en/engines/table-engines/mergetree-family/mergetree.md

@@ -1005,7 +1005,7 @@ They can be used for prewhere optimization only if we enable `set allow_statisti
 
 ## Column-level Settings {#column-level-settings}
 
-Certain MergeTree settings can be override at column level:
+Certain MergeTree settings can be overridden at column level:
 
 - `max_compress_block_size` — Maximum size of blocks of uncompressed data before compressing for writing to a table.
 - `min_compress_block_size` — Minimum size of blocks of uncompressed data required for compression when writing the next mark.

src/Storages/Statistics/Statistics.cpp

@@ -89,15 +89,17 @@ Float64 IStatistics::estimateLess(const Field & /*val*/) const
     throw Exception(ErrorCodes::LOGICAL_ERROR, "Less-than estimation is not implemented for this type of statistics");
 }
 
-/// -------------------------------------
-/// Implementation of the estimation:
-/// Note: Each statistics object supports certain types predicates natively, e.g.
-/// - TDigest: '< X' (less-than predicates)
-/// - Count-min sketches: '= X' (equal predicates)
-/// - Uniq (HyperLogLog): 'count distinct(*)' (column cardinality)
-/// If multiple statistics objects are available per column, it is sometimes also possible to combine them in a clever way.
-/// For that reason, all estimation are performed in a central place (here), and we don't simply pass the predicate to the first statistics
-/// object that supports it natively.
+/// Notes:
+/// - Statistics object usually only support estimation for certain types of predicates, e.g.
+///    - TDigest: '< X' (less-than predicates)
+///    - Count-min sketches: '= X' (equal predicates)
+///    - Uniq (HyperLogLog): 'count distinct(*)' (column cardinality)
+///
+/// If multiple statistics objects in a column support estimating a predicate, we want to try statistics in order of descending accuracy
+/// (e.g. MinMax statistics are simpler than TDigest statistics and thus worse for estimating 'less' predicates).
+///
+/// Sometimes, it is possible to combine multiple statistics in a clever way. For that reason, all estimation are performed in a central
+/// place (here), and we don't simply pass the predicate to the first statistics object that supports it natively.
 
 Float64 ColumnStatistics::estimateLess(const Field & val) const
 {

src/Storages/Statistics/StatisticsCountMinSketch.cpp

@@ -25,8 +25,8 @@ extern const int ILLEGAL_STATISTICS;
 static constexpr auto num_hashes = 7uz;
 static constexpr auto num_buckets = 2718uz;
 
-StatisticsCountMinSketch::StatisticsCountMinSketch(const SingleStatisticsDescription & stat_, DataTypePtr data_type_)
-    : IStatistics(stat_)
+StatisticsCountMinSketch::StatisticsCountMinSketch(const SingleStatisticsDescription & description, DataTypePtr data_type_)
+    : IStatistics(description)
     , sketch(num_hashes, num_buckets)
     , data_type(data_type_)
 {
@@ -84,7 +84,7 @@ void StatisticsCountMinSketch::deserialize(ReadBuffer & buf)
 }
 
 
-void countMinSketchStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type)
+void countMinSketchStatisticsValidator(const SingleStatisticsDescription & /*description*/, DataTypePtr data_type)
 {
     data_type = removeNullable(data_type);
     data_type = removeLowCardinalityAndNullable(data_type);
@@ -92,9 +92,9 @@ void countMinSketchStatisticsValidator(const SingleStatisticsDescription &, Data
         throw Exception(ErrorCodes::ILLEGAL_STATISTICS, "Statistics of type 'count_min' does not support type {}", data_type->getName());
 }
 
-StatisticsPtr countMinSketchStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr data_type)
+StatisticsPtr countMinSketchStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr data_type)
 {
-    return std::make_shared<StatisticsCountMinSketch>(stat, data_type);
+    return std::make_shared<StatisticsCountMinSketch>(description, data_type);
 }
 
 }

src/Storages/Statistics/StatisticsCountMinSketch.h

@@ -14,7 +14,7 @@ namespace DB
 class StatisticsCountMinSketch : public IStatistics
 {
 public:
-    StatisticsCountMinSketch(const SingleStatisticsDescription & stat_, DataTypePtr data_type_);
+    StatisticsCountMinSketch(const SingleStatisticsDescription & description, DataTypePtr data_type_);
 
     Float64 estimateEqual(const Field & val) const override;
 
@@ -31,8 +31,8 @@ private:
 };
 
 
-void countMinSketchStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type);
-StatisticsPtr countMinSketchStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr);
+void countMinSketchStatisticsValidator(const SingleStatisticsDescription & description, DataTypePtr data_type);
+StatisticsPtr countMinSketchStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr data_type);
 
 }
 

src/Storages/Statistics/StatisticsTDigest.cpp

@@ -10,8 +10,8 @@ extern const int ILLEGAL_STATISTICS;
 extern const int LOGICAL_ERROR;
 }
 
-StatisticsTDigest::StatisticsTDigest(const SingleStatisticsDescription & stat_)
-    : IStatistics(stat_)
+StatisticsTDigest::StatisticsTDigest(const SingleStatisticsDescription & description)
+    : IStatistics(description)
 {
 }
 
@@ -57,7 +57,7 @@ Float64 StatisticsTDigest::estimateEqual(const Field & val) const
     throw Exception(ErrorCodes::LOGICAL_ERROR, "Statistics 'tdigest' does not support estimating value of type {}", val.getTypeName());
 }
 
-void tdigestStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type)
+void tdigestStatisticsValidator(const SingleStatisticsDescription & /*description*/, DataTypePtr data_type)
 {
     data_type = removeNullable(data_type);
     data_type = removeLowCardinalityAndNullable(data_type);
@@ -65,9 +65,9 @@ void tdigestStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr
         throw Exception(ErrorCodes::ILLEGAL_STATISTICS, "Statistics of type 'tdigest' do not support type {}", data_type->getName());
 }
 
-StatisticsPtr tdigestStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr)
+StatisticsPtr tdigestStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr /*data_type*/)
 {
-    return std::make_shared<StatisticsTDigest>(stat);
+    return std::make_shared<StatisticsTDigest>(description);
 }
 
 }

src/Storages/Statistics/StatisticsTDigest.h

@@ -9,7 +9,7 @@ namespace DB
 class StatisticsTDigest : public IStatistics
 {
 public:
-    explicit StatisticsTDigest(const SingleStatisticsDescription & stat_);
+    explicit StatisticsTDigest(const SingleStatisticsDescription & description);
 
     void update(const ColumnPtr & column) override;
 
@@ -23,7 +23,7 @@ private:
     QuantileTDigest<Float64> t_digest;
 };
 
-void tdigestStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type);
-StatisticsPtr tdigestStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr);
+void tdigestStatisticsValidator(const SingleStatisticsDescription & description, DataTypePtr data_type);
+StatisticsPtr tdigestStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr data_type);
 
 }

src/Storages/Statistics/StatisticsUniq.cpp

@@ -11,8 +11,8 @@ namespace ErrorCodes
     extern const int ILLEGAL_STATISTICS;
 }
 
-StatisticsUniq::StatisticsUniq(const SingleStatisticsDescription & stat_, const DataTypePtr & data_type)
-    : IStatistics(stat_)
+StatisticsUniq::StatisticsUniq(const SingleStatisticsDescription & description, const DataTypePtr & data_type)
+    : IStatistics(description)
 {
     arena = std::make_unique<Arena>();
     AggregateFunctionProperties properties;
@@ -52,7 +52,7 @@ UInt64 StatisticsUniq::estimateCardinality() const
     return column->getUInt(0);
 }
 
-void uniqStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type)
+void uniqStatisticsValidator(const SingleStatisticsDescription & /*description*/, DataTypePtr data_type)
 {
     data_type = removeNullable(data_type);
     data_type = removeLowCardinalityAndNullable(data_type);
@@ -60,9 +60,9 @@ void uniqStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr da
         throw Exception(ErrorCodes::ILLEGAL_STATISTICS, "Statistics of type 'uniq' do not support type {}", data_type->getName());
 }
 
-StatisticsPtr uniqStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr data_type)
+StatisticsPtr uniqStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr data_type)
 {
-    return std::make_shared<StatisticsUniq>(stat, data_type);
+    return std::make_shared<StatisticsUniq>(description, data_type);
 }
 
 }

src/Storages/Statistics/StatisticsUniq.h

@@ -10,7 +10,7 @@ namespace DB
 class StatisticsUniq : public IStatistics
 {
 public:
-    StatisticsUniq(const SingleStatisticsDescription & stat_, const DataTypePtr & data_type);
+    StatisticsUniq(const SingleStatisticsDescription & description, const DataTypePtr & data_type);
     ~StatisticsUniq() override;
 
     void update(const ColumnPtr & column) override;
@@ -27,7 +27,7 @@ private:
 
 };
 
-void uniqStatisticsValidator(const SingleStatisticsDescription &, DataTypePtr data_type);
-StatisticsPtr uniqStatisticsCreator(const SingleStatisticsDescription & stat, DataTypePtr data_type);
+void uniqStatisticsValidator(const SingleStatisticsDescription & description, DataTypePtr data_type);
+StatisticsPtr uniqStatisticsCreator(const SingleStatisticsDescription & description, DataTypePtr data_type);
 
 }