[BugFix] fix using incorrect read ratio

Signed-off-by: stephen <[email protected]>
StarRocks · Jan 17, 2025 · e394a1b · e394a1b
1 parent 75aa13e
commit e394a1b
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 30 deletions.
diff --git a/fe/fe-core/src/main/java/com/starrocks/statistic/base/PartitionSampler.java b/fe/fe-core/src/main/java/com/starrocks/statistic/base/PartitionSampler.java
@@ -27,10 +27,10 @@
 import java.util.Map;
 
 public class PartitionSampler {
-    private static final double HIGH_WEIGHT_READ_RATIO = 0.001;
-    private static final double MEDIUM_HIGH_WEIGHT_READ_RATIO = 0.01;
-    private static final double MEDIUM_LOW_WEIGHT_READ_RATIO = 0.1;
-    private static final double LOW_WEIGHT_READ_RATIO = 0.8;
+    public static final double HIGH_WEIGHT_READ_RATIO = 0.001;
+    public static final double MEDIUM_HIGH_WEIGHT_READ_RATIO = 0.01;
+    public static final double MEDIUM_LOW_WEIGHT_READ_RATIO = 0.1;
+    public static final double LOW_WEIGHT_READ_RATIO = 0.8;
     private static final long HIGH_WEIGHT_ROWS_THRESHOLD = 10000000L;
     private static final long MEDIUM_HIGH_WEIGHT_ROWS_THRESHOLD = 1000000L;
     private static final long MEDIUM_LOW_WEIGHT_ROWS_THRESHOLD = 100000L;
@@ -56,26 +56,6 @@ public PartitionSampler(double highSampleRatio, double mediumHighRatio, double m
         this.sampleRowsLimit = sampleRowLimit;
     }
 
-    public double getHighRatio() {
-        return highRatio;
-    }
-
-    public double getMediumHighRatio() {
-        return mediumHighRatio;
-    }
-
-    public double getMediumLowRatio() {
-        return mediumLowRatio;
-    }
-
-    public double getLowRatio() {
-        return lowRatio;
-    }
-
-    public int getMaxSize() {
-        return maxSize;
-    }
-
     public long getSampleRowsLimit() {
         return sampleRowsLimit;
     }

diff --git a/fe/fe-core/src/main/java/com/starrocks/statistic/hyper/HyperStatisticSQLs.java b/fe/fe-core/src/main/java/com/starrocks/statistic/hyper/HyperStatisticSQLs.java
@@ -32,6 +32,11 @@
 import java.util.Objects;
 import java.util.stream.Collectors;
 
+import static com.starrocks.statistic.base.PartitionSampler.HIGH_WEIGHT_READ_RATIO;
+import static com.starrocks.statistic.base.PartitionSampler.LOW_WEIGHT_READ_RATIO;
+import static com.starrocks.statistic.base.PartitionSampler.MEDIUM_HIGH_WEIGHT_READ_RATIO;
+import static com.starrocks.statistic.base.PartitionSampler.MEDIUM_LOW_WEIGHT_READ_RATIO;
+
 public class HyperStatisticSQLs {
     private static final VelocityEngine DEFAULT_VELOCITY_ENGINE;
 
@@ -131,13 +136,13 @@ public static String buildSampleSQL(Database db, Table table, Partition p, List<
         List<String> groupSQLs = Lists.newArrayList();
         StringBuilder sqlBuilder = new StringBuilder();
         groupSQLs.add(generateRatioTable(tableName, sampler.getSampleRowsLimit(), info.getHighWeightTablets(),
-                sampler.getHighRatio(), "t_high"));
+                HIGH_WEIGHT_READ_RATIO, "t_high"));
         groupSQLs.add(generateRatioTable(tableName, sampler.getSampleRowsLimit(), info.getMediumHighWeightTablets(),
-                sampler.getMediumHighRatio(), "t_medium_high"));
+                MEDIUM_HIGH_WEIGHT_READ_RATIO, "t_medium_high"));
         groupSQLs.add(generateRatioTable(tableName, sampler.getSampleRowsLimit(), info.getMediumLowWeightTablets(),
-                sampler.getMediumLowRatio(), "t_medium_low"));
+                MEDIUM_LOW_WEIGHT_READ_RATIO, "t_medium_low"));
         groupSQLs.add(generateRatioTable(tableName, sampler.getSampleRowsLimit(), info.getLowWeightTablets(),
-                sampler.getLowRatio(), "t_low"));
+                LOW_WEIGHT_READ_RATIO, "t_low"));
         if (groupSQLs.stream().allMatch(Objects::isNull)) {
             groupSQLs.add("SELECT * FROM " + tableName + " LIMIT " + Config.statistic_sample_collect_rows);
         }

diff --git a/fe/fe-core/src/test/java/com/starrocks/statistic/hyper/HyperJobTest.java b/fe/fe-core/src/test/java/com/starrocks/statistic/hyper/HyperJobTest.java
@@ -43,6 +43,8 @@
 import com.starrocks.statistic.base.PartitionSampler;
 import com.starrocks.statistic.base.PrimitiveTypeColumnStats;
 import com.starrocks.statistic.base.SubFieldColumnStats;
+import com.starrocks.statistic.sample.SampleInfo;
+import com.starrocks.statistic.sample.TabletStats;
 import com.starrocks.utframe.StarRocksAssert;
 import mockit.Mock;
 import mockit.MockUp;
@@ -191,6 +193,13 @@ public void testFullJobs() {
     public void testSampleJobs() {
         Pair<List<String>, List<Type>> pair = initColumn(List.of("c1", "c2", "c3"));
 
+        new MockUp<SampleInfo>() {
+            @Mock
+            public List<TabletStats> getMediumHighWeightTablets() {
+                return List.of(new TabletStats(1, pid, 5000000));
+            }
+        };
+
         List<HyperQueryJob> jobs = HyperQueryJob.createSampleQueryJobs(connectContext, db, table, pair.first,
                 pair.second, List.of(pid), 1, sampler);
 
@@ -201,8 +210,8 @@ public void testSampleJobs() {
         List<String> sql = jobs.get(1).buildQuerySQL();
         Assert.assertEquals(1, sql.size());
 
-        assertContains(sql.get(0), "with base_cte_table as " +
-                "(SELECT * FROM `test`.`t_struct` LIMIT 200000) ");
+        assertContains(sql.get(0), "with base_cte_table as ( SELECT * FROM (SELECT *  FROM `test`.`t_struct` TABLET(1)" +
+                " SAMPLE('percent'='1')  LIMIT 200000)");
         assertContains(sql.get(0), "cast(IFNULL(SUM(CHAR_LENGTH(`c2`)) * 0/ COUNT(*), 0) as BIGINT), " +
                 "hex(hll_serialize(IFNULL(hll_raw(`c2`), hll_empty())))," +
                 " cast((COUNT(*) - COUNT(`c2`)) * 0 / COUNT(*) as BIGINT), " +