Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: edit default values for hive.stats.deserialization.factor

...

  • Default Value: 24
  • Added In: Hive 0.13 with HIVE-6229

Reserved length for postfix of statistics key. Currently only meaningful for counter type statistics which should keep the length of the full statistics key smaller than the maximum length configured by hive.stats.key.prefix.max.length. For counter type statistics, it should be bigger than the length of LB spec if exists.

hive.stats.max.variable.length
  • Default Value: 100
  • Added In: Hive 0.13 with HIVE-5369

To estimate the size of data flowing through operators in Hive/Tez (for reducer estimation etc.), average row size is multiplied with the total number of rows coming out of each operator. Average row size is computed from average column size of all columns in the row. In the absence of column statistics, for variable length columns (like string, bytes, etc.) this value will be used. For fixed length columns their corresponding Java equivalent sizes are used (float – 4 bytes, double – 8 bytes, etc.).

hive.analyze.stmt.collect.partlevel.stats
  • Default Value: true
  • Added In: Hive 0.14.0 with HIVE-7609

Prior to 0.14, on partitioned table, analyze statement used to collect table level statistics when no partition is specified. That behavior has changed beginning 0.14 to instead collect partition level statistics for all partitions. If old behavior of collecting aggregated table level statistics is desired, change the value of this config to false. This impacts only column statistics. Basic statistics are not impacted by this config.

hive.stats.list.num.entries
  • Default Value: 10
  • Added In: Hive 0.13 with HIVE-5369

To estimate the size of data flowing through operators in Hive/Tez (for reducer estimation etc.), average row size is multiplied with the total number of rows coming out of each operator. Average row size is computed from average column size of all columns in the row. In the absence of column statistics and for variable length complex columns like list, the average number of entries/values can be specified using this configuration property.

hive.stats.map.num.entries
  • Default Value: 10
  • Added In: Hive 0.13 with HIVE-5369

To estimate the size of data flowing through operators in Hive/Tez (for reducer estimation etc.), average row size is multiplied with the total number of rows coming out of each operator. Average row size is computed from average column size of all columns in the row. In the absence of column statistics and for variable length complex columns like map, the average number of entries/values can be specified using this configuration property.

hive.stats.map.parallelism

The Hive/Tez optimizer estimates the data size flowing through each of the operators. For the GROUPBY operator, to accurately compute the data size map-side parallelism needs to be known. By default, this value is set to 1 since the optimizer is not aware of the number of mappers during compile-time. This Hive configuration property can be used to specify the number of mappers for data size computation of the GROUPBY operator. (This configuration property was removed in release 0.14.0.)

hive.stats.fetch.partition.stats
  • Default Value: true
  • Added In: Hive 0.13 with HIVE-6298
  • Removed In: Hive 3.0.0 with HIVE-17932

Annotation of the operator tree with statistics information requires partition level basic statistics like number of rows, data size and file size. Partition statistics are fetched from the metastore. Fetching partition statistics for each needed partition can be expensive when the number of partitions is high. This flag can be used to disable fetching of partition statistics from the metastore. When this flag is disabled, Hive will make calls to the filesystem to get file sizes and will estimate the number of rows from the row schema.

hive.stats.fetch.column.stats
  • Default Value: false
  • Added In: Hive 0.13 with HIVE-5898

Annotation of the operator tree with statistics information requires column statistics. Column statistics are fetched from the metastore. Fetching column statistics for each needed column can be expensive when the number of columns is high. This flag can be used to disable fetching of column statistics from the metastore.

hive.stats.join.factor
  • Default Value: (float) 1.1
  • Added In: Hive 0.13 with HIVE-5921

The Hive/Tez optimizer estimates the data size flowing through each of the operators. The JOIN operator uses column statistics to estimate the number of rows flowing out of it and hence the data size. In the absence of column statistics, this factor determines the amount of rows flowing out of the JOIN operator.

hive.stats.deserialization.factor
  • Default Value:
    • Hive 0.13 to 2.x.x: (float) 1.0
    • Hive 3.0.0 and later: (float) 10.0
  • Added In: Hive 0.13 with HIVE-5921
  • Default value changed from 1.0 to 10.0 in Hive 3.0

The Hive/Tez optimizer estimates the data size flowing through each of the operators. In the absence of basic statistics like number of rows and data size, file size is used to estimate the number of rows and data size. Since files in tables/partitions are serialized (and optionally compressed) the estimates of number of rows and data size cannot be reliably determined. This factor is multiplied with the file size to account for serialization and compression.

hive.stats.avg.row.size
  • Default Value: 10000
  • Added In: Hive 0.13 with HIVE-5921

In the absence of table/partition statistics, average row size will be used to estimate the number of rows/data size.

...