Aggregation functions
AVG
¶
Since: 0.6.0
1 

Stream, Table
Return the average value for a given column.
COLLECT_LIST
¶
Since: 
1 

Stream, Table
Return an array containing all the values of col1
from each
input row (for the specified grouping and time window, if any).
Currently only works for simple types (not Map, Array, or Struct).
This version limits the size of the result Array to a maximum of
1000 entries and any values beyond this limit are silently ignored.
When using with a window type of session
, it can sometimes
happen that two session windows get merged together into one when a
outoforder record with a timestamp between the two windows is
processed. In this case the 1000 record limit is calculated by
first considering all the records from the first window, then the
outoforder record, then the records from the second window in
the order they were originally processed.
COLLECT_SET
¶
Since: 
1 

Stream
Return an array containing the distinct values of col1
from
each input row (for the specified grouping and time window, if any).
Currently only works for simple types (not Map, Array, or Struct).
This version limits the size of the result Array to a maximum of
1000 entries and any values beyond this limit are silently ignored.
When using with a window type of session
, it can sometimes
happen that two session windows get merged together into one when a
outoforder record with a timestamp between the two windows is
processed. In this case the 1000 record limit is calculated by
first considering all the records from the first window, then the
outoforder record, then the records from the second window in
the order they were originally processed.
COUNT
¶
Since: 
1 

1 

Stream, Table
Count the number of rows. When col1
is specified, the count
returned will be the number of rows where col1
is nonnull.
When *
is specified, the count returned will be the total
number of rows.
COUNT_DISTINCT
¶
Since: 0.7.0
1 

Stream, Table
Returns the approximate number of unique values of col1
in a group.
The function implementation uses HyperLogLog
to estimate cardinalities of 10^9 with a typical standard error of 2%.
EARLIEST_BY_OFFSET
¶
Since: 0.10.0
1 

Stream
Return the earliest value for a given column. Earliest here is defined as the value in the partition
with the lowest offset. Rows that have col1
set to null are ignored.
HISTOGRAM
¶
Since: 
1 

Stream, Table
Return a map containing the distinct String values of col1
mapped to the number of times each one occurs for the given window.
This version limits the number of distinct values which can be
counted to 1000, beyond which any additional entries are ignored.
When using with a window type of session
, it can sometimes
happen that two session windows get merged together into one when a
outoforder record with a timestamp between the two windows is
processed. In this case the 1000 record limit is calculated by
first considering all the records from the first window, then the
outoforder record, then the records from the second window in
the order they were originally processed.
LATEST_BY_OFFSET
¶
Since: 0.8.0
1 

Stream
Return the latest value for a given column. Latest here is defined as the value in the partition
with the greatest offset. Rows that have col1
set to null are ignored.
MAX
¶
Since: 
1 

Stream
Return the maximum value for a given column and window.
Rows that have col1
set to null are ignored.
MIN
¶
Since: 
1 

Stream
Return the minimum value for a given column and window.
Rows that have col1
set to null are ignored.
SUM
¶
Since: 
1 

Stream, Table
Sums the column values.
Rows that have col1
set to null are ignored.
TOPK
¶
Since: 
1 

Stream
Return the Top K values for the given column and window
Rows that have col1
set to null are ignored.
TOPKDISTINCT
¶
Since: 
1 

Stream
Return the distinct Top K values for the given column and window
Rows that have col1
set to null are ignored.