- Release Notes
- Introduction to CelerData Cloud Serverless
- Quick Start
- Sign up for CelerData Cloud Serverless
- A quick tour of the console
- Connect to CelerData Cloud Serverless
- Create an IAM integration
- Create and assign a warehouse
- Create an external catalog
- Load data from cloud storage
- Load data from Apache Kafka/Confluent Cloud
- Try your first query
- Invite new users
- Design data access control policy
- Warehouses
- Catalog, database, table, view, and MV
- Overview of database objects
- Catalog
- Table types
- Asynchronous materialized views
- Data Loading
- Data access control
- Networking and private connectivity
- Usage and Billing
- Organization and Account
- Integration
- Query Acceleration
- Reference
- AWS IAM policies
- Information Schema
- Overview
- be_bvars
- be_cloud_native_compactions
- be_compactions
- character_sets
- collations
- column_privileges
- columns
- engines
- events
- global_variables
- key_column_usage
- load_tracking_logs
- loads
- materialized_views
- partitions
- pipe_files
- pipes
- referential_constraints
- routines
- schema_privileges
- schemata
- session_variables
- statistics
- table_constraints
- table_privileges
- tables
- tables_config
- task_runs
- tasks
- triggers
- user_privileges
- views
- Data Types
- System Metadatabase
- Keywords
- SQL Statements
- Account Management
- Data Definition
- CREATE TABLE
- ALTER TABLE
- DROP CATALOG
- CREATE TABLE LIKE
- REFRESH EXTERNAL TABLE
- RESTORE
- SET CATALOG
- DROP TABLE
- RECOVER
- USE
- CREATE MATERIALIZED VIEW
- DROP DATABASE
- ALTER MATERIALIZED VIEW
- DROP REPOSITORY
- CANCEL RESTORE
- DROP INDEX
- DROP MATERIALIZED VIEW
- CREATE DATABASE
- CREATE TABLE AS SELECT
- BACKUP
- CANCEL BACKUP
- CREATE REPOSITORY
- CREATE INDEX
- Data Manipulation
- INSERT
- SHOW CREATE DATABASE
- SHOW BACKUP
- SHOW ALTER MATERIALIZED VIEW
- SHOW CATALOGS
- SHOW CREATE MATERIALIZED VIEW
- SELECT
- SHOW ALTER
- SHOW MATERIALIZED VIEW
- RESUME ROUTINE LOAD
- ALTER ROUTINE LOAD
- SHOW TABLES
- STREAM LOAD
- SHOW PARTITIONS
- CANCEL REFRESH MATERIALIZED VIEW
- SHOW CREATE CATALOG
- SHOW ROUTINE LOAD TASK
- SHOW RESTORE
- CREATE ROUTINE LOAD
- STOP ROUTINE LOAD
- SHOW DATABASES
- BROKER LOAD
- SHOW ROUTINE LOAD
- PAUSE ROUTINE LOAD
- SHOW SNAPSHOT
- SHOW CREATE TABLE
- CANCEL LOAD
- REFRESH MATERIALIZED VIEW
- SHOW REPOSITORIES
- SHOW LOAD
- Administration
- DESCRIBE
- SQL Functions
- Function List
- String Functions
- CONCAT
- HEX
- LOWER
- SPLIT
- LPAD
- SUBSTRING
- PARSE_URL
- INSTR
- REPEAT
- LCASE
- REPLACE
- HEX_DECODE_BINARY
- RPAD
- SPLIT_PART
- STRCMP
- SPACE
- CHARACTER_LENGTH
- URL_ENCODE
- APPEND_TAILING_CHAR_IF_ABSENT
- LTRIM
- HEX_DECODE_STRING
- URL_DECODE
- LEFT
- STARTS_WITH
- CONCAT
- GROUP_CONCAT
- STR_TO_MAP
- STRLEFT
- STRRIGHT
- MONEY_FORMAT
- RIGHT
- SUBSTRING_INDEX
- UCASE
- TRIM
- FIND_IN_SET
- RTRIM
- ASCII
- UPPER
- REVERSE
- LENGTH
- UNHEX
- ENDS_WITH
- CHAR_LENGTH
- NULL_OR_EMPTY
- LOCATE
- CHAR
- Predicate Functions
- Map Functions
- Binary Functions
- Geospatial Functions
- Lambda Expression
- Utility Functions
- Bitmap Functions
- BITMAP_SUBSET_LIMIT
- TO_BITMAP
- BITMAP_AGG
- BITMAP_FROM_STRING
- BITMAP_OR
- BITMAP_REMOVE
- BITMAP_AND
- BITMAP_TO_BASE64
- BITMAP_MIN
- BITMAP_CONTAINS
- SUB_BITMAP
- BITMAP_UNION
- BITMAP_COUNT
- BITMAP_UNION_INT
- BITMAP_XOR
- BITMAP_UNION_COUNT
- BITMAP_HAS_ANY
- BITMAP_INTERSECT
- BITMAP_AND_NOT
- BITMAP_TO_STRING
- BITMAP_HASH
- INTERSECT_COUNT
- BITMAP_EMPTY
- BITMAP_MAX
- BASE64_TO_ARRAY
- BITMAP_TO_ARRAY
- Struct Functions
- Aggregate Functions
- RETENTION
- MI
- MULTI_DISTINCT_SUM
- WINDOW_FUNNEL
- STDDEV_SAMP
- GROUPING_ID
- HLL_HASH
- AVG
- HLL_UNION_AGG
- COUNT
- BITMAP
- HLL_EMPTY
- SUM
- MAX_BY
- PERCENTILE_CONT
- COVAR_POP
- PERCENTILE_APPROX
- HLL_RAW_AGG
- STDDEV
- CORR
- COVAR_SAMP
- MIN_BY
- MAX
- VAR_SAMP
- STD
- HLL_UNION
- APPROX_COUNT_DISTINCT
- MULTI_DISTINCT_COUNT
- VARIANCE
- ANY_VALUE
- COUNT_IF
- GROUPING
- PERCENTILE_DISC
- Array Functions
- ARRAY_CUM_SUM
- ARRAY_MAX
- ARRAY_LENGTH
- ARRAY_REMOVE
- UNNEST
- ARRAY_SLICE
- ALL_MATCH
- ARRAY_CONCAT
- ARRAY_SORT
- ARRAY_POSITION
- ARRAY_DIFFERENCE
- ARRAY_CONTAINS
- ARRAY_JOIN
- ARRAY_INTERSECT
- CARDINALITY
- ARRAY_CONTAINS_ALL
- ARRAYS_OVERLAP
- ARRAY_MIN
- ARRAY_MAP
- ELEMENT_AT
- ARRAY_APPEND
- ARRAY_SORTBY
- ARRAY_TO_BITMAP
- ARRAY_GENERATE
- ARRAY_AVG
- ARRAY_FILTER
- ANY_MATCH
- REVERSE
- ARRAY_AGG
- ARRAY_DISTINCT
- ARRAY_SUM
- Condition Functions
- Math Functions
- Date and Time Functions
- DAYNAME
- MINUTE
- FROM_UNIXTIME
- HOUR
- MONTHNAME
- MONTHS_ADD
- ADD_MONTHS
- DATE_SUB
- PREVIOUS_DAY
- TO_TERA_DATA
- MINUTES_SUB
- WEEKS_ADD
- HOURS_DIFF
- UNIX_TIMESTAMP
- DAY
- DATE_SLICE
- DATE
- CURTIME
- SECONDS_SUB
- MONTH
- WEEK
- TO_DATE
- TIMEDIFF
- MONTHS_DIFF
- STR_TO_JODATIME
- WEEK_ISO
- MICROSECONDS_SUB
- TIME_SLICE
- MAKEDATE
- DATE_TRUNC
- JODATIME
- DAYOFWEEK
- YEARS_SUB
- TIMESTAMP_ADD
- HOURS_SUB
- STR2DATE
- TIMESTAMP
- FROM_DAYS
- WEEK_OF_YEAR
- YEAR
- TIMESTAMP_DIFF
- TO_TERA_TIMESTAMP
- DAYOFMONTH
- DAYOFYEAR
- DATE_FORMAT
- MONTHS_SUB
- NEXT_DAY
- MINUTES_DIFF
- DATA_ADD
- MINUTES_ADD
- CURDATE
- DAY_OF_WEEK_ISO
- CURRENt_TIMESTAMP
- STR_TO_DATE
- LAST_DAY
- WEEKS_SUB
- TO_DAYS
- DATEDIFF
- NOW
- TO_ISO8601
- TIME_TO_SEC
- QUARTER
- SECONDS_DIFF
- UTC_TIMESTAMP
- DATA_DIFF
- SECONDS_ADD
- ADDDATE
- WEEKSDIFF
- CONVERT_TZ
- MICROSECONDS_ADD
- SECOND
- YEARS_DIFF
- YEARS_ADD
- HOURS_ADD
- DAYS_SUB
- DAYS_DIFF
- Cryptographic Functions
- Percentile Functions
- Bit Functions
- JSON Functions
- Hash Functions
- Scalar Functions
- Table Functions
HLL (HyperLogLog)
Description
HLL is used for approximate count distinct.
The storage space used by HLL is determined by the distinct values in the hash value. The storage space varies depending on three conditions:
- HLL is empty. No value is inserted into HLL and the storage cost is the lowest, which is 80 bytes.
- The number of distinct hash values in HLL is less than or equal to 160. The highest storage cost is 1360 bytes (80 + 160 * 8 = 1360).
- The number of distinct hash values in HLL is greater than 160. The storage cost is fixed at 16,464 bytes (80 + 16 * 1024 = 16464).
In actual business scenarios, data volume and data distribution affect the memory usage of queries and the accuracy of the approximate result. You need to consider these two factors:
- Data volume: HLL returns an approximate value. A larger data volume results in a more accurate result. A smaller data volume results in larger deviation.
- Data distribution:In the case of large data volume and high-cardinality dimension column for GROUP BY,data computation will use more memory. HLL is not recommended in this situation. It is recommended when you perform no-group-by count distinct or GROUP BY on low-cardinality dimension columns.
- Query granularity: If you query data at a large query granularity, we recommend you use the Aggregate table or materialized view to pre-aggregate data to reduce data volume.
For details about using HLL, see Use HLL for approximate count distinct.
Examples
Specify the column type as HLL when you create a table and use the hll_union() function to aggregate data.
CREATE TABLE hllDemo
(
k1 TINYINT,
v1 HLL HLL_UNION
)
ENGINE=olap
AGGREGATE KEY(k1)
DISTRIBUTED BY HASH(k1);
In this article