GROUP BY Clause
GROUP BY
clause switches the SELECT
query into an aggregation mode, which works as follows:
GROUP BY
clause contains a list of expressions (or a single expression, which is considered to be the list of length one). This list acts as a “grouping key”, while each individual expression will be referred to as a “key expression”.- All the expressions in the SELECT, HAVING, and ORDER BY clauses must be calculated based on key expressions or on aggregate functions over non-key expressions (including plain columns). In other words, each column selected from the table must be used either in a key expression or inside an aggregate function, but not both.
- Result of aggregating
SELECT
query will contain as many rows as there were unique values of “grouping key” in source table. Usually, this significantly reduces the row count, often by orders of magnitude, but not necessarily: row count stays the same if all “grouping key” values were distinct.
When you want to group data in the table by column numbers instead of column names, enable the setting enable_positional_arguments.
There’s an additional way to run aggregation over a table. If a query contains table columns only inside aggregate functions, the GROUP BY clause
can be omitted, and aggregation by an empty set of keys is assumed. Such queries always return exactly one row.
NULL Processing
For grouping, ClickHouse interprets NULL as a value, and NULL==NULL
. It differs from NULL
processing in most other contexts.
Here’s an example to show what this means.
Assume you have this table:
┌─x─┬────y─┐
│ 1 │ 2 │
│ 2 │ ᴺᵁᴸᴸ │
│ 3 │ 2 │
│ 3 │ 3 │
│ 3 │ ᴺᵁᴸᴸ │
└───┴──────┘
The query SELECT sum(x), y FROM t_null_big GROUP BY y
results in:
┌─sum(x)─┬────y─┐
│ 4 │ 2 │
│ 3 │ 3 │
│ 5 │ ᴺᵁᴸᴸ │
└────────┴──────┘
You can see that GROUP BY
for y = NULL
summed up x
, as if NULL
is this value.
If you pass several keys to GROUP BY
, the result will give you all the combinations of the selection, as if NULL
were a specific value.
ROLLUP Modifier
ROLLUP
modifier is used to calculate subtotals for the key expressions, based on their order in the GROUP BY
list. The subtotals rows are added after the result table.
The subtotals are calculated in the reverse order: at first subtotals are calculated for the last key expression in the list, then for the previous one, and so on up to the first key expression.
In the subtotals rows the values of already "grouped" key expressions are set to 0
or empty line.
Mind that HAVING clause can affect the subtotals results.
Example
Consider the table t:
┌─year─┬─month─┬─day─┐
│ 2019 │ 1 │ 5 │
│ 2019 │ 1 │ 15 │
│ 2020 │ 1 │ 5 │
│ 2020 │ 1 │ 15 │
│ 2020 │ 10 │ 5 │
│ 2020 │ 10 │ 15 │
└──────┴───────┴─────┘
Query:
SELECT year, month, day, count(*) FROM t GROUP BY ROLLUP(year, month, day);
As GROUP BY
section has three key expressions, the result contains four tables with subtotals "rolled up" from right to left:
GROUP BY year, month, day
;GROUP BY year, month
(andday
column is filled with zeros);GROUP BY year
(nowmonth, day
columns are both filled with zeros);- and totals (and all three key expression columns are zeros).
┌─year─┬─month─┬─day─┬─count()─┐
│ 2020 │ 10 │ 15 │ 1 │
│ 2020 │ 1 │ 5 │ 1 │
│ 2019 │ 1 │ 5 │ 1 │
│ 2020 │ 1 │ 15 │ 1 │
│ 2019 │ 1 │ 15 │ 1 │
│ 2020 │ 10 │ 5 │ 1 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 2019 │ 1 │ 0 │ 2 │
│ 2020 │ 1 │ 0 │ 2 │
│ 2020 │ 10 │ 0 │ 2 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 2019 │ 0 │ 0 │ 2 │
│ 2020 │ 0 │ 0 │ 4 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 0 │ 0 │ 0 │ 6 │
└──────┴───────┴─────┴─────────┘
The same query also can be written using WITH
keyword.
SELECT year, month, day, count(*) FROM t GROUP BY year, month, day WITH ROLLUP;
See also
- group_by_use_nulls setting for SQL standard compatibility.
CUBE Modifier
CUBE
modifier is used to calculate subtotals for every combination of the key expressions in the GROUP BY
list. The subtotals rows are added after the result table.
In the subtotals rows the values of all "grouped" key expressions are set to 0
or empty line.
Mind that HAVING clause can affect the subtotals results.
Example
Consider the table t:
┌─year─┬─month─┬─day─┐
│ 2019 │ 1 │ 5 │
│ 2019 │ 1 │ 15 │
│ 2020 │ 1 │ 5 │
│ 2020 │ 1 │ 15 │
│ 2020 │ 10 │ 5 │
│ 2020 │ 10 │ 15 │
└──────┴───────┴─────┘
Query:
SELECT year, month, day, count(*) FROM t GROUP BY CUBE(year, month, day);
As GROUP BY
section has three key expressions, the result contains eight tables with subtotals for all key expression combinations:
GROUP BY year, month, day
GROUP BY year, month
GROUP BY year, day
GROUP BY year
GROUP BY month, day
GROUP BY month
GROUP BY day
- and totals.
Columns, excluded from GROUP BY
, are filled with zeros.
┌─year─┬─month─┬─day─┬─count()─┐
│ 2020 │ 10 │ 15 │ 1 │
│ 2020 │ 1 │ 5 │ 1 │
│ 2019 │ 1 │ 5 │ 1 │
│ 2020 │ 1 │ 15 │ 1 │
│ 2019 │ 1 │ 15 │ 1 │
│ 2020 │ 10 │ 5 │ 1 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 2019 │ 1 │ 0 │ 2 │
│ 2020 │ 1 │ 0 │ 2 │
│ 2020 │ 10 │ 0 │ 2 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 2020 │ 0 │ 5 │ 2 │
│ 2019 │ 0 │ 5 │ 1 │
│ 2020 │ 0 │ 15 │ 2 │
│ 2019 │ 0 │ 15 │ 1 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 2019 │ 0 │ 0 │ 2 │
│ 2020 │ 0 │ 0 │ 4 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 0 │ 1 │ 5 │ 2 │
│ 0 │ 10 │ 15 │ 1 │
│ 0 │ 10 │ 5 │ 1 │
│ 0 │ 1 │ 15 │ 2 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 0 │ 1 │ 0 │ 4 │
│ 0 │ 10 │ 0 │ 2 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 0 │ 0 │ 5 │ 3 │
│ 0 │ 0 │ 15 │ 3 │
└──────┴───────┴─────┴─────────┘
┌─year─┬─month─┬─day─┬─count()─┐
│ 0 │ 0 │ 0 │ 6 │
└──────┴───────┴─────┴─────────┘
The same query also can be written using WITH
keyword.
SELECT year, month, day, count(*) FROM t GROUP BY year, month, day WITH CUBE;
See also
- group_by_use_nulls setting for SQL standard compatibility.
WITH TOTALS Modifier
If the WITH TOTALS
modifier is specified, another row will be calculated. This row will have key columns containing default values (zeros or empty lines), and columns of aggregate functions with the values calculated across all the rows (the “total” values).
This extra row is only produced in JSON*
, TabSeparated*
, and Pretty*
formats, separately from the other rows:
- In
XML
andJSON*
formats, this row is output as a separate ‘totals’ field. - In
TabSeparated*
,CSV*
andVertical
formats, the row comes after the main result, preceded by an empty row (after the other data). - In
Pretty*
formats, the row is output as a separate table after the main result. - In
Template
format, the row is output according to specified template. - In the other formats it is not available.
totals is output in the results of SELECT
queries, and is not output in INSERT INTO ... SELECT
.
WITH TOTALS
can be run in different ways when HAVING is present. The behavior depends on the totals_mode
setting.
Configuring Totals Processing
By default, totals_mode = 'before_having'
. In this case, ‘totals’ is calculated across all rows, including the ones that do not pass through HAVING and max_rows_to_group_by
.
The other alternatives include only the rows that pass through HAVING in ‘totals’, and behave differently with the setting max_rows_to_group_by
and group_by_overflow_mode = 'any'
.
after_having_exclusive
– Don’t include rows that didn’t pass through max_rows_to_group_by
. In other words, ‘totals’ will have less than or the same number of rows as it would if max_rows_to_group_by
were omitted.
after_having_inclusive
– Include all the rows that didn’t pass through ‘max_rows_to_group_by’ in ‘totals’. In other words, ‘totals’ will have more than or the same number of rows as it would if max_rows_to_group_by
were omitted.
after_having_auto
– Count the number of rows that passed through HAVING. If it is more than a certain amount (by default, 50%), include all the rows that didn’t pass through ‘max_rows_to_group_by’ in ‘totals’. Otherwise, do not include them.
totals_auto_threshold
– By default, 0.5. The coefficient for after_having_auto
.
If max_rows_to_group_by
and group_by_overflow_mode = 'any'
are not used, all variations of after_having
are the same, and you can use any of them (for example, after_having_auto
).
You can use WITH TOTALS
in subqueries, including subqueries in the JOIN clause (in this case, the respective total values are combined).
GROUP BY ALL
GROUP BY ALL
is equivalent to listing all the SELECT-ed expressions that are not aggregate functions.
For example:
SELECT
a * 2,
b,
count(c),
FROM t
GROUP BY ALL
is the same as
SELECT
a * 2,
b,
count(c),
FROM t
GROUP BY a * 2, b
For a special case that if there is a function having both aggregate functions and other fields as its arguments, the GROUP BY
keys will contain the maximum non-aggregate fields we can extract from it.
For example:
SELECT
substring(a, 4, 2),
substring(substring(a, 1, 2), 1, count(b))
FROM t
GROUP BY ALL
is the same as
SELECT
substring(a, 4, 2),
substring(substring(a, 1, 2), 1, count(b))
FROM t
GROUP BY substring(a, 4, 2), substring(a, 1, 2)
Examples
Example:
SELECT
count(),
median(FetchTiming > 60 ? 60 : FetchTiming),
count() - sum(Refresh)
FROM hits
As opposed to MySQL (and conforming to standard SQL), you can’t get some value of some column that is not in a key or aggregate function (except constant expressions). To work around this, you can use the ‘any’ aggregate function (get the first encountered value) or ‘min/max’.
Example:
SELECT
domainWithoutWWW(URL) AS domain,
count(),
any(Title) AS title -- getting the first occurred page header for each domain.
FROM hits
GROUP BY domain
For every different key value encountered, GROUP BY
calculates a set of aggregate function values.
GROUPING SETS modifier
This is the most general modifier. This modifier allows manually specifying several aggregation key sets (grouping sets). Aggregation is performed separately for each grouping set, and after that, all results are combined. If a column is not presented in a grouping set, it's filled with a default value.
In other words, modifiers described above can be represented via GROUPING SETS
.
Despite the fact that queries with ROLLUP
, CUBE
and GROUPING SETS
modifiers are syntactically equal, they may perform differently.
When GROUPING SETS
try to execute everything in parallel, ROLLUP
and CUBE
are executing the final merging of the aggregates in a single thread.
In the situation when source columns contain default values, it might be hard to distinguish if a row is a part of the aggregation which uses those columns as keys or not.
To solve this problem GROUPING
function must be used.
Example
The following two queries are equivalent.
-- Query 1
SELECT year, month, day, count(*) FROM t GROUP BY year, month, day WITH ROLLUP;
-- Query 2
SELECT year, month, day, count(*) FROM t GROUP BY
GROUPING SETS
(
(year, month, day),
(year, month),
(year),
()
);
See also
- group_by_use_nulls setting for SQL standard compatibility.
Implementation Details
Aggregation is one of the most important features of a column-oriented DBMS, and thus it’s implementation is one of the most heavily optimized parts of ClickHouse. By default, aggregation is done in memory using a hash-table. It has 40+ specializations that are chosen automatically depending on “grouping key” data types.
GROUP BY Optimization Depending on Table Sorting Key
The aggregation can be performed more effectively, if a table is sorted by some key, and GROUP BY
expression contains at least prefix of sorting key or injective functions. In this case when a new key is read from table, the in-between result of aggregation can be finalized and sent to client. This behaviour is switched on by the optimize_aggregation_in_order setting. Such optimization reduces memory usage during aggregation, but in some cases may slow down the query execution.
GROUP BY in External Memory
You can enable dumping temporary data to the disk to restrict memory usage during GROUP BY
.
The max_bytes_before_external_group_by setting determines the threshold RAM consumption for dumping GROUP BY
temporary data to the file system. If set to 0 (the default), it is disabled.
When using max_bytes_before_external_group_by
, we recommend that you set max_memory_usage
about twice as high. This is necessary because there are two stages to aggregation: reading the data and forming intermediate data (1) and merging the intermediate data (2). Dumping data to the file system can only occur during stage 1. If the temporary data wasn’t dumped, then stage 2 might require up to the same amount of memory as in stage 1.
For example, if max_memory_usage was set to 10000000000 and you want to use external aggregation, it makes sense to set max_bytes_before_external_group_by
to 10000000000, and max_memory_usage
to 20000000000. When external aggregation is triggered (if there was at least one dump of temporary data), maximum consumption of RAM is only slightly more than max_bytes_before_external_group_by
.
With distributed query processing, external aggregation is performed on remote servers. In order for the requester server to use only a small amount of RAM, set distributed_aggregation_memory_efficient
to 1.
When merging data flushed to the disk, as well as when merging results from remote servers when the distributed_aggregation_memory_efficient
setting is enabled, consumes up to 1/256 * the_number_of_threads
from the total amount of RAM.
When external aggregation is enabled, if there was less than max_bytes_before_external_group_by
of data (i.e. data was not flushed), the query runs just as fast as without external aggregation. If any temporary data was flushed, the run time will be several times longer (approximately three times).
If you have an ORDER BY with a LIMIT after GROUP BY
, then the amount of used RAM depends on the amount of data in LIMIT
, not in the whole table. But if the ORDER BY
does not have LIMIT
, do not forget to enable external sorting (max_bytes_before_external_sort
).