Creates a table named 'name' in the 'db' database or the current database if 'db' is not set, with the structure specified in brackets and the 'engine' engine. The structure of the table is a list of column descriptions. If indexes are supported by the engine, they are indicated as parameters for the table engine.
Creates a table with the same structure as another table. You can specify a different engine for the table. If the engine is not specified, the same engine will be used as for the 'db2.name2' table.
If an expression for the default value is not defined, the default values will be set to zeros for numbers, empty strings for strings, empty arrays for arrays, and 0000-00-00 for dates or 0000-00-00 00:00:00 for dates with time. NULLs are not supported.
If the default expression is defined, the column type is optional. If there isn't an explicitly defined type, the default expression type is used. Example: ``EventDate DEFAULT toDate(EventTime)`` - the 'Date' type will be used for the 'EventDate' column.
If the data type and default expression are defined explicitly, this expression will be cast to the specified type using type casting functions. Example: ``Hits UInt32 DEFAULT 0`` means the same thing as ``Hits UInt32 DEFAULT toUInt32(0)``.
Default expressions may be defined as an arbitrary expression from table constants and columns. When creating and changing the table structure, it checks that expressions don't contain loops. For INSERT, it checks that expressions are resolvable - that all columns they can be calculated from have been passed.
Materialized expression. Such a column can't be specified for INSERT, because it is always calculated.
For an INSERT without a list of columns, these columns are not considered.
In addition, this column is not substituted when using an asterisk in a SELECT query. This is to preserve the invariant that the dump obtained using SELECT * can be inserted back into the table using INSERT without specifying the list of columns.
When using the ALTER query to add new columns, old data for these columns is not written. Instead, when reading old data that does not have values for the new columns, expressions are computed on the fly by default. However, if running the expressions requires different columns that are not indicated in the query, these columns will additionally be read, but only for the blocks of data that need it.
If you add a new column to a table but later change its default expression, the values used for old data will change (for data where values were not stored on the disk). Note that when running background merges, data for columns that are missing in one of the merging parts is written to the merged part.
In all cases, if TEMPORARY is specified, a temporary table will be created. Temporary tables have the following characteristics:
- Temporary tables disappear when the session ends, including if the connection is lost.
- A temporary table is created with the Memory engine. The other table engines are not supported.
- The DB can't be specified for a temporary table. It is created outside of databases.
- If a temporary table has the same name as another one and a query specifies the table name without specifying the DB, the temporary table will be used.
- For distributed query processing, temporary tables used in a query are passed to remote servers.
In most cases, temporary tables are not created manually, but when using external data for a query, or for distributed (GLOBAL) IN. For more information, see the appropriate sections.
Queries ``CREATE``, ``DROP``, ``ALTER``, ``RENAME`` support distributed execution on cluster.
For example, the following query creates ``Distributed`` table ``all_hits`` for each host of the cluster ``cluster``:
..code-block:: sql
CREATE TABLE IF NOT EXISTS all_hits ON CLUSTER cluster (p Date, i Int32) ENGINE = Distributed(cluster, default, hits)
To correctly execute such queries you need to have equal definitions of the cluster on each host (you can use :ref:`ZooKeeper substitutions <configuration_files>` to syncronize configs on hosts) and connection to ZooKeeper quorum.
Local version of the query will be eventually executed on each host of the cluster, even if some hosts are temporary unavaiable; on each host queries are executed stictly sequentually.
At the moment, ``ALTER`` queries for replicated tables are not supported yet.
Normal views don't store any data, but just perform a read from another table. In other words, a normal view is nothing more than a saved query. When reading from a view, this saved query is used as a subquery in the FROM clause.
When creating a materialized view, you can specify ENGINE - the table engine for storing data. By default, it uses the same engine as for the table that the SELECT query is made from.
A materialized view is arranged as follows: when inserting data to the table specified in SELECT, part of the inserted data is converted by this SELECT query, and the result is inserted in the view.
If you specify POPULATE, the existing table data is inserted in the view when creating it, as if making a CREATE TABLE ... AS SELECT ... query. Otherwise, the query contains only the data inserted in the table after creating the view. We don't recommend using POPULATE, since data inserted in the table during the view creation will not be inserted in it.
The SELECT query can contain DISTINCT, GROUP BY, ORDER BY, LIMIT ... Note that the corresponding conversions are performed independently on each block of inserted data. For example, if GROUP BY is set, data is aggregated during insertion, but only within a single packet of inserted data. The data won't be further aggregated. The exception is when using an ENGINE that independently performs data aggregation, such as SummingMergeTree.
- The query doesn't create data on the disk, but assumes that data is already in the appropriate places, and just adds information about the table to the server.
After executing an ATTACH query, the server will know about the existence of the table.
This query is used when starting the server. The server stores table metadata as files with ATTACH queries, which it simply runs at launch (with the exception of system tables, which are explicitly created on the server).
This does not delete the table's data or metadata. On the next server launch, the server will read the metadata and find out about the table again. Similarly, a "detached" table can be re-attached using the ATTACH query (with the exception of system tables, which do not have metadata stored for them).
All tables are renamed under global locking. Renaming tables is a light operation. If you indicated another database after TO, the table will be moved to this database. However, the directories with databases must reside in the same file system (otherwise, an error is returned).
Adds a new column to the table with the specified name, type, and default expression (see the section "Default expressions"). If you specify 'AFTER name_after' (the name of another column), the column is added after the specified one in the list of table columns. Otherwise, the column is added to the end of the table. Note that there is no way to add a column to the beginning of a table. For a chain of actions, 'name_after' can be the name of a column that is added in one of the previous actions.
Adding a column just changes the table structure, without performing any actions with data. The data doesn't appear on the disk after ALTER. If the data is missing for a column when reading from the table, it is filled in with default values (by performing the default expression if there is one, or using zeros or empty strings). The column appears on the disk after merging data parts (see MergeTree).
Changes the 'name' column's type to 'type' and/or the default expression to 'default_expr'. When changing the type, values are converted as if the 'toType' function were applied to them.
Only the first stage takes time. If there is a failure at this stage, the data is not changed.
If there is a failure during one of the successive stages, data can be restored manually. The exception is if the old files were deleted from the file system but the data for the new files did not get written to the disk and was lost.
The ALTER query lets you create and delete separate elements (columns) in nested data structures, but not whole nested data structures. To add a nested data structure, you can add columns with a name like 'name.nested_name' and the type 'Array(T)'. A nested data structure is equivalent to multiple array columns with a name that has the same prefix before the dot.
There is no support for deleting of columns in the primary key or the sampling key (columns that are in the ENGINE expression). Changing the type of columns in the primary key is allowed only if such change doesn't entail changing the actual data (e.g. adding the value to an Enum or changing the type from DateTime to UInt32 is allowed).
If the ALTER query is not sufficient for making the table changes you need, you can create a new table, copy the data to it using the INSERT SELECT query, then switch the tables using the RENAME query and delete the old table.
The ALTER query blocks all reads and writes for the table. In other words, if a long SELECT is running at the time of the ALTER query, the ALTER query will wait for the SELECT to complete. At the same time, all new queries to the same table will wait while this ALTER is running.
For tables that don't store data themselves (Merge and Distributed), ALTER just changes the table structure, and does not change the structure of subordinate tables. For example, when running ALTER for a Distributed table, you will also need to run ALTER for the tables on all remote servers.
The ALTER query for changing columns is replicated. The instructions are saved in ZooKeeper, then each replica applies them. All ALTER queries are run in the same order. The query waits for the appropriate actions to be completed on the other replicas. However, a query to change columns in a replicated table can be interrupted, and all actions will be performed asynchronously.
A partition in a table is data for a single calendar month. This is determined by the values of the date key specified in the table engine parameters. Each month's data is stored separately in order to simplify manipulations with this data.
``active`` - Only count active parts. Inactive parts are, for example, source parts remaining after merging to a larger part - these parts are deleted approximately 10 minutes after merging.
On an operating server, you can't manually change the set of parts or their data on the file system, since the server won't know about it. For non-replicated tables, you can do this when the server is stopped, but we don't recommended it. For replicated tables, the set of parts can't be changed in any case.
The 'detached' directory contains parts that are not used by the server - detached from the table using the ALTER ... DETACH query. Parts that are damaged are also moved to this directory, instead of deleting them. You can add, delete, or modify the data in the 'detached' directory at any time - the server won't know about this until you make the ALTER TABLE ... ATTACH query.
After the query is executed, you can do whatever you want with the data in the 'detached' directory — delete it from the file system, or just leave it.
The query is replicated - data will be moved to the 'detached' directory and forgotten on all replicas. The query can only be sent to a leader replica. To find out if a replica is a leader, perform SELECT to the 'system.replicas' system table. Alternatively, it is easier to make a query on all replicas, and all except one will throw an exception.
Similar to the DETACH operation. Deletes data from the table. Data parts will be tagged as inactive and will be completely deleted in approximately 10 minutes. The query is replicated - data will be deleted on all replicas.
The query is replicated. Each replica checks whether there is data in the 'detached' directory. If there is data, it checks the integrity, verifies that it matches the data on the server that initiated the query, and then adds it if everything is correct. If not, it downloads data from the query requestor replica, or from another replica where the data has already been added.
Creates a local backup of one or multiple partitions. The name can be the full name of the partition (for example, 201403), or its prefix (for example, 2014) - then the backup will be created for all the corresponding partitions.
The query does the following: for a data snapshot at the time of execution, it creates hardlinks to table data in the directory /var/lib/clickhouse/shadow/N/...
/var/lib/clickhouse/ is the working ClickHouse directory from the config.
The backup is created almost instantly (but first it waits for current queries to the corresponding table to finish running). At first, the backup doesn't take any space on the disk. As the system works, the backup can take disk space, as data is modified. If the backup is made for old enough data, it won't take space on the disk.
After creating the backup, data from ``/var/lib/clickhouse/shadow/`` can be copied to the remote server and then deleted on the local server. The entire backup process is performed without stopping the server.
As an alternative, you can manually copy data from the ``/var/lib/clickhouse/data/database/table directory``. But if you do this while the server is running, race conditions are possible when copying directories with files being added or changed, and the backup may be inconsistent. You can do this if the server isn't running - then the resulting data will be the same as after the ALTER TABLE t FREEZE PARTITION query.
``ALTER TABLE ... FREEZE PARTITION`` only copies data, not table metadata. To make a backup of table metadata, copy the file ``/var/lib/clickhouse/metadata/database/table.sql``
Replication provides protection from device failures. If all data disappeared on one of your replicas, follow the instructions in the "Restoration after failure" section to restore it.
Backups protect against human error (accidentally deleting data, deleting the wrong data or in the wrong cluster, or corrupting data). For high-volume databases, it can be difficult to copy backups to remote servers. In such cases, to protect from human error, you can keep a backup on the same server (it will reside in /var/lib/clickhouse/shadow/).
It downloads the specified partition from the shard that has its ZooKeeper path specified in the FROM clause, then puts it in the 'detached' directory for the specified table.
The path to ZooKeeper is specified in the FROM clause. For example, ``/clickhouse/tables/01-01/visits``.
Before downloading, the system checks that the partition exists and the table structure matches. The most appropriate replica is selected automatically from the healthy replicas.
The ALTER ... FETCH PARTITION query is not replicated. The partition will be downloaded to the 'detached' directory only on the local server. Note that if after this you use the ALTER TABLE ... ATTACH query to add data to the table, the data will be added on all replicas (on one of the replicas it will be added from the 'detached' directory, and on the rest it will be loaded from neighboring replicas).
For non-replicatable tables, all ALTER queries are performed synchronously. For replicatable tables, the query just adds instructions for the appropriate actions to ZooKeeper, and the actions themselves are performed as soon as possible. However, the query can wait for these actions to be completed on all the replicas.
**user** is the user who made the query. Keep in mind that for distributed processing, queries are sent to remote servers under the 'default' user. SHOW PROCESSLIST shows the username for a specific query, not for a query that this query initiated.
**address** is the name of the host that the query was sent from. For distributed processing, on remote servers, this is the name of the query requestor host. To track where a distributed query was originally made from, look at SHOW PROCESSLIST on the query requestor server.
**rows_read**, **bytes_read** - How many rows and bytes of uncompressed data were read when processing the query. For distributed processing, data is totaled from all the remote servers. This is the data used for restrictions and quotas.
**query_id** - The query identifier. Non-empty only if it was explicitly defined by the user. For distributed processing, the query ID is not passed to remote servers.
Returns a single UInt8-type column, which contains the single value 0 if the table or database doesn't exist, or 1 if the table exists in the specified database.
Lets you set the 'param' setting to 'value'. You can also make all the settings from the specified settings profile in a single query. To do this, specify 'profile' as the setting name. For more information, see the section "Settings". The setting is made for the session, or for the server (globally) if GLOBAL is specified.
When making a global setting, the setting is not applied to sessions already running, including the current session. It will only be used for new sessions.
The data itself comes after 'format', after all space symbols up to the first line break if there is one and including it, or after all space symbols if there isn't a line break. We recommend writing data starting from the next line (this is important if the data starts with space characters).
For more information about data formats, see the section "Formats". The "Interfaces" section describes how to insert data separately from the query when using the command-line client or the HTTP interface.
The query may optionally specify a list of columns for insertion. In this case, the default values are written to the other columns.
Default values are calculated from DEFAULT expressions specified in table definitions, or, if the DEFAULT is not explicitly defined, zeros and empty strings are used. If the 'strict_insert_default' setting is set to 1, all the columns that do not have explicit DEFAULTS must be specified in the query.
If the query omits the DISTINCT, GROUP BY, and ORDER BY clauses and the IN and JOIN subqueries, the query will be completely stream processed, using O(1) amount of RAM.
Otherwise, the query may consume too much RAM, if appropriate restrictions are not defined (max_memory_usage, max_rows_to_group_by, max_rows_to_sort, max_rows_in_distinct, max_bytes_in_distinct, max_rows_in_set, max_bytes_in_set, max_rows_in_join, max_bytes_in_join, max_bytes_before_external_sort, max_bytes_before_external_group_by). For more information, see the section "Settings". It is possible to use external sorting (saving temporary tables to a disk) and external aggregation. Merge join is not implemented.
The FROM clause specifies the table to read data from, or a subquery, or a table function; ARRAY JOIN and the regular JOIN may also be included (see below).
Instead of a table, the SELECT subquery may be specified in brackets. In this case, the subquery processing pipeline will be built into the processing pipeline of an external query.
In contrast to standard SQL, a synonym does not need to be specified after a subquery. For compatibility, it is possible to write 'AS name' after a subquery, but the specified name isn't used anywhere.
To execute a query, all the columns listed in the query are extracted from the appropriate table. Any columns not needed for the external query are thrown out of the subqueries.
If a query does not list any columns (for example, SELECT count() FROM t), some column is extracted from the table anyway (the smallest one is preferred), in order to calculate the number of rows.
The FINAL modifier can be used only for a SELECT from a CollapsingMergeTree table. When you specify FINAL, data is selected fully "collapsed". Keep in mind that using FINAL leads to a selection that includes columns related to the primary key, in addition to the columns specified in the SELECT. Additionally, the query will be executed in a single stream, and data will be merged during query execution. This means that when using FINAL, the query is processed more slowly. In most cases, you should avoid using FINAL. For more information, see the section "CollapsingMergeTree engine".
Approximated query processing is only supported by MergeTree\* type tables, and only if the sampling expression was specified during table creation (see the section "MergeTree engine").
In the first case, the query will be executed on 'k' percent of data. For example, ``SAMPLE 0.1`` runs the query on 10% of data.
In the second case, the query will be executed on a sample of no more than 'n' rows. For example, ``SAMPLE 10000000`` runs the query on a maximum of 10,000,000 rows.
In this example, the query is executed on a sample from 0.1 (10%) of data. Values of aggregate functions are not corrected automatically, so to get an approximate result, the value 'count()' is manually multiplied by 10.
When using something like ``SAMPLE 10000000``, there isn't any information about which relative percent of data was processed or what the aggregate functions should be multiplied by, so this method of writing is not always appropriate to the situation.
A sample with a relative coefficient is "consistent": if we look at all possible data that could be in the table, a sample (when using a single sampling expression specified during table creation) with the same coefficient always selects the same subset of possible data. In other words, a sample from different tables on different servers at different times is made the same way.
For example, a sample of user IDs takes rows with the same subset of all the possible user IDs from different tables. This allows using the sample in subqueries in the IN clause, as well as for manually correlating results of different queries with samples.
An alias can be specified for an array in the ARRAY JOIN clause. In this case, an array item can be accessed by this alias, but the array itself by the original name. Example:
Multiple arrays of the same size can be comma-separated in the ARRAY JOIN clause. In this case, JOIN is performed with them simultaneously (the direct sum, not the direct product).
When specifying names of nested data structures in ARRAY JOIN, the meaning is the same as ARRAY JOIN with all the array elements that it consists of. Example:
The corresponding conversion can be performed before the WHERE/PREWHERE clause (if its result is needed in this clause), or after completing WHERE/PREWHERE (to reduce the volume of calculations).
Performs joins with data from the subquery. At the beginning of query execution, the subquery specified after JOIN is run, and its result is saved in memory. Then it is read from the "left" table specified in the FROM clause, and while it is being read, for each of the read rows from the "left" table, rows are selected from the subquery results table (the "right" table) that meet the condition for matching the values of the columns specified in USING.
The table name can be specified instead of a subquery. This is equivalent to the ``SELECT * FROM table`` subquery, except in a special case when the table has the Join engine - an array prepared for joining.
If INNER is specified, the result will contain only those rows that have a matching row in the right table.
If LEFT is specified, any rows in the left table that don't have matching rows in the right table will be assigned the default value - zeros or empty rows. LEFT OUTER may be written instead of LEFT; the word OUTER does not affect anything.
Using ALL corresponds to the normal JOIN semantic from standard SQL.
Using ANY is optimal. If the right table has only one matching row, the results of ANY and ALL are the same. You must specify either ANY or ALL (neither of them is selected by default).
When using a normal ``JOIN``, the query is sent to remote servers. Subqueries are run on each of them in order to make the right table, and the join is performed with this table. In other words, the right table is formed on each server separately.
When using ``GLOBAL ... JOIN``, first the requestor server runs a subquery to calculate the right table. This temporary table is passed to each remote server, and queries are run on them using the temporary data that was transmitted.
When running JOINs, there is no optimization of the order of execution in relation to other stages of the query. The join (a search in the right table) is run before filtering in WHERE and before aggregation. In order to explicitly set the order of execution, we recommend running a JOIN subquery with a subquery.
Subqueries don't allow you to set names or use them for referencing a column from a specific subquery.
The columns specified in USING must have the same names in both subqueries, and the other columns must be named differently. You can use aliases to change the names of columns in subqueries (the example uses the aliases 'hits' and 'visits').
The USING clause specifies one or more columns to join, which establishes the equality of these columns. The list of columns is set without brackets. More complex join conditions are not supported.
Each time a query is run with the same JOIN, the subquery is run again - the result is not cached. To avoid this, use the special 'Join' table engine, which is a prepared array for joining that is always in RAM. For more information, see the section "Table engines, Join".
In some cases, it is more efficient to use IN instead of JOIN. Among the various types of JOINs, the most efficient is ANY LEFT JOIN, then ANY INNER JOIN. The least efficient are ALL LEFT JOIN and ALL INNER JOIN.
If you need a JOIN for joining with dimension tables (these are relatively small tables that contain dimension properties, such as names for advertising campaigns), a JOIN might not be very convenient due to the bulky syntax and the fact that the right table is re-accessed for every query. For such cases, there is an "external dictionaries" feature that you should use instead of JOIN. For more information, see the section "External dictionaries".
This clause has the same meaning as the WHERE clause. The difference is in which data is read from the table. When using PREWHERE, first only the columns necessary for executing PREWHERE are read. Then the other columns are read that are needed for running the query, but only those blocks where the PREWHERE expression is true.
It makes sense to use PREWHERE if there are filtration conditions that are not suitable for indexes that are used by a minority of the columns in the query, but that provide strong data filtration. This reduces the volume of data to read.
Keep in mind that it does not make much sense for PREWHERE to only specify those columns that have an index, because when using an index, only the data blocks that match the index are read.
If the 'optimize_move_to_prewhere' setting is set to 1 and PREWHERE is omitted, the system uses heuristics to automatically move parts of expressions from WHERE to PREWHERE.
If there is a GROUP BY clause, it must contain a list of expressions. Each expression will be referred to here as a "key".
All the expressions in the SELECT, HAVING, and ORDER BY clauses must be calculated from keys or from aggregate functions. In other words, each column selected from the table must be used either in keys or inside aggregate functions.
If a query contains only table columns inside aggregate functions, the GROUP BY clause can be omitted, and aggregation by an empty set of keys is assumed.
However, in contrast to standard SQL, if the table doesn't have any rows (either there aren't any at all, or there aren't any after using WHERE to filter), an empty result is returned, and not the result from one of the rows containing the initial values of aggregate functions.
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'.
A constant can't be specified as arguments for aggregate functions. Example: sum(1). Instead of this, you can get rid of the constant. Example: ``count()``.
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).
In JSON* formats, this row is output as a separate 'totals' field. In TabSeparated 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.
``WITH TOTALS`` can be run in different ways when HAVING is present. The behavior depends on the 'totals_mode' setting.
By default, totals_mode = 'before_having'. In this case, 'totals' is calculated across all rows, including the ones that don't 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.
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').
It is possible to turn on spilling temporary data to disk to limit memory consumption during the execution of GROUP BY. Value of ``max_bytes_before_external_group_by`` setting determines the maximum memory consumption before temporary data is dumped to the file system. If it is 0 (the default value), the feature is turned off.
When using ``max_bytes_before_external_group_by`` it is advisable to set ``max_memory_usage`` to an approximately twice greater value. The reason for this is that aggregation is executed in two stages: reading and generation of intermediate data (1) and merging of intermediate data (2). Spilling data to the filesystem can be performed only on stage 1. If the spilling did not happen, then stage 2 could consume up to the same amount of memory as stage 1.
For example: if ``max_memory_usage`` is equal 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. If dumping data to the file system happened at least once during the execution, maximum memory consumption would be just a little bit higher than ``max_bytes_before_external_group_by``.
During distributed query execution external aggregation is performed on the remote servers. If you want the memory consumption on the originating server to be small, set ``distributed_aggregation_memory_efficient`` to 1. If ``distributed_aggregation_memory_efficient`` is turned on then during merging of the dumped data and also during merging of the query results from the remote servers, total memory consumption is no more than 1/256 * number of threads of the total amount of memory.
If external aggregation is turned on and total memory consumption was less than ``max_bytes_before_external_group_by`` (meaning that no spilling took place), the query performance is the same as when external aggregation is turned off. If some data was dumped, then execution time will be several times longer (approximately 3x).
If you have an ORDER BY clause with some small LIMIT after a GROUP BY, then ORDER BY will not consume significant amount of memory. But if no LIMIT is provided, don't forget to turn on external sorting (``max_bytes_before_external_sort``).
LIMIT ``N`` BY ``COLUMNS`` allows you to restrict top ``N`` rows per each group of ``COLUMNS``. ``LIMIT N BY`` is unrelated to ``LIMIT`` clause. Key for ``LIMIT N BY`` could contain arbitrary number of columns or expressions.
Allows filtering the result received after GROUP BY, similar to the WHERE clause.
WHERE and HAVING differ in that WHERE is performed before aggregation (GROUP BY), while HAVING is performed after it. If aggregation is not performed, HAVING can't be used.
The ORDER BY clause contains a list of expressions, which can each be assigned DESC or ASC (the sorting direction). If the direction is not specified, ASC is assumed. ASC is sorted in ascending order, and DESC in descending order. The sorting direction applies to a single expression, not to the entire list. Example: ``ORDER BY Visits DESC, SearchPhrase``
For sorting by String values, you can specify collation (comparison). Example: ``ORDER BY SearchPhrase COLLATE 'tr'`` - for sorting by keyword in ascending order, using the Turkish alphabet, case insensitive, assuming that strings are UTF-8 encoded. COLLATE can be specified or not for each expression in ORDER BY independently. If ASC or DESC is specified, COLLATE is specified after it. When using COLLATE, sorting is always case-insensitive.
We only recommend using COLLATE for final sorting of a small number of rows, since sorting with COLLATE is less efficient than normal sorting by bytes.
Rows that have identical values for the list of sorting expressions are output in an arbitrary order, which can also be nondeterministic (different each time).
If the ORDER BY clause is omitted, the order of the rows is also undefined, and may be nondeterministic as well.
When floating point numbers are sorted, NaNs are separate from the other values. Regardless of the sorting order, NaNs come at the end. In other words, for ascending sorting they are placed as if they are larger than all the other numbers, while for descending sorting they are placed as if they are smaller than the rest.
Less RAM is used if a small enough LIMIT is specified in addition to ORDER BY. Otherwise, the amount of memory spent is proportional to the volume of data for sorting. For distributed query processing, if GROUP BY is omitted, sorting is partially done on remote servers, and the results are merged on the requestor server. This means that for distributed sorting, the volume of data to sort can be greater than the amount of memory on a single server.
If there is not enough RAM, it is possible to perform sorting in external memory (creating temporary files on a disk). Use the setting max_bytes_before_external_sort for this purpose. If it is set to 0 (the default), external sorting is disabled. If it is enabled, when the volume of data to sort reaches the specified number of bytes, the collected data is sorted and dumped into a temporary file. After all data is read, all the sorted files are merged and the results are output. Files are written to the /var/lib/clickhouse/tmp/ directory in the config (by default, but you can use the 'tmp_path' parameter to change this setting).
Running a query may use more memory than ``'max_bytes_before_external_sort'``. For this reason, this setting must have a value significantly smaller than 'max_memory_usage'. As an example, if your server has 128 GB of RAM and you need to run a single query, set 'max_memory_usage' to 100 GB, and 'max_bytes_before_external_sort' to 80 GB.
The expressions specified in the SELECT clause are analyzed after the calculations for all the clauses listed above are completed.
More specifically, expressions are analyzed that are above the aggregate functions, if there are any aggregate functions. The aggregate functions and everything below them are calculated during aggregation (GROUP BY). These expressions work as if they are applied to separate rows in the result.
If DISTINCT is specified, only a single row will remain out of all the sets of fully matching rows in the result.
The result will be the same as if GROUP BY were specified across all the fields specified in SELECT without aggregate functions. But there are several differences from GROUP BY:
- When ORDER BY is omitted and LIMIT is defined, the query stops running immediately after the required number of different rows has been read. In this case, using DISTINCT is much more optimal.
- Data blocks are output as they are processed, without waiting for the entire query to finish running.
Only UNION ALL is supported. The regular UNION (UNION DISTINCT) is not supported. If you need UNION DISTINCT, you can write SELECT DISTINCT from a subquery containing UNION ALL.
The structure of results (the number and type of columns) must match for the queries, but the column names can differ. In this case, the column names for the final result will be taken from the first query.
Queries that are parts of UNION ALL can't be enclosed in brackets. ORDER BY and LIMIT are applied to separate queries, not to the final result. If you need to apply a conversion to the final result, you can put all the queries with UNION ALL in a subquery in the FROM clause.
Specify 'FORMAT format' to get data in any specified format.
You can use this for convenience, or for creating dumps. For more information, see the section "Formats".
If the FORMAT clause is omitted, the default format is used, which depends on both the settings and the interface used for accessing the DB. For the HTTP interface and the command-line client in batch mode, the default format is TabSeparated. For the command-line client in interactive mode, the default format is PrettyCompact (it has attractive and compact tables).
When using the command-line client, data is passed to the client in an internal efficient format. The client independently interprets the FORMAT clause of the query and formats the data itself (thus relieving the network and the server from the load).
If the left side is a single column that is in the index, and the right side is a set of constants, the system uses the index for processing the query.
Don't list too many values explicitly (i.e. millions). If a data set is large, put it in a temporary table (for example, see the section "External data for query processing"), then use a subquery.
The right side of the operator can be a set of constant expressions, a set of tuples with constant expressions (shown in the examples above), or the name of a database table or SELECT subquery in brackets.
If the right side of the operator is the name of a table (for example, ``UserID IN users``), this is equivalent to the subquery ``UserID IN (SELECT * FROM users)``. Use this when working with external data that is sent along with the query. For example, the query can be sent together with a set of user IDs loaded to the 'users' temporary table, which should be filtered.
If the right side of the operator is a table name that has the Set engine (a prepared data set that is always in RAM), the data set will not be created over again for each query.
There are two versions of INs with subqueries (and for JOINs): the regular ``IN`` / ``JOIN``, and ``GLOBAL IN`` / ``GLOBAL JOIN``. They differ in how they are run for distributed query processing.
When using ``GLOBAL IN`` / ``GLOBAL JOIN``, first all the subqueries for ``GLOBAL IN`` / ``GLOBAL JOIN`` are run, and the results are collected in temporary tables. Then the temporary tables are sent to each remote server, where the queries are run using this temporary data.
Let's look at some examples. Assume that each server in the cluster has a normal local_table. Each server also has a **distributed_table** table with the Distributed type, which looks at all the servers in the cluster.
and run on each of them in parallel, until it reaches the stage where intermediate results can be combined. Then the intermediate results will be returned to the requestor server and merged on it, and the final result will be sent to the client.
In other words, the data set in the IN clause will be collected on each server independently, only across the data that is stored locally on each of the servers.
This will work correctly and optimally if you are prepared for this case and have spread data across the cluster servers such that the data for a single UserID resides entirely on a single server. In this case, all the necessary data will be available locally on each server. Otherwise, the result will be inaccurate. We refer to this variation of the query as "local IN".
To correct how the query works when data is spread randomly across the cluster servers, you could specify **distributed_table** inside a subquery. The query would look like this:
Each of the remote servers will start running the subquery. Since the subquery uses a distributed table, each remote server will re-send the subquery to every remote server, as
For example, if you have a cluster of 100 servers, executing the entire query will require 10,000 elementary requests, which is generally considered unacceptable.
and the temporary table '_data1' will be sent to every remote server together with the query (the name of the temporary table is implementation-defined).
#. When creating a temporary table, data is not made unique. To reduce the volume of data transmitted over the network, specify DISTINCT in the subquery. (You don't need to do this for a normal IN.)
#. The temporary table will be sent to all the remote servers. Transmission does not account for network topology. For example, if 10 remote servers reside in a datacenter that is very remote in relation to the requestor server, the data will be sent 10 times over the channel to the remote datacenter. Try to avoid large data sets when using GLOBAL IN.
#. When transmitting data to remote servers, restrictions on network bandwidth are not configurable. You might overload the network.
#. Try to distribute data across servers so that you don't need to use GLOBAL IN on a regular basis.
#. If you need to use GLOBAL IN often, plan the location of the ClickHouse cluster so that in each datacenter, there will be at least one replica of each shard, and there is a fast network between them - for possibility to process query with transferring data only inside datacenter.
It also makes sense to specify a local table in the GLOBAL IN clause, in case this local table is only available on the requestor server and you want to use data from it on remote servers.
In addition to results, you can also get minimum and maximum values for the results columns. To do this, set the 'extremes' setting to '1'. Minimums and maximums are calculated for numeric types, dates, and dates with times. For other columns, the default values are output.
An extra two rows are calculated - the minimums and maximums, respectively. These extra two rows are output in JSON*, TabSeparated*, and Pretty* formats, separate from the other rows. They are not output for other formats.
In JSON* formats, the extreme values are output in a separate 'extremes' field. In TabSeparated formats, the row comes after the main result, and after 'totals' if present. It is preceded by an empty row (after the other data). In Pretty formats, the row is output as a separate table after the main result, and after 'totals' if present.
Extreme values are calculated for rows that have passed through LIMIT. However, when using 'LIMIT offset, size', the rows before 'offset' are included in 'extremes'. In stream requests, the result may also include a small number of rows that passed through LIMIT.
You can put an asterisk in any part of a query instead of an expression. When the query is analyzed, the asterisk is expanded to a list of all table columns (excluding the ``MATERIALIZED`` and ALIAS columns). There are only a few cases when using an asterisk is justified:
* When creating a table dump.
* For tables containing just a few columns, such as system tables.
* For getting information about what columns are in a table. In this case, set ``'LIMIT 1'``. But it is better to use the ``DESC TABLE`` query.
* When there is strong filtration on a small number of columns using ``PREWHERE``.
* In subqueries (since columns that aren't needed for the external query are excluded from subqueries).
In all other cases, we don't recommend using the asterisk, since it only gives you the drawbacks of a columnar DBMS instead of the advantages.
