ClickHouse Driver for Kipchak

A ClickHouse driver for Kipchak: run analytics queries, stream large results, and insert events in batches from your Kipchak API. ClickHouse pairs naturally with the Kafka driver: ingest with Kafka, query with ClickHouse.

• Queries: select() returns rows with their column types; value() and first() for single answers. Values go in {name:Type} placeholders and travel separately from the SQL, so they cannot change the query.
• Streaming: each() yields rows as ClickHouse sends them, so a result of any size uses little memory. Stopping early cancels the query on the server.
• Inserts: insert() sends rows in batches as JSON lines, gzip-compressed, with server-side batching (async_insert) for many small writes.
• Statements: execute() for DDL and INSERT ... SELECT; cancel() stops a running query by its ID.
• Production settings: several hosts with failover, timeouts, HTTPS with CA files and client certificates, and per-query settings. Nothing is held open between requests, so it is safe in FrankenPHP's long-lived workers.
• Built on Kipchak's drivers: requests go through the HTTP driver, warnings to the Logger driver, and results can be cached in worker memory, with the Filecache driver or with the Valkey driver.

Uses ClickHouse's HTTP interface, so no PHP extension is needed. Tested against ClickHouse 25.8.

Composer Package

kipchak/driver-clickhouse

Installation

composer require kipchak/driver-clickhouse

Initialise it in drivers/drivers.php, after the Logger and HTTP drivers (and after Filecache or Valkey, if you cache results with them):

use Kipchak\Driver\ClickHouse\ClickHouse;

ClickHouse::initialise($container);

Configuration

Copy sample.config.php to config/kipchak.clickhouse.php:

return [
    'connections' => [
        'default' => [
            'hosts' => [env('CLICKHOUSE_URL', 'http://clickhouse:8123')],
            'database' => env('CLICKHOUSE_DATABASE', 'default'),
            'username' => env('CLICKHOUSE_USER', 'default'),
            'password' => env('CLICKHOUSE_PASSWORD', ''),
        ],
    ],
];

Add more entries under connections for more clusters or databases, and fetch them by name: ClickHouse::get('archive').

Usage

Queries

use Kipchak\Driver\ClickHouse\ClickHouse;

$clickhouse = ClickHouse::get();

$result = $clickhouse->select(
    'SELECT toDate(at) AS day, count() AS hits
     FROM events
     WHERE user_id = {user:UInt64} AND at >= {from:DateTime}
     GROUP BY day ORDER BY day',
    ['user' => $userId, 'from' => new DateTimeImmutable('-7 days')],
);

foreach ($result as $row) {
    $row['day'];    // '2026-10-29'
    $row['hits'];   // 42
}
$result->types;          // ['day' => 'Date', 'hits' => 'UInt64']
$result->column('hits'); // [40, 42, ...]
$result->rowsRead;       // rows ClickHouse read to answer
$result->queryId;

$clickhouse->value('SELECT count() FROM events');             // 1250000
$clickhouse->first('SELECT * FROM users WHERE id = {id:UInt64}', ['id' => 7]);

Write each value as a placeholder, {name:Type}, and pass it in the parameters array. ClickHouse parses it as that type, so a value can never change the query. PHP values are sent as follows:

Table and column names can be placeholders too: SELECT count() FROM {table:Identifier}.

Results come back as PHP values. Integers that fit in a PHP int are ints, and larger ones (UInt64 above PHP_INT_MAX, Int128, Int256) are strings. Decimals are strings, so no precision is lost. DateTime values are strings in the column's time zone; set session_timezone in settings to choose one.

select() sets the output format itself, so leave FORMAT out of the query. ClickHouse's HTTP interface runs one statement per request.

Streaming large results

foreach ($clickhouse->each('SELECT * FROM events WHERE day = {day:Date}', ['day' => '2026-10-29']) as $row) {
    fwrite($csv, implode(',', $row) . "\n");
}

each() holds one row in memory at a time. The query is sent when the loop starts. The timeout limits each wait for the next part of the result, not the whole stream, so long exports are not cut off. Breaking out of the loop closes the connection and ClickHouse cancels the query. An error part-way through throws after the rows before it.

Inserting

$summary = $clickhouse->insert('events', [
    ['user_id' => 7, 'action' => 'login', 'at' => new DateTimeImmutable()],
    ['user_id' => 8, 'action' => 'logout', 'at' => new DateTimeImmutable()],
]);
$summary->writtenRows;   // 2

Rows are arrays (or objects) of column => value; missing columns get their defaults. insert() accepts any iterable, including a generator, and sends insert_batch_size rows per INSERT. Each batch is atomic for a MergeTree table; if one fails, earlier batches stay inserted.

ClickHouse prefers fewer, larger inserts. When many requests each insert a few rows, let ClickHouse batch them:

$clickhouse->insert('events', [$event], ['async_insert' => 1, 'wait_for_async_insert' => 1]);

With async_insert, the server writes the rows with the next flush, so writtenRows is 0.

Statements and cancelling

$clickhouse->execute('ALTER TABLE events DELETE WHERE user_id = {user:UInt64}', ['user' => 7]);
$summary = $clickhouse->execute('INSERT INTO daily SELECT toDate(at), count() FROM events GROUP BY 1');

$result = $clickhouse->select($sql, queryId: 'report-2026-10');
$clickhouse->cancel('report-2026-10');   // from another request

Every method takes ClickHouse settings for that call (settings: ['max_execution_time' => 5]) and an optional queryId; otherwise each call gets a UUID. Results, summaries and exceptions carry the query ID, so you can find the query in system.query_log.

Caching

$result = $clickhouse->select($dashboardSql, $params, cacheTtl: 60);
$result->cached;   // true when it came from the cache

Only calls with a cacheTtl are cached, keyed by the SQL, parameters and settings. Writes do not clear the cache; the TTL decides. clearCache() forgets the connection's cached results.

Errors

• ClickHouseConnectionException: no host could be reached, the TLS handshake failed, or the query timed out. Answer 503.
• ClickHouseException: ClickHouse refused or failed the query. errorCode and errorName are ClickHouse's (60 UNKNOWN_TABLE, 62 SYNTAX_ERROR, 159 TIMEOUT_EXCEEDED, 241 MEMORY_LIMIT_EXCEEDED, 516 AUTHENTICATION_FAILED, ...), with httpStatus and queryId.

Queries running when ClickHouse shuts down fail with QUERY_WAS_CANCELLED, the same code a cancel() gives.

Diagnostics

$clickhouse->health();          // whether ClickHouse answers with this connection's credentials; never throws
$clickhouse->serverVersion();   // '25.8.33.6'

Failover and FrankenPHP worker mode

One client per connection is kept for the life of each worker. It holds nothing between requests except the list of hosts that recently failed.

When a host refuses the connection or its name does not resolve, the statement never reached it, so any statement moves on to the next host. Reads also move on after other connection failures. Timeouts are never retried, because the query may still be running. A host that failed is skipped for host_retry_after seconds, and the warning goes to the Logger driver. When every host has failed recently, they are all tried again.

Testing

composer install
vendor/bin/phpunit --testsuite Unit

# Integration tests run against ClickHouse in Docker, over HTTP and HTTPS:
tests/Integration/start-services.sh
vendor/bin/phpunit
docker compose -f tests/Integration/compose.yaml down -v

What is a Kipchak Driver?

Kipchak Drivers are used to connect Kipchak to various data sources or storage systems.

They provide a standardized interface for interacting with different data sources, allowing developers to focus on building their applications rather than dealing with the complexities of each data source.

Drivers are basically Container Dependencies injected into Kipchak's Service Container.

Licence

MIT