ywnsyage/clickhouse-php-client

Clickhouse client over HTTP and MySQL


License
GPL-3.0

Documentation

Clickhouse Client

Build Status Coverage Status

Package was written as client for Clickhouse.

Client uses Guzzle for sending Http requests to Clickhouse servers.

Requirements

php7.1

Install

Composer

composer require ywnsyage/clickhouse-php-client

Usage

Client works with alone server and cluster. Also, client can make async select and insert (from local files) queries.

Alone server

$server = new Ywnsyage\Clickhouse\Server('127.0.0.1', '8123', 'default', 'user', 'pass');
$serverProvider = (new Ywnsyage\Clickhouse\ServerProvider())->addServer($server);

$client = new Ywnsyage\Clickhouse\Client($serverProvider);

Cluster

$testCluster = new Ywnsyage\Clickhouse\Cluster('cluster-name', [
    'server-1' => [
        'host' => '127.0.0.1',
        'port' => '8123',
        'database' => 'default',
        'user' => 'user',
        'password' => 'pass'
    ],
    'server-2' => new Ywnsyage\Clickhouse\Server('127.0.0.1', '8124', 'default', 'user', 'pass')
]);

$anotherCluster = new Ywnsyage\Clickhouse\Cluster('cluster-name', [
    [
        'host' => '127.0.0.1',
        'port' => '8125',
        'database' => 'default',
        'user' => 'user',
        'password' => 'pass'
    ],
    new Ywnsyage\Clickhouse\Server('127.0.0.1', '8126', 'default', 'user', 'pass')
]);

$serverProvider = (new Ywnsyage\Clickhouse\ServerProvider())->addCluster($testCluster)->addCluster($anotherCluster);

$client = (new Ywnsyage\Clickhouse\Client($serverProvider));

Before execute any query on cluster, you should provide cluster name and client will run all queries on specified cluster.

$client->onCluster('test-cluster');

By default client will use random server in given list of servers or in specified cluster. If you want to perform request on specified server you should use using($hostname) method on client and then run query. Client will remember hostname for next queries:

$client->using('server-2')->select('select * from table');

Server tags

$firstServerOptionsWithTag = (new \Ywnsyage\Clickhouse\Common\ServerOptions())->setTag('tag');
$secondServerOptionsWithAnotherTag = (new \Ywnsyage\Clickhouse\Common\ServerOptions())->setTag('another-tag');

$server = new Ywnsyage\Clickhouse\Server('127.0.0.1', '8123', 'default', 'user', 'pass', $firstServerOptionsWithTag);

$cluster = new Ywnsyage\Clickhouse\Cluster('cluster', [
    new Ywnsyage\Clickhouse\Server('127.0.0.2', '8123', 'default', 'user', 'pass', $secondServerOptionsWithAnotherTag)
]);

$serverProvider = (new Ywnsyage\Clickhouse\ServerProvider())->addServer($server)->addCluster($cluster);

$client = (new Ywnsyage\Clickhouse\Client($serverProvider));

To use server with tag, you should call usingServerWithTag function before execute any query.

$client->usingServerWithTag('tag');

Select queries

Any SELECT query will return instance of Result. This class implements interfaces \ArrayAccess, \Countable и \Iterator, which means that it can be used as an array.

Array with result rows can be obtained via rows property

$rows = $result->rows;
$rows = $result->getRows();

Also you can get some statistic of your query execution:

  1. Number of read rows
  2. Number of read bytes
  3. Time of query execution
  4. Rows before limit at least

Statistic can be obtained via statistic property

$statistic = $result->statistic;
$statistic = $result->getStatistic();

echo $statistic->rows;
echo $statistic->getRows();

echo $statistic->bytes;
echo $statistic->getBytes();

echo $statistic->time;
echo $statistic->getTime();

echo $statistic->rowsBeforeLimitAtLeast;
echo $statistic->getRowsBeforeLimitAtLeast();

Sync

$result = $client->readOne('select number from system.numbers limit 100');

foreach ($result as $number) {
    echo $number['number'].PHP_EOL;
}

Using local files

You can use local files as temporary tables in Clickhouse. You should pass as third argument array of TempTable instances. instance.

In this case will be sent one file to the server from which Clickhouse will extract data to temporary table. Structure of table will be:

  • number - UInt64

If you pass such an array as a structure:

['UInt64']

Then each column from file wil be named as _1, _2, _3.

$result = $client->readOne('select number from system.numbers where number in _numbers limit 100', new TempTable('_numbers', 'numbers.csv', [
    'number' => 'UInt64'
]));

foreach ($result as $number) {
    echo $number['number'].PHP_EOL;
}

You can provide path to file or pass FileInterface instance as second argument.

There is some other types of file streams which could be used to send to server:

  • File - simple file stored on disk;
  • FileFromString - stream created from string. For example: new FileFromString('1'.PHP_EOL.'2'.PHP_EOL.'3'.PHP_EOL)
  • MergedFiles - stream which includes many files and merges them all in one. You should pass to constructor file path, which contains list of files which should be megred in one stream.
  • TempTable - wrapper to any of FileInterface instance and contains structure. Usefull to make inserts using with MergedFiles.

Async

Unlike the readOne method, which returns Result, the read method returns an array of Result for each executed query.

list($clicks, $visits, $views) = $client->read([
    ['query' => "select * from clicks where date = '2017-01-01'"],
    ['query' => "select * from visits where date = '2017-01-01'"],
    ['query' => "select * from views where date = '2017-01-01'"],
]);

foreach ($clicks as $click) {
    echo $click['date'].PHP_EOL;
}

In read method, you can pass the parameter $concurrency which is responsible for the maximum simultaneous number of requests.

Using local files

As with synchronous select request you can pass files to the server:

list($clicks, $visits, $views) = $client->read([
    ['query' => "select * from clicks where date = '2017-01-01' and userId in _users", new TempTable('_users', 'users.csv', ['number' => 'UInt64'])],
    ['query' => "select * from visits where date = '2017-01-01'"],
    ['query' => "select * from views where date = '2017-01-01'"],
]);

foreach ($clicks as $click) {
    echo $click['date'].PHP_EOL;
}

With asynchronous requests you can pass multiple files as with synchronous request.

Insert queries

Insert queries always returns true or throws exceptions in case of error.

Data can be written row by row or from local CSV or TSV files.

$client->writeOne("insert into table (date, column) values ('2017-01-01',1), ('2017-01-02',2)");
$client->write([
    ['query' => "insert into table (date, column) values ('2017-01-01',1), ('2017-01-02',2)"],
    ['query' => "insert into table (date, column) values ('2017-01-01',1), ('2017-01-02',2)"],
    ['query' => "insert into table (date, column) values ('2017-01-01',1), ('2017-01-02',2)"]
]);

$client->writeFiles('table', ['date', 'column'], [
    new Ywnsyage\Clickhouse\Common\File('/file-1.csv'),
    new Ywnsyage\Clickhouse\Common\File('/file-2.csv')
]);

$client->insertFiles('table', ['date', 'column'], [
    new Ywnsyage\Clickhouse\Common\File('/file-1.tsv'),
    new Ywnsyage\Clickhouse\Common\File('/file-2.tsv')
], Ywnsyage\Clickhouse\Common\Format::TSV);

In case of writeFiles queries executes asynchronously. If you have butch of files and you want to insert them in one insert query, you can use our ccat utility and MergedFiles instance instead of File. You should put list of files to insert into one file:

file-1.tsv
file-2.tsv

Building ccat

ccat sources placed into utils/ccat directory. Just run make && make install to build and install library into bin directory of package. There are already compiled binary of ccat in bin directory, but it may not work on some systems.

In writeFiles method, you can pass the parameter $concurrency which is responsible for the maximum simultaneous number of requests.

Other queries

In addition to SELECT and INSERT queries, you can execute other queries :) There is statement method for this purposes.

$client->writeOne('DROP TABLE table');

Testing

$ composer test

Roadmap

  • Add ability to save query result in local file

Contributing

Please send your own pull-requests and make suggestions on how to improve anything. We will be very grateful.

Thx!