datastore/tests/performance/limit_by.xml

<test>
    <substitutions>
        <substitution>
            <name>cardinality</name>
            <values>
                <value>1000</value>
                <value>100000</value>
            </values>
        </substitution>
    </substitutions>

    <!-- LIMIT with offset + length -->
    <query>
        SELECT number % {cardinality} AS k, number AS v
        FROM numbers_mt(10000000)
        LIMIT 5, 10 BY k
        FORMAT NULL
    </query>

    <!-- Expensive BY expression -->
    <query>
        SELECT number, concat(toString(number), '-', lowerUTF8(toString(number * 31)), '-', toString(number % 97)) AS bucket
        FROM numbers_mt(10000000)
        LIMIT 10 BY bucket
        FORMAT NULL
    </query>

    <!-- Skewed key distribution: ~90% of rows fall into a single hot key, interleaved with cold keys -->
    <query>
        SELECT if(number % 10 = 0, number % {cardinality}, 0) AS k, number AS v
        FROM numbers_mt(10000000)
        LIMIT 10 BY k
        FORMAT NULL
    </query>

    <!-- UInt8 BY key: narrow integer type, 256 distinct groups -->
    <query>
        SELECT toUInt8(number % {cardinality}) AS k, number AS v
        FROM numbers_mt(10000000)
        LIMIT 10 BY k
        FORMAT NULL
    </query>

    <!-- LowCardinality(String) BY key -->
    <query>
        SELECT toLowCardinality(toString(number % {cardinality})) AS k, number AS v
        FROM numbers_mt(10000000)
        LIMIT 10 BY k
        FORMAT NULL
    </query>

    <!-- Two String BY columns -->
    <query>
        SELECT toString(number % {cardinality}) AS k1, toString(number % 97) AS k2, number AS v
        FROM numbers_mt(10000000)
        LIMIT 10 BY k1, k2
        FORMAT NULL
    </query>
</test>