Introduce validation to the ColumnArray constructor

[IyeOnline]

The ColumnArray constructor would just accept incompatible value- and
offset-arrays and the client would then happily send them, causing
issues in the ClickHouse server.

This introduces validation of the arrays.

Fixes #426

[CLAassistant]

All committers have signed the CLA.

[IyeOnline]

Code that triggers the issue and would fail without the change

#include <clickhouse/client.h>

using namespace clickhouse;

/// Produces the offset array for a column of `n_rows`, which each row being an
/// array of 2 elements
auto make_offsets(const size_t n_rows) -> std::shared_ptr<ColumnUInt64>
{
    auto column_offsets = std::make_shared<ColumnUInt64>();
    for ( size_t i=0; i < n_rows; ++i) {
      column_offsets->Append(i*2);
    }
    return column_offsets;
}

/// Produces non-monotonic offsets that still satisfy the protocol
auto make_broken_offsets(const size_t n_rows) -> std::shared_ptr<ColumnUInt64>
{
    auto column_offsets = std::make_shared<ColumnUInt64>();
    for ( size_t i=1; i < n_rows+1; ++i) {
      column_offsets->Append(i%2 ? i : i*2);
    }
    return column_offsets;
}


/// Produces the values array for a column of `n_rows`, which each row being an
/// array of 2 elements
auto make_values(const size_t n_rows) -> std::shared_ptr<ColumnString>
{
    auto column_values = std::make_shared<ColumnString>();
    for ( size_t i=0; i < n_rows; ++i) {
      column_values->Append("some_long_string "+std::to_string(2*i));
      column_values->Append("some_long_string "+std::to_string(2*i+1));
    }
    return column_values;
}

/// Simply sends the column array as a block
void send(Client &client, std::shared_ptr<ColumnString> values, std::shared_ptr<ColumnUInt64> offsets)
{
    Block block;
    auto column = std::make_shared<ColumnArray>(std::move(values), std::move(offsets));
    block.AppendColumn("col", std::move(column));
    client.Insert("crash_showcase", block);
}


int main()
{
    Client client(ClientOptions().SetHost("localhost"));
    client.Execute("CREATE TABLE IF NOT EXISTS crash_showcase (col Array(String)) ENGINE = Memory");

    /// Missing elements in the value array causes a timeout in the clickhouse server
    /// Solutions:
    /// * The client should at least do some rudimentary checks of offset array vs
    ///    the data array
    /// * The server should presumably do some verification of the data it gets
    send( client, make_values(5), make_offsets(10));


    /// Having too many elements in the data array causes the server to just start
    /// interpreting the excess values as new messages, which is really bad
    /// Solutions:
    /// * The client should at least do some rudimentary checks of offset array vs
    ///    the data array
    /// * The server should absolutely verify that there is not more elements than
    ///   expected
    send( client, make_values(6000),  make_offsets(100));


    /// The protocol at the very least specifies that the last offset is the column count:
    /// https://clickhouse.com/docs/native-protocol/columns#array
    /// While non-monotonic offsets *seem to work* (at least no error happens when sending), I am not sure about that.
    send( client, make_values(10),  make_broken_offsets(10));
}

[Copilot]

Pull Request Overview

This PR adds validation logic in the ColumnArray constructor to ensure that the provided value and offset arrays are compatible, preventing inconsistent data from being sent to the ClickHouse server.

• Added a check to ensure that the offsets array is monotonically increasing.
• Validated that the size of the data array matches the expected number of values derived from the offsets.

[mshustov]

@IyeOnline could you check the failing tests, please?