Bug: Serialize.writeInt64() truncates Long to Int causing data corruption

[jeremysong]

Description

The Serialize.writeInt64() method in com.clickhouse.utils.Serialize incorrectly truncates Long values to 32-bit integers before writing them as 64-bit values. This causes any Int64 value larger than Integer.MAX_VALUE (2,147,483,647) to be corrupted when inserted using RowBinary format.

Steps to reproduce

1. Create a ClickHouse table with a Nullable(Int64) column
2. Use Serialize.writeInt64() to write a value larger than 2,147,483,647 (e.g., timestamp 1704067200000L)
3. Insert the data using RowBinary format
4. Query the data back
5. The value is corrupted (e.g., 1704067200000 becomes -1034816512)

Error Log or Exception StackTrace

No exception thrown - data is silently corrupted.
Expected: 1704067200000
Actual: -1034816512

Expected Behaviour

Serialize.writeInt64() should correctly write the full 64-bit Long value without truncation. Values like 1704067200000L should be stored and retrieved correctly.

Code Example

// Bug in Serialize.java line 208-212
public static void writeInt64(OutputStream out, Long value, boolean defaultsSupport, 
                             boolean isNullable, ClickHouseDataType dataType, 
                             boolean hasDefault, String column) throws IOException {
   if (writeValuePreamble(out, defaultsSupport, isNullable, dataType, hasDefault, column, value)) {
       BinaryStreamUtils.writeInt64(out, convertToInteger(value));  // BUG: Should be convertToLong(value)
   }
}

// convertToInteger() truncates Long to Int
public static Integer convertToInteger(Object value) {
   if (value instanceof Number) {
       return ((Number) value).intValue();  // Loses upper 32 bits!
   }
   ...
}

Root Cause: Line 210 calls convertToInteger(value) which truncates the 64-bit Long to a 32-bit Integer using .intValue(). The truncated value is then sign-extended when written as 64 bits.

Example:

• Input: 1704067200000L (0x0000018C_C251F400)
• After convertToInteger(): -1034816512 (0xC251F400 as signed 32-bit)
• After sign-extension to 64-bit: 0xFFFFFFFF_C251F400 (-1034816512 as signed 64-bit)

Fix: Change line 210 to use convertToLong(value) instead:

BinaryStreamUtils.writeInt64(out, convertToLong(value));

Configuration

Client Configuration

ClickHouseClientConfig config = new ClickHouseClientConfig(
   "http://127.0.0.1:8123",
   "default",
   "",
   "default",
   "test_table"
);
Client client = config.createClient();

Environment

• Cloud
• Connector version: flink-connector-clickhouse-base (latest)
• Language version: Java 11
• OS: macOS / Linux

ClickHouse Server

• ClickHouse Server version: 24.x
• ClickHouse Server non-default settings, if any: None
• CREATE TABLE statements for tables involved:

CREATE TABLE test_int64 (
   id String,
   timestamp_value Nullable(Int64)
) ENGINE = MergeTree() ORDER BY id

• Sample data: Any Int64 value > 2,147,483,647 will reproduce the issue (e.g., millisecond timestamps like 1704067200000)

[mzitnik]

@jeremysong

Thanks