Enhance insert() API to support SELECT with input() for computed columns v2 #2726
Description
Current Limitation
The insert() method only supports direct table insertion, not SELECT statements
with transformations.
Desired Enhancement
Allow insert() to accept SELECT statements with input() function:
client.insert()
.table("target_table")
.query("SELECT col1, col2, col1 + col2 as computed FROM input('col1 Int32, col2 Int32')")
.data(inputStream)
.format(ClickHouseFormat.CSV)
.send();Or alternative syntax:
client.insert("INSERT INTO target_table SELECT col1, col2, col1 + col2 as computed FROM input('col1 Int32, col2 Int32')")
.data(inputStream)
.format(ClickHouseFormat.CSV)
.send();Use Case
We need to stream data with computed columns during insertion:
- Stream large datasets efficiently
- Apply transformations/computed columns on the fly
- Leverage ClickHouse's input() function capabilities
- Maintain the performance benefits of the streaming insert() API
Business Value
This is a common pattern in ClickHouse for ETL workloads where data needs
transformation during insertion without materializing intermediate results.
Current Workaround (Working but Undocumented)
We're successfully using this pattern in production:
String loadQuery = tableName + "(" + headers + ")" +
" select " + parametersQuery + "* from input('" + csvHeadersStructure + "')";
try (InputStream inputStream = Files.newInputStream(Paths.get(filepath))) {
InsertResponse response = client.insert(loadQuery, inputStream, ClickHouseFormat.CSVWithNames)
.get(1200, TimeUnit.SECONDS);
}This works but:
- Not documented in official API docs
- Unclear if this is intended behavior or might break in future versions
- No IDE autocomplete/type safety for this pattern