Yong Jie Fong

Data engineer at Jupiter. I work on real-time analytics, distributed systems, and ClickHouse.

Writing

Stop polling ClickHouse for changes

Using materialized views, HTTP, and Kafka to turn inserts into recomputation signals.

Protecting early LIMIT in ClickHouse

A before-and-after rewrite that separates page selection from optional enrichment.

A ClickHouse TTL is scheduled merge work

Why retention and rollups create merge pressure—and when the cost is worth it.

FINAL vs argMax in ClickHouse

Choosing query-time deduplication by part overlap, key cardinality, and correctness.

When IN beats a JOIN in ClickHouse

Why a small IN filter pruned granules that a hash join still read.

Ordering keys, projections, or skip indexes?

Choose each physical structure by access pattern and write cost.

About

I lead the data team at Jupiter. This site is where I write down notes on ClickHouse and data systems. Examples use synthetic data and simplified schemas.