Collections and streams

1. "Walk me through a stream pipeline as a SQL query."

stream().filter(p) is WHERE p. .map(f) is SELECT f(x). .flatMap(f) is UNNEST. .collect(groupingBy(k, summingDouble(v))) is GROUP BY k with SUM(v). .sorted(c) is ORDER BY. .distinct() is DISTINCT. Intermediate ops are lazy. The terminal op runs the pipeline.

2. "map vs flatMap?"

map is 1→1 (project). flatMap maps each element to a stream and flattens the streams, so it is 1→many. Examples: split a CSV line into tokens, or explode an array column.

3. "How do you count occurrences per key?"

Use map.merge(key, 1, Integer::sum). It inserts 1 if the key is absent. Otherwise it adds 1. You can also use stream().collect(Collectors.groupingBy(k, Collectors.counting())). Avoid map.put(k, map.get(k)+1). On the first occurrence, get returns null, and the +1 throws an NPE.

4. "When is a parallel stream a bad idea?"

It is a bad idea in three cases:

• The input is small (the pool overhead is greater than the benefit).
• The lambda has side effects or shared mutable state.
• The reduce combiner is not associative.

All three give wrong or non-deterministic results. Real parallelism belongs to Spark and Flink, not to a parallelStream().

5. "Why doesn't my second use of a stream work?"

The terminal op consumes a stream. You cannot reuse it. Java reports "stream has already been operated upon or closed". Build a fresh stream from the source List. Or collect to a List and reuse that.

6. "List.of(...) then .add(...) throws. Why, and the fix?"

List.of returns an immutable list. If you change it, Java throws UnsupportedOperationException. When you need to add or remove elements, wrap it: new ArrayList<>(List.of(...)).