Pseudocode → Java

Pseudocode is under-specified on purpose. The Java you produce has to commit to everything pseudocode leaves open: types, nullability, error handling, mutability, collection implementations.

Decisions pseudocode doesn't make (but Java requires)

Pseudocode says	Java must decide
let S be a set	HashSet? TreeSet? LinkedHashSet? (Does order matter?)
x ← lookup(k)	Returns null on miss? Optional<V>? Throws?
list of numbers	int[]? List<Integer>? IntStream? (Boxed vs primitive matters.)
error: ...	Checked Exception? RuntimeException? Return sentinel?
for each x in S	Enhanced for? Stream? (Mutation during iteration → ConcurrentModificationException.)
procedure f(x)	Static method? Instance method? What class does it live on?

Default choices (absent a reason to deviate):

• Sets/maps → HashSet/HashMap unless order is mentioned; LinkedHash* if insertion order matters; Tree* if sorted iteration is used.
• Missing lookup → Optional for new code, null if matching an existing API.
• Number list → List<Integer> for flexibility, int[] if the pseudocode indexes and size is fixed.
• Errors → IllegalArgumentException for bad inputs, IllegalStateException for bad state, checked only if the caller can meaningfully recover.
• Home for the method → a final utility class with static methods, unless there's obvious object state.

Type inference from pseudocode

Pseudocode variables are used before typed. Infer from usage:

Usage in pseudocode	Inferred Java type
x ← 0; x ← x + 1	int (or long if values look big)
x ← []; x.append(y)	List<typeof(y)> (ArrayList)
x ← {}; x[k] ← v	Map<typeof(k), typeof(v)>
x ← ∅; x ← x ∪ {y}	Set<typeof(y)>
for i ← 1 to n	int i (1-indexed — watch off-by-one)
x ← head(list); rest(list)	Consider Deque or LinkedList

Worked example

Pseudocode (Dijkstra, textbook style):

function shortest_paths(G, source):
    dist ← map with dist[source] = 0, all others = ∞
    visited ← ∅
    Q ← priority queue of all vertices, keyed by dist

    while Q is not empty:
        u ← extract-min(Q)
        visited ← visited ∪ {u}
        for each neighbor v of u where v ∉ visited:
            alt ← dist[u] + weight(u, v)
            if alt < dist[v]:
                dist[v] ← alt
                decrease-key(Q, v, alt)

    return dist

Java:

import java.util.*;

public final class Dijkstra {
    private Dijkstra() {}

    public record Edge(int to, int weight) {}

    /**
     * Computes shortest distances from {@code source} to all reachable vertices.
     * @param adj adjacency list: adj.get(u) is the outgoing edges from u
     * @param source starting vertex
     * @return map from vertex to shortest distance; unreachable vertices absent
     * @throws IllegalArgumentException if source is not in adj
     */
    public static Map<Integer, Integer> shortestPaths(
            Map<Integer, List<Edge>> adj, int source) {
        if (!adj.containsKey(source)) {
            throw new IllegalArgumentException("source vertex not in graph");
        }

        Map<Integer, Integer> dist = new HashMap<>();
        dist.put(source, 0);
        Set<Integer> visited = new HashSet<>();

        // Java's PriorityQueue has no decrease-key. Standard workaround:
        // insert duplicates, skip stale entries on poll.
        PriorityQueue<int[]> pq = new PriorityQueue<>(Comparator.comparingInt(a -> a[1]));
        pq.offer(new int[]{source, 0});

        while (!pq.isEmpty()) {
            int[] top = pq.poll();
            int u = top[0], d = top[1];
            if (visited.contains(u)) continue;          // stale entry — already settled
            visited.add(u);

            for (Edge e : adj.getOrDefault(u, List.of())) {
                if (visited.contains(e.to())) continue;
                int alt = d + e.weight();
                Integer cur = dist.get(e.to());
                if (cur == null || alt < cur) {
                    dist.put(e.to(), alt);
                    pq.offer(new int[]{e.to(), alt});   // "decrease-key" = insert new, ignore old
                }
            }
        }
        return dist;
    }
}

Decisions made (and why):

• ∞ → absent from map. dist.get(v) == null means infinity. Avoids Integer.MAX_VALUE overflow when adding weights.
• decrease-key → Java's PQ doesn't have it. Lazy deletion: insert duplicates, skip visited on poll. O((V+E) log V) — same asymptotic.
• int[]{vertex, dist} in PQ instead of a record — cheap, no boxing. Comment explains.
• adj.getOrDefault(u, List.of()) — graph might have vertices with no outgoing edges.
• Unreachable vertices absent from result — documented. Caller checks containsKey.

Preserving annotations from verified pseudocode

If the pseudocode came from → verified-pseudocode-extractor, it has // PRE:, // POST:, // INV: annotations. Keep them as Javadoc and comments:

• PRE → @throws IllegalArgumentException if ... or @param ... must be ...
• POST → @return clause
• INV → comment at the loop head (don't delete — it's why the loop is correct)

Do not

• Do not use Integer.MAX_VALUE for infinity without guarding additions. MAX_VALUE + weight overflows to negative. Use absent-from-map, or Long, or check before adding.
• Do not mutate a collection while iterating it with for-each. ConcurrentModificationException. Use an explicit iterator with .remove(), or iterate over a copy.
• Do not translate 1-indexed pseudocode without adjusting. for i ← 1 to n → for (int i = 0; i < n; i++) and every a[i] stays a[i] (not a[i-1]) — or keep 1-indexed loops and adjust all accesses. Pick one, consistently.
• Do not make everything a static utility method if the pseudocode has persistent state between calls. That's an object.

Output format

## Type decisions
| Pseudocode var | Java type | Why |
| -------------- | --------- | --- |

## Code
<java>

## Deviations from pseudocode
<decrease-key workaround, ∞ encoding, index base — anything where the Java isn't a direct reading>

## Javadoc
<pre/post conditions translated>