Topological Sort for Ads Tasks

You are given numTasks tasks labeled from 0 to numTasks - 1 and a list of prerequisite edges. Each edge [before, after] means task before must be completed before task after.
Return one valid order to complete all tasks. If more than one task is available at the same time, choose the smaller task id first so the result is deterministic. If it is impossible to complete all tasks because the dependency graph contains a cycle, return an empty array.

Function
findTaskOrder(numTasks: int, edges: int[][]) → int[]

Examples
Example 1
numTasks = 4
edges = [[0, 1], [0, 2], [1, 3], [2, 3]]
return = [0, 1, 2, 3]
Task 0 unlocks tasks 1 and 2. The smaller available id is chosen first.
Example 2
numTasks = 2
edges = [[0, 1], [1, 0]]
return = []
The two tasks depend on each other, so no valid topological order exists.

Constraints
0 <= before, after < numTasks
Use topological sorting with cycle detection.
When multiple tasks have in-degree zero, choose the smallest task id first.

Build an adjacency list and an in-degree array from the edges. Push every task with in-degree zero into a min-heap. Pop the smallest id, append it to the result, then decrement the in-degree of each neighbor. Any neighbor that hits zero goes into the heap. When the heap empties, check the result length. If it equals numTasks, return it. Otherwise there's a cycle, so return an empty array. The common pitfall is using a regular FIFO queue. That gives a valid order but not the smallest-id-first one, so the expected output won't match. Another miss is forgetting the cycle check, or building the edge direction backwards. Edge [before, after] means before points to after. Complexity is O((V + E) log V). If you freeze on the heap detail during the live OA, StealthCoder is the hedge that gets you unstuck.

If you see this problem in your OA tomorrow, the play is to recognize the pattern in 30 seconds. StealthCoder buys you that recognition.