Reverse multi-source reachability

Route Water to Both Drain Edges

Original & variantPacific Atlantic Water Flow ↗

Original scenario and examples. The task states the exact input adapters and deterministic output rules; serialization, cloning and operation sequences return observable JSON results.

01. The question

Water can move to an orthogonal neighbor of equal or lower height. Return cells that can drain to both top/left and bottom/right boundary groups. Sort [row,column] pairs in row-major order.

Drainage heights3 rows × 3 columns
r / c012
0123
1232
2321

Top/left and bottom/right are the two drainage boundary groups.

Before revealing the idea, predict the result and describe what information you would need to track.

02. Worked example

Input[[1,2,3],[2,3,2],[3,2,1]]

Output[[0,2],[1,1],[2,0]]

  1. Start one search from each boundary group.
  2. Walk to neighbors at least as high.
  3. Intersect the reachable sets.

Try your code

Define solve(data, parameter="") and return your result. Your code runs locally, not on a server.

Edit your test cases

Use 1-10 objects with data and an optional string parameter. Expected outputs are computed from the lesson. An optional expected field must be correct.

Input contract & reference code

Example data: [[1,2,3],[2,3,2],[3,2,1]]. Parameter: "". Numeric parameters arrive as strings; convert them with int(parameter) when needed.

The example and generated cases use the question’s exact variant. This checks behavior on a small test batch—not performance on every possible input.

Python runtime notices

Python downloads only when you run code, using compressed runtime assets. Code is not saved or uploaded. Enable JavaScript to run it.

Five-second execution limit per batch. Standard-library Python only; additional packages and network access are unavailable. Stop ends execution without blocking the lesson.

03. The idea

Try sketching your own approach first. What would you track, and why would it work?

Reveal the idea
  1. Start one search from each boundary group.
  2. Walk to neighbors at least as high.
  3. Intersect the reachable sets.

04. Watch it happen

Use Step to move slowly, or Play to follow the whole trace. Change a preset or load your own input.

Input & settings

Small verified inputs: at most 15 binary slots, 10 graph vertices, 12 trie operations or a 4×4 board. Binary children use indexed slots 2i+1 and 2i+2. See the task for the object format.

Red = current item / blue = active window or cells / green = completed work. A value docks at its actual destination; the original input stays unchanged. Positions start at 0.

Working state / what changes

Ready to begin

Enable JavaScript to step through the animation. The worked example and solution below remain readable without it.

Python / why it changes

The line behind the step

def solve(data, parameter=""):
    rows, cols = len(data), len(data[0])
    def reach(starts):
        seen = set(starts)
        queue = list(starts)
        for r, c in queue:
            for dr, dc in [(1, 0), (-1, 0), (0, 1), (0, -1)]:
                nr, nc = r + dr, c + dc
                if 0 <= nr < rows and 0 <= nc < cols and (nr, nc) not in seen and data[nr][nc] >= data[r][c]:
                    seen.add((nr, nc))
                    queue.append((nr, nc))
        return seen
    first = reach([(0, c) for c in range(cols)] + [(r, 0) for r in range(rows)])
    second = reach([(rows - 1, c) for c in range(cols)] + [(r, cols - 1) for r in range(rows)])
    return [list(cell) for cell in sorted(first & second)]
How to use this walkthrough

Hover or focus a control for its label. Play forms the input and follows values into the working state. Step shows one change; Back and the slider restore exact snapshots. Reduced motion uses instant updates. Saved snapshots and sorted heap displays are teaching overhead, not algorithm complexity.

05. Python solution

def solve(data, parameter=""):
    rows, cols = len(data), len(data[0])
    def reach(starts):
        seen = set(starts)
        queue = list(starts)
        for r, c in queue:
            for dr, dc in [(1, 0), (-1, 0), (0, 1), (0, -1)]:
                nr, nc = r + dr, c + dc
                if 0 <= nr < rows and 0 <= nc < cols and (nr, nc) not in seen and data[nr][nc] >= data[r][c]:
                    seen.add((nr, nc))
                    queue.append((nr, nc))
        return seen
    first = reach([(0, c) for c in range(cols)] + [(r, 0) for r in range(rows)])
    second = reach([(rows - 1, c) for c in range(cols)] + [(r, cols - 1) for r in range(rows)])
    return [list(cell) for cell in sorted(first & second)]

06. Cost & edge cases

Time
O(RC)
Space
O(RC)
  • Validate the exact input format and bounds before tracing.
  • Equal labels do not collapse vertex identity; missing nodes and invalid pointers are explicit.

Keep going