From 792d0ddc02177119f079b1dcda0662917f21c6aa Mon Sep 17 00:00:00 2001
From: cupcakearmy <hi@nicco.io>
Date: Tue, 13 Dec 2022 10:49:30 +0100
Subject: [PATCH] day 12

---
 2022/12/README.md      |  18 ++++++
 2022/12/python/main.py | 139 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 157 insertions(+)
 create mode 100644 2022/12/README.md
 create mode 100644 2022/12/python/main.py

diff --git a/2022/12/README.md b/2022/12/README.md
new file mode 100644
index 0000000..7fafbfb
--- /dev/null
+++ b/2022/12/README.md
@@ -0,0 +1,18 @@
+# 12
+
+What a ride...
+
+So first I tried without A-Star. Well that of course did not stand a chance against the actual dataset.
+
+I think my a start implementation is flawed, of the heuristics are not optimal. It starts in a straight line, but ends up checking every field anyways. So it's more like a Dijkstra. But it works.
+The difference between Manhattan and Euclidean distance as `h(x)` did not help two. Went form 6939 iterations to 6974.
+
+The second part was quick. I reversed the search, starting from the end, removed the heuristic, so ti's now a classic Dijkstra, and instead of searching for a specific node, i stopped at the first encounter of an `a` or `0` elevation.
+
+<details>
+  <summary>Solutions</summary>
+  <ol>
+    <li>497</li>
+    <li>492</li>
+  </ol>
+</details>
diff --git a/2022/12/python/main.py b/2022/12/python/main.py
new file mode 100644
index 0000000..df3d8e5
--- /dev/null
+++ b/2022/12/python/main.py
@@ -0,0 +1,139 @@
+#!/usr/bin/env python
+
+from collections import defaultdict
+from copy import deepcopy
+from dataclasses import dataclass
+from os.path import dirname, join
+
+# Day 12
+
+# Common
+
+
+def read_input(filename):
+    data = join(dirname(__file__), '..', filename)
+    with open(data) as f:
+        return f.read().strip()
+
+
+test = read_input('test.txt')
+data = read_input('input.txt')
+
+
+@dataclass(unsafe_hash=True)
+class Point:
+    y: int
+    x: int
+
+    def __add__(self, other: 'Point') -> 'Point':
+        return Point(self.y+other.y, self.x+other.x)
+
+    def distance(self, other: 'Point') -> int:
+        return int(((self.y-other.y)**2 + (self.x-other.x)**2)**(1/2))
+        return abs(self.y - other.y) + abs(self.x - other.y)
+
+
+class Map:
+    def __init__(self, grid: list[list[int]], start: Point, end: Point) -> None:
+        self.start = start
+        self.end = end
+        self.grid = grid
+
+        self._max_y = len(self.grid)
+        self._max_x = len(self.grid[0])
+        self._step: tuple[Point, ...] = (Point(0, 1), Point(0, -1), Point(1, 0), Point(-1, 0))
+
+        self._display = [
+            [chr(c + ord('a')) for c in y]
+            for y in self.grid
+        ]
+
+    def display_points(self, points: list[Point]):
+        display = deepcopy(self._display)
+        for p in points:
+            display[p.y][p.x] = '#'
+        print('\n'+'\n'.join([''.join(y) for y in display]))
+
+    def get_neighbors(self, point: Point):
+        points: list[Point] = []
+        for step in self._step:
+            delta = point + step
+            if delta.y > -1 and delta.y < self._max_y and delta.x > -1 and delta.x < self._max_x:
+                points.append(delta)
+        return points
+
+    def get_height(self, point: Point) -> int:
+        return self.grid[point.y][point.x]
+
+    def a_star(self, two: bool = False) -> int:
+        start = self.end if two else self.start
+        end = self.end
+        visited: list[Point] = []  # Our graph is not directed, so we need to keep track of already visited nodes
+        came_from: dict[Point, Point] = {}  # To reconstruct the path once done
+        g: dict[Point, int] = {start: 0}  # G Value for each node
+        h: dict[Point, int] = {}
+        to_consider: list[tuple[Point, float]] = [(start, g[start] + 0 if two else start.distance(self.end))]
+
+        # i = 0
+        while True:
+            to_consider = sorted(to_consider, key=lambda x: x[1])
+            current, f = to_consider.pop(0)
+
+            current_height = self.get_height(current)
+            if two:
+                if current_height == 0:
+                    end = current
+                    break
+            else:
+                if current == end:
+                    break
+
+            visited.append(current)
+            neighbours = self.get_neighbors(current)
+            for neighbour in neighbours:
+                delta = self.get_height(neighbour) - current_height
+                invalid = delta < -1 if two else delta > 1
+                if neighbour in visited or invalid:
+                    continue
+                tmp_g = g[current] + 1  # We only have simple 1 step costs in our graph
+                if neighbour not in g or tmp_g < g[neighbour]:
+                    came_from[neighbour] = current
+                    g[neighbour] = tmp_g
+                    if neighbour not in h:
+                        h[neighbour] = 0 if two else neighbour.distance(self.end)
+                    tmp_h = h[neighbour]
+                    tmp_f = tmp_g + tmp_h
+                    to_consider.append((neighbour, tmp_f))
+
+        l: list[Point] = [end]
+        while True:
+            current = l[-1]
+            if current not in came_from:
+                self.display_points(l)
+                return len(l) - 1
+            prev = came_from[current]
+            l.append(prev)
+
+    @staticmethod
+    def parse(data: str) -> 'Map':
+        offset = ord('a')
+        start: Point
+        end: Point
+        grid: list[list[int]] = []
+        for y, line in enumerate(data.splitlines()):
+            grid.append([])
+            for x, char in enumerate(line):
+                if char == 'S':
+                    start = Point(y, x)
+                    char = 'a'
+                if char == 'E':
+                    end = Point(y, x)
+                    char = 'z'
+                height = ord(char) - offset
+                grid[y].append(height)
+        return Map(grid, start, end)
+
+
+# Running
+m = Map.parse(data)
+print(m.a_star(two=True))