For part 1 I was happy with my one pass approach. Making it more generic for part 2 somewhat ruined the neatness. Rolling a column north just keeps track of the next valid destination as it goes, so no nested loops needed (swapping ranges with indexing for clarity): int64_t dst_row = 0; for (row...) { if (grid[row][col] == 'O') { swap(grid[dst_row++][col], grid[row][col]); } else if (grid[row][col] == '#') { dst_row = row + 1; } } Still can't get part under 19ms, though. All the copies to keep a history up to the first cycle must be killing it.
![Day 14: Parabolic Reflector Dish [AOC2023 Python]](/img/n.gif)
I like your videos a lot. thanks for sharing.👍
For part 1 I was happy with my one pass approach. Making it more generic for part 2 somewhat ruined the neatness. Rolling a column north just keeps track of the next valid destination as it goes, so no nested loops needed (swapping ranges with indexing for clarity):
int64_t dst_row = 0;
for (row...) {
if (grid[row][col] == 'O') {
swap(grid[dst_row++][col], grid[row][col]);
}
else if (grid[row][col] == '#') {
dst_row = row + 1;
}
}
Still can't get part under 19ms, though. All the copies to keep a history up to the first cycle must be killing it.
Why there is "-1" in calculating cycle number in part 2 on line 92? Can someone explain?
the problem uses 1-based indexing but I'm using 0-based indexing, so cycle 1000000000 is actually cycle 99999999.
Oh I see ❤ Thank you 🎉
Your explanation is not superb