Solve 2023:15 "Lens Library"
WALLOFTEXT for part 2, took me 90 minutes to find this important text: > Each step begins with a sequence of letters that > indicate the label of the lens on which the step > operates. The result of running the HASH algorithm > on the label indicates the correct box for that > step. It also clarifies how part 2 and part 1 relates.
This commit is contained in:
parent
c832e30dcf
commit
a90269f7f9
1 changed files with 63 additions and 0 deletions
63
2023-python/output/day_15.py
Normal file
63
2023-python/output/day_15.py
Normal file
|
|
@ -0,0 +1,63 @@
|
|||
from collections import OrderedDict, defaultdict
|
||||
|
||||
from output import answer
|
||||
|
||||
n = 15
|
||||
title = "Lens Library"
|
||||
|
||||
|
||||
@answer(1, "Sum of HASH algorithm results: {}")
|
||||
def part_1(presolved):
|
||||
return presolved[0]
|
||||
|
||||
|
||||
@answer(2, "Focusing power of the resulting configuration: {}")
|
||||
def part_2(presolved):
|
||||
return presolved[1]
|
||||
|
||||
|
||||
def presolve(data):
|
||||
def h(s):
|
||||
v = 0
|
||||
for a in s:
|
||||
if a == "\n":
|
||||
continue
|
||||
v += ord(a)
|
||||
v *= 17
|
||||
v = v % 256
|
||||
return v
|
||||
|
||||
p1 = sum(h(c) for c in data.split(","))
|
||||
|
||||
b = defaultdict(OrderedDict)
|
||||
for lr in data.split(","):
|
||||
if "=" in lr:
|
||||
l, r = lr.split("=")
|
||||
if r == "":
|
||||
continue
|
||||
k = h(l)
|
||||
b[k][l] = r
|
||||
if "-" in lr:
|
||||
l, _r = lr.split("-")
|
||||
k = h(l)
|
||||
if l in b[k]:
|
||||
del b[k][l]
|
||||
p2 = 0
|
||||
for i, c in b.items():
|
||||
for j, f in enumerate(b[i].values(), 1):
|
||||
p2 += (i + 1) * j * int(f)
|
||||
|
||||
return p1, p2
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
with open("./input/16.txt", "r") as f:
|
||||
inp = f.read().strip()
|
||||
|
||||
inp = presolve(inp)
|
||||
|
||||
a = part_1(inp)
|
||||
b = part_2(inp)
|
||||
|
||||
assert a == 509784
|
||||
assert b == 230197
|
||||
Loading…
Add table
Reference in a new issue