Advent of Code 2023 Day 5 (BaseA Style)

Discussion thread for day 5 of the Advent of Code - https://adventofcode.com/2023/day/5

Accepted answers

All comments

My brain is dead, but boy was that a fun challenge and a great puzzle! Thinking through how to solve a problem of this magnitude was very exciting (... although I almost crashed my computer, jumping straight to the Generate Rows Tool... Next time I'll check how big the numbers are).

Spoiler

First problem used a standard macro:
Standard Macro.png

Second Part an Iterative Macro:
Iterative Macro.png

Both in a beautiful cascading chain (Main Workflow):
Main Screenshot.png

Solution.yxzp

Tokimatsu

There may be many problems, but I solved them anyway.

Spoiler

Gawa

Please don't ask me how my WF works, because I can't explain😂

0.9sec with AMP engine.

Spoiler

I'm so tired...

AoC2023_Day5_gawa.yxzp

Gawa

I refined my WF, and now, I can explain how it works. Happy to achieve less than 1 sec!

Spoiler

AoC2023_Day5_gawa_rev1.yxzp

DaisukeTsuchiya

Part2 was so hard.

Spoiler

I spend a lot for Part2. I was suprised that the answer was matched luckly in the middle of development as a trial.
I wonder why it matched then I carefully checked the map file and got the concept.

I skip the complicated part in the workflow then add the records by investigating the map in Excel. My workflow needs to be configured when input is changed...

=> I revisit WF and update. Now I have a clear picture.

スクリーンショット 2023-12-06 094703.png

< Macro Part2>

patrick_digan

I feel good about my part 1 solve using my multi-row-field-column tool, but part 2 is one of those that you just find a way to solve and move on!

Spoiler

Part 1:

Part 2:

Day 5.yxmd

Day 5 part2.yxmd

phottovy

Attached is my sort of cleaned up workflow:

Spoiler

For part 1, I could have built an iterative macro but ended up reusing containers to accomplish the same thing.

For part 2, I had to adjust my approach quite a bit. There were simply way too many rows to send through my workflow so I eventually switched to a sampling approach. I would send random rows from each seed through my workflow and analyze which seeds and individual rows consistently had the lowest result. Eventually, I was able to isolate which seed had the lowest result and I was able to start sending ranges of approximately 250K- 500K rows through my workflow until I was able to isolate the lowest outcome.

aoc_ph_2023_05 final.yxzp

mmontgomery

No macro and 1 second to run. Ironically got p2 in a much faster time than P1.

Spoiler

P2 was a matter of finding a trend in the data at a large scale and then doing subsamples. I plotted random points and noticed it was always going up except for random drops. Once I found enough random drops, I found the min of those and then tested smaller subsets until I got it. ~1.5 hours for p1. ~30 min for p2

2023_Day5.yxzp

DavidP

Had to rethink my strategy for part 2 to think in ranges rather than lists. Rough steps are:

1. Check if start of seed range is within one of the source ranges

2. If true, check if the end is within that source range

3. If yes, happy days, assign the offset for the next level

4. If no, break the range into two - the part that fits and the remainder

5. The remainder forms a new range on which to preform steps 1-4.

I used an iterative macro for steps 1-4 and then ran it for each mapping. Runtime 0.5 secs

Spoiler

AoC day5 part2 iterative.yxmc

Day5_2.yxmd

Yoshiro_Fujimori

With a hint from @gawa , I could not come to the solution of Part 2 before going to bed.

This morning I found a way and it worked.

Thanks @gawa (and my good sleep)!

Spoiler

Main

Prep Macro

Mapping Macro

AoC05_Part2_YFujimori.yxzp

DaisukeTsuchiya

I revisit my WF and update. I got the clear picture now. WF completes in 0.7sec.

Spoiler

<WF>

<Macro Part1>
スクリーンショット 2023-12-05 230635.png

<Macro Part2>
スクリーンショット 2023-12-05 230659.png

estherb47

My brain definitely hurts after this one......

ccanterman

Solved. Part 1 worked beautifully but Part 2 failed spectacularly (those Joins seemed like such a good idea during Part 1), so I ended up rebuilding the workflow just for Part 2 with an Iterative Macro (which I probably should've done in the first place). Good fun

Spoiler

FrederikE

Pure brute force. The workflow ran all night and finished 20 min before the AoC Day ended.

Screenshot 2023-12-06 053904.png

Spoiler

5/5 so far.

05.yxzp

cgoodman

Spoiler

Brutal
Tried to brute force part 2, then spend a crazy amount of time trying to optimise.
Finally in the evening attempted the approach of creating ranges with big jumps, and then finding the seed number and then doing +1 increments around this value, which resulted in a fast solve.

And yes I probably have should have put this in a macro!

AncientPandaman

Spoiler

workflow

macro for part 1
Screenshot 2023-12-06 140445.png

macro for part 2:
1. split to group based on the seed range
2. update the destination, source, and range
3. go for next soil etc.

i did not update the previous range, because it not required.
Screenshot 2023-12-06 140449.png

Day 05.yxzp

AkimasaKajitani

My solution.

Spoiler

I completed part 2 just before sleeping. This is that workflow.

To be the honest, I was going to do debug more time, but I got the correct result. Some of this is not clear to me, but I'll leave it at that for now. But the workflow above is messy, so I remade it using Iterative macro.

Part 1 is normal brute force. But that was not working for Part 2. So I made it using range approach.

Part 1 macro :

Part 2 macro :

mceleavey

I did it.

I now hate the world.

jarrodthuener

Day 5 Part 2 was tricky even with the right "logic". Ranges are easy to conceptualize, but a bit harder to implement... but when it finally works, it solves both parts in 0.6 seconds. 😁

Spoiler

Day5.yxzp

JeffF

Brute force.

Spoiler

starkey

@gawa Nice work on this! By far the most elegant solution I've seen for this challenge.

alisonpitt

My heart leapt out of my chest when I ran my first try of part 2 without looking how big the numbers were 😱 But handling the ranges of numbers ended up being a really interesting exercise. I feel like I probably could have used a macro for this, but with the number of steps, I went with brute-force.

Spoiler

Day 5 - AP.yxmd

Qiu

At first I believed brutal force and finally getting the result as below. It was an epic failure.
Then moved to the Range comparing approach, with some hints from here, I got the answer and I decided to move on even I can not explain the logic in my flow.

Interesting thing is that I am still getting the incorrect answer when I use my input with the most of workflows uploaded here. 😁
So I guess its just something we can forget and leave behind and continue our life, continue to be tortured by Day 7 or Day xx. 😅 (Day 6, you are nice), continue to hope to retire to beach one day. @mceleavey
giphy

Spoiler

kelsey_kincaid

I was on the cusp of solving this one for... many many hours. @PangHC 's workflow looked most similar to my approach, so I ended up having to open that up to help me troubleshoot why I was getting the right answer on the sample data but not my prod data. It was super helpful - even though I was getting the right answer on the sample data, my approach was wrong, and Pang_Hee_Choy's mapping table helped me identify those issues. Thank you for your work!

Spoiler

Part 1: I started with my favorite - a brute force generate rows approach. Shame on me for not first looking at my prod data! I had to re-tool to evaluate ranges instead. I did that by building an iterative macro that loops through each mapping table to find the matches.

Part 2: Beat my head against a wall for a long time on this. I didn't read the problem closely enough and wasn't subtracting 1 from the range in a number of calculations. I also had to tweak the way I was handling matches within partial ranges. Since locations without any mapping to previous values was my final output, I had a hard time tracing my work for troubleshooting. Lesson learned there too I think. I may need another night of sleep before I can fully explain how this works.

Workflow:

Part 1 Macro:

Part 2 Macro:

AOC2023_Day5_KelseyKincaid.yxzp

StevenP

While going the route of brute force may have gotten me to the answer faster I refused to give in and kept banging my head to get this to run efficiently. Ultimately I succeeded and got my overall run time down to 1.1 seconds.

giphy

Spoiler

For Part 2 I reverse-engineered the location back into the seed based on the given ranges:

Day 5 If You Give A Seed A Fertilizer.yxzp

SuguruYoshinaga

Part 2 was very difficult.

Spoiler

Macro for part 1

Macro for part 2

day5.yxzp

leonhekkert

Actually quite proud that I got the solution to part 2! Did have to start all over though as my approach to part 1 was the brute force kind.

Part two is done with a batch and iterative macro and takes less then 2 seconds!

Spoiler

Day5.yxmd

I ran into the same trap as everyone. My is definitely Ugly and I went the range route.

Advent2023Day5.yxmd

Laurap1228

Spoiler

I built two Macros for this one.

Advent of Code 2023_5.yxzp

SeanAdams

Finally got this done (took a few days to clear out the brain fog).

And built a good few AOC reusable macros in the process!

Spoiler

The recipe for part 2 is as follows:
- Do the prep and break stuff down.
- Create a loop for each block of translation
- For each iteration - split the seed ranges based on overlaps with the mapping ranges. There's a trick here - you take the mapping ranges and break them into break points, and then iteratively break up each of the seed ranges
- For each of the newly subdivided ranges - look to see if there's a hit on the mappings and then update the ranges
- Finally compress the ranges to get rid of overlaps

Screenshot 2023-12-09 210220.png

2023_Day05-Part2.yxzp

DaisukeTsuchiya

I post again with solution in yxdb.

Spoiler

Day5_Rev4.yxzp

Quick Links

This months top contributors

atcodedog05 19458

Qiu 15866

binu_acs 15708

MarqueeCrew 13708

apathetichell 13703