Advent of Code 2023 Day 3 (BaseA Style)

AlteryxCommunityTeam

Discussion thread for day 3 of the Advent of Code - https://adventofcode.com/2023/day/3

Advent of Code

Accepted answers

All comments

Gawa

I had to use Appendix Field...I mean brute force attack! (Appreciate AMP engine)

Spoiler

phottovy

Joins to the left of me, joins to the right, here I am stuck in the middle with you:

Spoiler

aoc_ph_2023_03.yxmd

kelsey_kincaid

Definitely spun my wheels, spent too much time overthinking, and reworked things too many times, but got there!

Spoiler

Part 1: Spent a fair amount of time figuring out the best way to split out the values in each row to keep adjacent digits together as a single value, but also determine each number's position in the row. To account for that I created way too many IDs - one for the Row, one for the value, one for the column... I'm embarrassed to say those aren't the only ones. I've learned that when I'm not sure what I'm doing I find comfort in creating IDs for some reason 😂. Separated numbers from symbols and did a brute force append to find all possible combinations of numbers & symbols to determine if they were adjacent.

Part 2: Luckily, I was able to use most of my Part 1 solution to figure out Part 2. Just had to isolate possible gears and then determine how many numbers were adjacent to each possible gear.

AOC2023_Day3.yxmd

AkimasaKajitani

My solution. The Map style question is too early on Day3.

Spoiler

I made the neighbor list which has not numeric string and then I joined the numeric list and neighbor list. The neighbor list also worked well at part 2, so it contribute to solving the part 2 efficiently.

スクリーンショット 2023-12-03 145349.png

NicoleJ

Oooof. Same, @phottovy . Same.
Full transparency - I did temporarily output data to Excel so I could conditionally format the data (parts were green, potential gears were red) just so I could wrap my head around what I was looking at... which might have actually led me to my "I wonder if this matters" aha! moment in the spoiler. So... thanks Excel?

Spoiler

I struggled with Part 2 until I was able to visualize the puzzle, and noticed that there were gears with parts on either side (rather than just above & below, as in the example), and I started wondering if that mattered/was impacting something with the wrong answers I was getting... and then I realized I was using a count distinct on the part row field instead of a count to identify cases where there were 2 parts to a gear... and I really can't put into logical words how I got there... but I figured out there was a problem and then I started annotating my steps until I could rationally explain what was happening in each one and then I visualized it and spotted something that got my spidey senses tingling and then I smacked my head because Count Distinct and Count are very much not the same thing.

BUT!! I got to use that new Summarize Product function for the 2nd day in a row!!! So all is well, and now I can go to bed. 😁

My sweet Excel visualization:

DaisukeTsuchiya

It was tough as Day3.

Spoiler

joshbennett

Spoiler

For Part 1, I identified the xy coordinates of each character. Then, I broke out numeric characters and grouped them into identifiers for each non-breaking numeric sequence within a given record and joined them back to the individual numeric characters, giving me the xy coordinates of each digit in each non-breaking numeric sequence. I then broke out the symbol characters and identified horizontal, vertical, and diagonal neighbor xy coordinates for each symbol, at which point I joined this to the xy coordinates of the non-breaking numeric sequences to identify every "part number" (i.e., numeric sequence having 1+ symbol neighbors), and then summed the part numbers to get the answer.

For Part 2, I used Summarize on the join results from Part 1 to identify asterisk symbols with exactly 2 neighboring part numbers and then multiplied each symbol's part numbers and summed the products across all records.

Screenshot 2023-12-02 232552.png

I wish there was a more intuitive way to handle grid operations in Alteryx, but it is more fun than it should be to figure out. Looking forward to seeing others' solution strategies.

aoc_2023_03_joshbennett.yxzp

clmc9601

This was a chewy challenge for Day 3. I enjoyed it without getting too frustrated!

Spoiler

@AkimasaKajitani your generate rows for finding adjacent coordinates is so elegant!

2023_AoC_clmc9601_Day_03.yxmd

Tokimatsu

We're still good.

Spoiler

IraWatt

This one was interesting. I found it more challenging then day 1 or 2. Great to see all the approaches here, of the half dozen ways I tried I never even considered generate rows!!

Spoiler

Part 1

Part 2

Day 3.yxmd

Yoshiro_Fujimori

My solution.

Spoiler

AoC03.yxmd

caltang

Hard coded some stuff but it works I guess! Forgot that REGEX_CountMatches exists when I started…

Also, special credit to @clmc9601 ! Her guide and previous AOC entry really helped me conceptualise this better. I added a comment to visualise the “surrounding” values of the special characters! Thanks Claire!!!!

Spoiler

AOC23_Day3_caltang.yxzp

suwenchuan

Day 3 down! Off to learning from others' solution 🚀

Spoiler

AOC Day 3 workflow.yxzp

FrederikE

It took me 30 min of troubleshooting to figure out my workflow was fine, but I had to turn AMP engine off.

Spoiler

03.yxmd

patrick_digan

Spoiler

Day 3.yxmd

Qiu

Not so pretty flow but get the number correctly. 😁

CoG

These workflows are more work and less flow. This was tough to think through, and made tougher by random inconveniences like me forgetting to check my data types.

Workflow Part2.yxmd

Workflow.yxmd

starkey

Spoiler

I definitely had a flashback to previous years of AOC when reading the prompt for this one! Keeping the data long and thin and making use of IDs was the key to helping me wrap my head around how to solve this one without over complicating it.

aoc_2023_day_3_jstarkey.yxzp

afv2688

For sure can be optimized but don't want to spend more time here XD

Spoiler

Went with mathematical aproach and looked for vector distance between special characters and numbers
Sin título.png

mmontgomery

A bit wonky, but it makes sense to me, and got there!

Spoiler

2023_Day3.yxzp

DataNath

Day 3 done! Join and Summarize galore today. Didn't expect the co-ordinate/grid stuff to pop up already but fun nevertheless!

Spoiler

Day 3.yxmd

SeanAdams

:-) now we're warming up - brings back memories of all the matrix /grid problems from prev years!

Got to use the new formula tool that does a product! Thanks for pointing this out @NicoleJ

Spoiler

Part 1:
Prep & Cleanup:
- Break down the raw data into rows and columns
- Tag which ones are numbers vs. which ones are symbols
- For the numbers - use a multi-row tool to get all contiguous numbers
The actual problem:
- For each symbol - work out each surrounding cell (using a cartesian join)
- If any of these surrounding cells match a location with a number in - then you have a match

Break into cells

Parse and clean numbers and symbols

Add all adjacents

Part 2:
Same as part 1 really - but just requires you to check for 2 matches rather than just 1.

Screenshot 2023-12-03 121531.png

2023_Day03-Part2.yxzp

jarrodthuener

nothing tooooo crazy here... ended up using quite a bit from part 1.

Spoiler

Day3.yxzp

mbarone

Egads; had to break this one up too (part 1 not much issue, but part 2 I had to come back to as I had free time today).

Not pretty, but fully dynamic, and made a grid out of it (see browse tool) to help me out.

Spoiler

Day03.yxmd

cgoodman

Spoiler

Made some effort to comment what I am doing here. Here are some of my tricks which may have helped
1) I used to calculate the grid positions using a multirow formula, but I think I learnt a few years ago from @Aguisande to use the tile tool as with unique values on the recordID and this would create tile_number_sequence which essentially did the same thing, but avoids the risk of a typo in the multirow.
2) Once I got the X and Y position, I concatenated these together but padded them out so coordinate 1,3 becomes '001003' this helps with joining later on. I'm sure you could concat in short form but I felt this was too risky / the padding might help later on
3) Creating the 8 positions around the point using the formula tool and then transposed into long and thin data structure (however I did a typo in this stage which didn't affect the test file so I scratched my head why my original part 1 answer was incorrect).
4) Luckily all of the values were on the horizontal, so because I filtered out the '.' a multi-row formula was able to easily group the values together to get the value.
5) In part B, filtered out the * and used the formula to get the 8 positions around this to find if there were any values from Part 1 adjacent then focused only on those where there were 2.

cgoodman

@caltang similar approach to me to get the surrounding positions, but instead of creating 8 individual formula tools, I created 8 formulas in one tool and then transposed it to get from 8 fields to 8 rows. 👍

mceleavey

I don't think I've done anything special here.

Spoiler

I parsed everything out into a single row for each character and gave it a Record ID and column ID.
I split the stream into numbers and symbols, then generated the rows and appended to essentially create a 3x3 grid. I then denoted what the "cells" were in association to each symbol, and then joined those cells to the main stream. This created the "cells" to which each symbol was linked.
At this point, I solved Part 1 by cross-tabbing and concatenating the individual numbers into a cell, creating the full number and summed the value.
Part 2 was achieved by isolating the "*", grouping to find which were associated to exactly two number groups (the groups were used when the numbers were split to rows), then it was another round of cross-tabbing to get the numbers into two columns on which I could apply the calculations. Yes, I'm aware I could have done it in fewer tools if I upgraded and had the Product function...)

Easier than part two of day one!

Day 3.yxmd

Samantha_Jayne

Spoiler

I used vectors to map a 3x3 grid around my symbol to then understand if I was intersecting a number or not.

By creating the co-ordinates of the symbol against the 9 possible intersections, this enabled me to easily match and group for part 2.

It took me a while to get my workflow into something that was readable, I had far too many tools and offshoots as my example to input didn't work as expected again. Got there in the end though 😁 I hate those spanners!

I hate it when you get the example right almost immediately but spend ages working out why the main input doesn't work! That was this day for me. However another one completed with Base A. Whoop!

AoC_2023_SamiJayne_Day3.yxmd

Qiu

@clmc9601
I think our approach is a bit similar by finding all the 8 neighbours. 😁

JeffF

Spoiler

DavidP

Any workflow where I get to use the Tile tool, or where I can change both the data contents and data type using a single Multi-Field formula, is my favorite workflow!!!

When I get to do both, well, that's just an irresponsible amount of fun!!!

Spoiler

Day3.yxmd

ccanterman

That was HARD. I can't explain what I did, but it made sense to me I also gave up on the pretty annotations by the time I got to Part 2

Spoiler