Day 01 - Advent of Code 2023

[@chrisbiscardi]

I've also added an example that uses aho-corasick: one of the underlying matching libraries that powers the Rust regex crate. It is a bit slower (20 milliseconds vs 20 microseconds github.com/ChristopherBiscardi/advent-of-code/blob/dca401065808666c7553b4ba157eb0cbb31a6abf/2023/rust/benchmarks.txt ) but applies very well to the problem. github.com/ChristopherBiscardi/advent-of-code/blob/dca401065808666c7553b4ba157eb0cbb31a6abf/2023/rust/day-01/src/part2_aho_corasick.rs

[@QuentinMcCormick-o4q]

I am very new to Rust coming from a little bit of C, Javascript & C# mostly, so i am using advent of code to sharpen up my basics and understanding of syntax etc.. however i probably spent close to 3.5 hours trying to get Part 1 and 2 done in a way that i thought looked right and understood. But after seeing this video, I realize I still used so many levels of useless abstraction i did not need. And it has shown me allot of concepts in rust I either didn't know about or didn't understand how they were implemented. Subbed, great video 👍

[@mariothrowsfireballsuntitl1841]

same. I think this way of solving the problem then watching the video is really helpful

[@nomadvagabond1263]

Same😂, i solved those problems as if i was writing C.

[@TheUkearchy]

The amount of detail and explanation in this is extremely educational, thanks so much!

[@MaikDiepenbroek]

Great stuff, this year i'm gonna dive into rust myself. So your setup video helped me out a lot! GL with this years AoC!

[@kevincodes674]

I took a different approach on part 2. Created an array of tuples like ("one", "o1e"), ("two","t2o"),("three","t3e"), etc... then looped with a replace. Luckily the numbers will only overlap by one letter max, so I just kept first and last letters. Then, just pulled out digits in the same order 😊

[@tapu_]

Despite trying not to participate, I did and instantly did a war crime with .replace("one","one1one").replace("two","two2two").replace("three","three3three").replace("four","four4four").replace("five","five5five").replace("six","six6six").replace("seven","seven7seven").replace("eight","eight8eight").replace("nine","nine9nine").split(" ")[:]

[@lielfr]

almost as I did 😊 I was just lazier and did "one" to "one1one" etc.

[@orizach01]

I had a list of all the numbers as strings, then looped over each line and each number and checked if line from i to i + len(num) == num, then I add the number to a new line and also all the normal numbers, then I do the sum

[@chrisbiscardi]

It seems a lot of people did end up with this approach. Its quite nice for re-using part1's code

[@srgibberd]

To get the last digit, I reversed the line, then used pattern matching on "eno", "owt", "1", etc.

[@endrevestb895]

This is going to be a great year for AOC :D Thx! Looking foreward to tomorrow!

[@ExpertOfNil]

I'm glad I wasn't the only one that got hung up by the overlapping number words!

[@ThePixelSlime]

These are simply amazing! I'm learning Rust and the deep dives are super helpful. :D

[@oghenemarho5694]

Thanks you for this Tutorial Chris, it really helped me with the part 2 of the challenge.

[@pwbandwidth]

This is such a great concept for a series. I've been solving the problems mostly in JS or Python and then just translating them to Rust, but this series really helps in learning how to think in Rust.

[@diegoparraca]

That's definitely awesome content for us beginners :D, please keep doing it!

[@ericaustinbarber4186]

Thank you Chris!

[@siddharthbisht8522]

Wrote my first rust code from scratch today. Took me 1 hour to get both stars but was worth it. Thanks to your previous video, I learned about Advent of Code, which also led me to discover Advent of Cyber.

[@siddharthbisht8522]

@@Gers217 ✌️

[@florianbopp187]

very interesting, I took a very different approach, which might not be performant, but I am just doing this to learn rust. Basically I had a couple of helper functions: get_first_and_last_digit(line: &str) -> (usize, char) // returns tuple of index and found digit as char get_first_and_last_sting_digit(line: &str) -> (usize, char) both of them use the string.find or rfind methods to get the index of first or last occurence, for the string based digits I had another helper funciton which matched from "one" => '1' etc. so both function would have the same output. then in the main() I compared both functions output and chose the lower index version for the first digit and higher index verison for the second input. concatenated with the same forma().parse() and tada it worked

[@chrisbiscardi]

nice! You might be interested in aho-corasick, which does multiple searches at the same time and seems like an evolution from your current approach -- github.com/ChristopherBiscardi/advent-of-code/blob/dca401065808666c7553b4ba157eb0cbb31a6abf/2023/rust/day-01/src/part2_aho_corasick.rs

[@yumbuboyumbubo8230]

Hello fellow rfind and find user! My approach is similar, but I used a single array and instead of mapping I used index % 9 + 1

[@kentbowling7513]

The "eightwothree" actually not mattering in the test input was so evil (either way would return 83)🤣 Your solution is much more elegant than mine. In part 2 I abused a regex, and in the process learned the std Rust crate doesn't have "lookahead assertions" to do it cleanly. There's another crate called "fancy_regex" that does, but I instead just matched multiple times and advanced where I was looking 1 index past the start of the last hit. If I had to guess, my approach would only work for ASCII and I'd be in trouble if someone for instance put a poop emoji in the input 💩. I'm not sure if yours is affected, but I'll probably go check it out for fun. Your more "functional programming" approach is so much cleaner in the end, and I need to train myself to take those sorts of opportunities.

[@chrisbiscardi]

as far as elegance goes, the "aho corasick" implementation in the repo is far better than even what I show in this video, although the benchmarks show it slower in this case (ms vs microseconds).

[@markday3145]

This was the first time I made an effort to make my string processing Unicode-friendly. I ended up using str::char_indices to loop over valid character offsets. That also returns the character at that offset. So I checked if that character is_ascii_digit(), and called to_digit(10); otherwise, I looped over the words "one", ..., "nine" and did line[i..].starts_with(word), and using the word's index plus 1 as the digit.

[@flyingsquirrel3271]

This is fun, thanks for making these videos! A much more efficient solution for "concatenating" two decimal digits would be to just multiply the first one by 10 and then add the second one. Allocating a string, parsing it and then deallocating it is kinda excessive, I'd say ;) On the other hand, your solution is probably more helpful for rust newbs that are watching since they learn how to do the string stuff.

[@chrisbiscardi]

Yeah, I refactored to do the math in the repo afterwards

[@markday3145]

I keep forgetting about filter_map(). At least this time, I did .filter(...).next().unwrap().to_digit(10) instead of .map(|c| c.to_digit(10)), so clippy did't warn me about it. I feel marginally better about forgetting filter_map when I never actually used map(). I liked the way you did match it.last() to handle the case where the line only had one digit. In part 1, I ended up iterating over each line twice (once to get the first digit, and once for the last digit). In part 2, I was using a for loop instead of an iterator chain, so I used Option variables to keep track of whether I had seen a first or last digit yet.

[@servalp]

very nice to see some else's approach. I totally forgot about starts_with. You can "clean up" that if-else tree with a match, like so: match &word { w if w.starts_with("one") => Some(1), w if w.starts_with("two") => Some(2), w if w.starts_with("three") => Some(3), ... _ => None, }

[@necuz]

Checking this out after having completed day 1 (in Go, sorry but I have Rust burnout). I totally also tried cheesing part 2 with replace all ones etc. and then left-to-right search and replace. Liked the trickiness of this, but they should have had a test case that ended in like eightwo or something.

[@chrisbiscardi]

you don't need to apologize for using a different language. I definitely wish the test output for day 01 was comprehensive, but that's definitely something that is a factor in advent of code every year.

[@Agryphos]

I missed that for lines with only one number you just take the found number twice, messed me up for a bit with part 2

[@jamstoast]

for part 1 I used rusts' char.is_ascii_digit() and then for part 2, used an index to iterate, still using char.is_ascii_digit() for the normal digits and looked ahead for the spelled versions

[@LloydHZA]

This is great - just looking into Rust now and this has been super helpful. How did you do the selection in VSCode and paste at 12:41?

[@chrisbiscardi]

If you option+click you'll get another cursor. If the number of items you copied matches the number you're pasting into it will put one in each spot.

[@IvanoColtellacci]

In the video at a moment you togle the inlay hints of the rust-analyser, can you tell me how do you achive thta ?

[@chrisbiscardi]

kzbin.infowBTAMLp-gnc

[@chrsolr]

@chrisbiscardi do you have a video to make vscode do clean and simple? Theme, fonts, etc?

[@chrisbiscardi]

I have my setup described here: kzbin.info/www/bejne/l6GoiaCAhZqrrtk alongwith a link to some settings I was using at that time in the description. Its basically the same today.

[@Ankstek]

I got stuck on part 2, had the example been "threeightwo" then I'd understand that I needed to account for overlap.

[@chrisbiscardi]

yeah, really unfortunate that the test input didn't fully exercise what happens in the actual input

[@teuluPaul]

Thanks for the video Chris! I am very new to rust, and had a go a solving the problems before watching your video. I did succeed with part 1, but not part 2. I had an interesting issue with testing. If I put the test input into a file, and reference it in the same way as the code, I get a file not found error which I could only resolve by putting the full path in rather than a relative path. Which working directory does cargo use when running tests?

[@chrisbiscardi]

include_str! Is relative to the file that you use it in. doc.rust-lang.org/std/macro.include_str.html

[@teuluPaul]

@@chrisbiscardi Thanks for the reply!

[@kusadasi8641]

Im new at programming, does two pointer work here? Just iterate the front and back until you get a number then combine them and then sum them. I have no idea how rust works... (Only part 1)

[@chrisbiscardi]

yep, in Rust there's find and rfind for searching from the front/back. Here's an example of that: github.com/ChristopherBiscardi/advent-of-code/blob/c8ccc5cb0f90d31565070be2187f9b499cfedca4/2023/rust/day-01/src/part1.rs#L4-L34

[@mykalimba]

11:00 For Part 2 on this challenge, I found that their examples of how words should be converted to actual numbers was not sufficient in explaining how the conversion needed to be performed. For example, if the input line is "abconeightdef", you need to convert that to "abc18def" for your checksum to match what they're expecting.

[@mykalimba]

@@fnmatrix And I assert that is is not "perfectly clear", in that there is ambiguity in what "last valid number" means. There is no concrete example of how an input that ends with something like "oneight" would resolve. Should it become "1ight", "18", or "on8"? _That_ is ambiguity.

[@cryoexn7307]

Also I was wondering if you are manually updating your bench results with the table ligatures? I am adding benchmarking to my AOC setup and was wondering if there was anything fancy going on there :)

[@chrisbiscardi]

Nothing fancy, just running it and copy/pasting the terminal output. I'm not even doing the table layout, that's the bench tool output. I thought about automating it but didn't really feel like it was worth it.

[@SkylerLinux]

Part 2: that massive "ugly" if-else tree is crying out to be a switch, maybe an enum too

[@chrisbiscardi]

yeah it really is isn't it. idk why I didn't go for match guards to start with -- github.com/ChristopherBiscardi/advent-of-code/blob/3b00d6e270a90456289763e4697ceb608f7447a1/2023/rust/day-01/src/part2.rs#L17-L30

[@ovidiu_nl]

first * 10 + last 😅

[@chrisbiscardi]

github.com/ChristopherBiscardi/advent-of-code/commit/3def2b0b86239cadd1973de9e9ba7906a23b6e50

[@husziahazban]

yeah ,maybe this is the only reason my solution was a bit faster: fn process(input: String) -> u32 { let c_iter = input.chars(); let mut digits: Vec = Vec::with_capacity(2); let mut sum: u32 = 0; for c in c_iter { if c == 0xA as char { sum += digits[0] * 10 + digits[1]; digits.clear(); } if c.is_ascii_digit() && digits.is_empty() { digits.push(c.to_digit(10).unwrap()); digits.push(c.to_digit(10).unwrap()); } if c.is_ascii_digit() { digits[1] = c.to_digit(10).unwrap(); } } sum += digits[0] * 10 + digits[1]; sum }

[@misterwoody_]

English is unfortunately an imprecise language. I think from the spec it mentioned the "first" and "last" digit in a line of input (I could be wrong) so I thought of it as first "digit" could be found with a left to right search (stop looking at this point) last digit could be found with a right to left search. I think that does give the correct results without worrying about the overlapping and also picks up the single "digit" situation. I'm not sure about efficiency but hopefully that's a way of thinking about the problem in a clearer way? Many thanks

[@chrisbiscardi]

after some refactoring, that's exactly what I ended up with -- github.com/ChristopherBiscardi/advent-of-code/blob/dafde982166a4edca9a24f978c6a202fc7175dd2/2023/rust/day-01/src/part1.rs#L18-L27

[@Metruzanca]

On the topic of turbo-fish syntax "e.g. sum::()". Why do we need the turbofish here and can't just do sum()?

[@chrisbiscardi]

Basically because if you remove the :: then it's ambiguous as to whether you mean "type argument" or "comparison operator" (greater than/less than)

[@CrazyLuke11]

Hey! Love your content!❤ Could you maybe to a tutorial on justbuild? Thank you 😊

[@chrisbiscardi]

Yep, is there anything you want to know specifically?

[@xyzg34]

Is that a window manager for macos?

[@chrisbiscardi]

Yeah, it's yabai

[@zzzzzzzzzzzzzzzzzzzzzzzzzzzz__]

if I try and run just work day-01 part1 it tells me the project root does not exist :/ not too used to this tooling I guess. all of it does seem really helpful though. but I guess perhaps I will use a basic cargo new for my aoc this year. thanks for the good videos anyway

[@chrisbiscardi]

Did you create the day 01 directory using ` just create day-01`?

[@zzzzzzzzzzzzzzzzzzzzzzzzzzzz__]

@@chrisbiscardi i did, I tried it again just now, just create says the project was created, but running just work inside the day-01 folder gives the project root does not exist. Recipe work failed on line 3 with exit code 1

[@lpanebr]

Regular expressions

[@vyorkin]

damn, I tried to use nom for part2 parsing at first

[@chrisbiscardi]

you can definitely use nom for part 2 -- github.com/ChristopherBiscardi/advent-of-code/blob/14c04a3a056e648737d614da0e85d000e65af9ec/2023/rust/day-01/src/part2_nom.rs

[@luckystrike91]

looks strange for me in part1, much simpler and efficien approach: let first = it.next().expect("line must contain at least one digit"); let last = it.next_back().unwrap_or(first); 10 * first + last

[@chrisbiscardi]

the code in the repo does get refactored after the initial day is completed (github.com/ChristopherBiscardi/advent-of-code/blob/c33be5e886003701134165bb938c5e5dad87c3f3/2023/rust/day-01/src/part1.rs). Day 1 also includes some interesting alternatives at this point that show off crates like aho-corasick, which powers the regex crate -- github.com/ChristopherBiscardi/advent-of-code/blob/c33be5e886003701134165bb938c5e5dad87c3f3/2023/rust/day-01/src/part2_aho_corasick.rs#L22

[@steamedeggeggegg]

doesnt expect it to be so hard in rust got it done in javascript, but i think my code is all reproducable in rust

[@Gers217]

First comment 😊

[@Enan_Andrew]

It will more readable and concise way for combining and conversion. let mut it = line.chars().filter_map(|character| character.to_digit(10)); let first = it.next().expect("must be a number"); let last = it.last().map(|num| num.to_string()).unwrap_or_else(|| first.clone().to_string()); format!("{}{}", first, last).parse::().expect("a number")

[@flwi]

Nice solution! I went with a regex like this. Let's see if youtube likes it ;-) .*?(one|two|\d).*(one|two|\d).*?