[lhumphrey][5][Ada]https://github.com/lhumphrey/AoC-2025-In-Ada/tree/main/day05
[lhumphrey][6][Ada]https://github.com/lhumphrey/AoC-2025-In-Ada/tree/main/day06
[lhumphrey][7][Ada]https://github.com/lhumphrey/AoC-2025-In-Ada/tree/main/day07
Hello everyone, thanks for your contributions!
Here are the preliminary results. Let me know if you think thereās a mistake somewhere, we will adjust before publishing the results.
- jcmoyer: 12
- wutka: 12
- Daniel-Larraz: 12
- Max: 12
- Heziode: 12
- EthanLuisMcDonough: 12
- zertovitch: 11
- Paul-Farmer: 9
- jrcarter: 8
- AdaOrbit: 7
- cunger: 7
- Rodrigo-Garcia: 7
- lhumphrey: 7
- smionean: 6
- JeremyGrosser: 2
- Fabien-Chouteau: 2
Solved with Ada: 98
Solved with SPARK: 40
Completed: 6
Ada donation: $980
SPARK donation: $800
Completed donation: $600
Total Donation to Ada Academy: $2380
6 Likes
Sorry for the slow reply! For some broader context, Iām a research engineer with ~16 years experience. I donāt do āproduction level code,ā but I do regularly work on relatively complex codebases across a few different languages (C++, Python, SPARK/Ada, and a little Java). I have some comments on SPARK/Ada vs. Python that are influenced by both Advent of Code (AoC) and my own work experience.
I enjoy using Python for small, self-contained coding projects. Python nowadays is super easy to install and get packages for. Itās also easy to start coding quickly, since you can just create a .py file and immediately run it without having to create any project files. I also find that being able to run Python commands interactively often makes it easier to build understanding about algorithms Iām developing. Itās also very flexible due to its dynamic typing, and lots of common operations can be done very concisely, like for string processing or processing data in standard containers. For AoC, I found that I could code up solutions more quickly in Python than SPARK/Ada. Similarly, I use Python in my work to write simple data processing scripts, and I often make use of Python plotting libraries for data visualization. I also use it for scripting other relatively simple automated tasks. Itās great for that type of thing since it can run cross-platform without needing to be compiled, so I frequently see people use Python scripts in place of e.g. shell scripts that are shared across a team.
Iāve seen some of the major downsides of Python in my work, though. We have some relatively complex Python code for running simulations and processing data, and itās incredibly difficult to modify because itās much less readable, mainly because of the dynamic typing. Since I work on research-grade software, the team does not always use things like type hints, and we donāt have much documentation. So I end up spending way too much time trying to figure out what data various functions are actually passing around in order to make even the most simple changes. Itās also really hard to debug when something goes wrong.
So for working on large coding projects, I definitely prefer SPARK/Ada. Having a strong type system makes the code so much easier to understand, navigate, modify, and extend. Even on small coding projects like AoC, I think SPARK/Ada ends up being more readable, at least for code I write, because the design of the language makes me naturally organize concepts better. It was a couple of weeks between the time I wrote my code in Python and started translating it to SPARK/Ada, and even in that short amount of time, I had a hard time remembering exactly what some of my Python code was doing. When I write in SPARK/Ada, I generally take more time to organize my code and define proper data types and functions, so itās easier to understand if I come back to it after a while. The main downsides for me are that it takes more up-front coding to feel like I can get started on core algorithms (defining data types, instantiating packages, etc.), and I canāt ātinkerā with algorithms Iām developing as easily since thereās no interactive interpreter. For AoC, I also found it a bit annoying to set up projects. Alire is generally helpful, but I have some complaints. I can see why it works the way it does, but it feels very heavy weight for simple coding projects. For example, I didnāt have time to investigate how to change default compiler flags, and I really didnāt want many of the defaults, so I ended up modifying them manually until I gave up and just lived with a lot of warnings. I also had to edit some of them manually to add the flag for Ada 2022 features, and I had to pull in the sparklib project file manually in order to use SPARK containers. Iām also not interested in publishing crates for AoC, so it was a bit annoying to enter a bunch of information for each project (even though I left most of it blank). Maybe there are some commands that I donāt know about, but I just wanted to get coding without learning a lot about Alire, so I didnāt investigate it too much.
Another random thought: a friend did AoC a couple of years ago using Ada for the first time. He tried using ChatGPT to quickly learn to do simple things (e.g. āhow do I make a map in Adaā, āhow do I iterate over a vectorā, āhow do I create a subtypeā), since itās very good at answering those types of questions in C++ and Python. However, for Ada, it frequently produced code that looked okay at first glance but actually had syntax errors and wouldnāt compile. It also somewhat frequently explained concepts incorrectly using terminology from other languages. So he frequently had to come to me to get answers to questions. However, it looks like ChatGPT has improved on Ada quite a bit. Itās been a while since Iāve coded in Ada, so I asked it similar questions to get up to speed, and it didnāt make any mistakes this year. It was also able to answer basic questions about Alire commands, though it did mess up syntax sometimes (e.g. adding āāā to commands that didnāt need it).
So, those are some of my thoughts on Python vs. SPARK/Ada. Hopefully it was helpful!
3 Likes
Thank you for the detailed answer! Indeed some people have now started using AI for Ada and they have started getting pretty good results; things seem to be improving quite fast on that front.
Also, I know the pain of (large) Python codebases created by scientistsā¦ 
I find interesting the mix of languages that you work on sounds aerospacy 
Best regards,
Fer
Claude Code got very good with Ada!
I forgot to post the results here: Advent of Ada/SPARK 2025 Results | AdaCore
Congrats everyone!
5 Likes