frozen VS tailspin-v0

I found this extension that implements "frozen" versions of some C++ containers, but I was wondering if there is a good solution available in the standard library.

A library exists that can produce constexpr hash table based containers.

I use the Frozen library for that. Since the conversions should be known at compile time you can make constexpr hash tables for lookups.

The constexpr dependency of note here is frozen.

In C++, a drop-in replacement for your DSA can provide significant improvements over the standard library. Particularly the standard unordered_map class can be improved by 50% to 100% (e.g. https://github.com/greg7mdp/parallel-hashmap, or for static maps https://github.com/serge-sans-paille/frozen). Of course, recognize that creating a DS/A from scratch is an entire project, and you shouldn't roll your own for an independent codebase.

Technologies (for curious folks): C++17, SFML, Entt, Frozen, Protobuf, spdlog, GoogleTest, GoogleBenchmark, CMake and Dear ImGui for debug purpose.

In the language, I added anonymous array literals. I did some cleanup in the compiler and updated to LLVM 12 from 10 (which was pretty trivial, surprisingly). I also added frozen, a C++ perfect-hashing library, as a dependency to speed up the lookup of keywords in my lexer. The library exploits C++’s constexpr features to generate a perfect hash at compile-time without any separate build step, which is great, and it also provides a drop-in replacement for std::unordered_map that uses the hash.

This is where I plug Frozen :-] https://github.com/serge-sans-paille/frozen

I’ve no idea how hard it is to implement, but return type polymorphism would be nice. Especially returning different things based on the constexpress of the result. And then add Frozen eqivalents of associative containers to the STL, so that, for example constexpr auto set = std::make_set(...) would be frozen::set, and auto set = std::make_set(...) would be std::set.

In which case, I believe the answer your question would be yes: the frozen map.

I eventually tend to do all days in Tailspin. The ones I have done so far are in directories ending in "tt" (the others are in Pyret, just to get a feel for it) https://github.com/tobega/aoc2023/tree/main

As a general tip, it is often helpful to first try to think of how you would like to represent the data in your program. Then you need to parse the data into that structure. I'd recommend you to look at a PEG-parser, for example. Or if you like, look at my Tailspin programming language which has a very visual parser syntax and also very visual ways of creating data structures (if that should happen to be your mental affinity). Look at my day1 for example. Or if you're more mathematical, maybe a functional language (I also did day1 in Pyret)

My Tailspin language is based on processing streams of values, you might want to look at it https://github.com/tobega/tailspin-v0

Many people had trouble with the day 7 problem. Paradoxically, good developers probably had more trouble. Here some of the difficulties are explained and implementations are provided in imperative, functional and OO styles, written in the Tailspin programming language.

I test almost entirely from my language, that way the tests are independent of the implementation. Currently the tests are implemented in java because that fits the interpreter implementation https://github.com/tobega/tailspin-v0/tree/master/test/tailspin/samples

Finished off the implementation of typed and offset array indices in Tailspin

I ended up enabling left recursion in Tailspin's composer (parser) syntax. Much cleaner calculator example now.

I think it's possible that Tailspin might be suitable for you.

I think gradual typing is definitely something worth exploring more. I thought it was a shame when Dart abandoned that path. Have you seen Shen ? I guess my small offering, Tailspin, is currently evolving to gradual typing as well.