Lots of places; Dictionaries.jl’s packages are just much faster than Base Dicts to iterate over. There’s a reason most package designers use them rather than Dicts.

Yes. I think Andy writes about this in several places. Like in Getting Started in the README. The three main difference to Dict are that it preserves the order of elements, it iterates much faster, and it iterates values rather than key-value pairs. His JuliaCon talk is very clear. If you do a…

AFAIK there is no great way to se a Base.Dict to implement a “lazy accumulator”, like this: if haskey(dict, key) dict[key] += 1 else dict[key] = 1 end This needs three lookups. We can reduce this to two via get!: dict[key] = 1 + get!(dict, key, 0) But I am not aware of way to …

Add modify! function for lookup/update/insert/delete in one go by tkf · Pull Request #33758 · JuliaLang/julia · GitHub enables this, and to my understanding is a reason why a token-based API doesn’t currently exist in Base. Maybe the current thread will reignite efforts on landing one of the two app…

[image] fingolfin: But I am not aware of way to get it down to 1 Neither was I… but now I think the following incrementkey! does it: using OrderedCollections incrementkey!(d, k) = ( mergewith!(+, d, LittleDict((k,), (1,))) ;nothing) classic_incrementkey!(d, k) = ( d[k] = 1 + get!(d, k, 0) ;…

The last time I looked into it (a long time ago), you have to go behind the API for Dict. Whereas (as discussed above) avoiding looking up twice is part of the API for Dictionary. I wrote some things towards giving Dict and Dictionary a common interface here DictTools.jl . In one place, it avoids com…

Fwiw, adding a token API to Dict would be a reasonable thing. Someone just need to make a proposal/implementation. Could be based on Dictionaries.

Agreed - there’s a lot of AbstractDict API work that could happen in Base without breaking changes.

I tried to use Dictionares instead of the basic Dict, but keeping the same structure, in this elegant script by @Daniel_Gonzalez that solves the problem of the AOC 2024 d11, but I got no improvements. I don’t know if the failure is due to a clumsy application of the Dictionares.jl package or if the…

Since the posters above seem to suggest that Dictionaries.jl is faster for iteration but possibly slower for random access, it’s normal to not expect a speedup for your code which involves random access of the d_n Dict. Having said that, the following part your code could potentially be made faster…

Another version of solve saving hash calculation using mergewith!: using OrderedCollections function solve(stones, n) d_new = Dict{Int,Int}() d_old = Dict(stones .=> 1) for _ in 1:n for (s, n) ∈ d_old t1, t2 = blink(s) mergewith!(+, d_new, …

If I did not measure badly, the performance does not improve. On the contrary … julia> @btime solve(stones, 25) # 260 us 3.339 ms (13913 allocations: 1.56 MiB) 197357 julia> @btime solve(stones, 75) # 12ms 155.595 ms (516906 allocations: 49.95 MiB) 234568186890978

Performance of Dictionaries.jl vs Base.Dict

Internals & Design

Topic		Replies	Views
[ANN] Dictionaries.jl - Improved productivity and performance of dictionaries in Julia Package Announcements dictionary , dictionaries	22	6487	December 15, 2019
Poor time performance on Dict? Performance	26	19657	March 12, 2018
Replacement for Dict, is anyone using isslotfilled? General Usage	22	2160	October 27, 2020
Dictionaries.jl token API General Usage	5	1124	December 31, 2022
Fast Dict value modification by accessing Dict.vals Performance	16	3896	March 14, 2019

Performance of Dictionaries.jl vs Base.Dict

Related topics