Allow using const values in agent macro

[Tortar]

This should make it possible to use default values addressing #709 , let me know what do you think. Experimenting a bit, there shouldn't be any perf regression applying @kwdef as default

(I also put in the api page of the documentation the replicate! function, forgot to do that in the other pr :D)

edit: now it also includes the possibility to declare constants, which should help performance, but need some good tests

[codecov-commenter]

Codecov Report

Merging #831 (81686fc) into main (de57ea3) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #831      +/-   ##
==========================================
+ Coverage   70.13%   70.17%   +0.04%     
==========================================
  Files          42       42              
  Lines        2719     2723       +4     
==========================================
+ Hits         1907     1911       +4     
  Misses        812      812              

Impacted Files	Coverage Δ
src/Agents.jl	100.00% <100.00%> (ø)
src/core/agents.jl	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Tortar]

Implemented the first way described in #728

using Agents

@agent A NoSpaceAgent begin
    f1::Int
    f2::Int
    consts = (:f1, :f2)
end

agent = A(1, 3, 5)
agent.f1 = 10

returns

ERROR: setfield!: const field .f1 of type A cannot be changed
Stacktrace:
 [1] setproperty!(x::A, f::Symbol, v::Int64)
   @ Base .\Base.jl:38
 [2] top-level scope
   @ REPL[4]:1

[Tortar]

this should be it, all new tests pass!

[Datseris]

does it allow inhereting from a type defined with constants? if we change NoSpaceAgent to have const id::Int, does @agent work with inhereting from NoSpaceAgent?

[Tortar]

Yes it allows it, any const will be propagated to any inheriting new type, this is so because we actually check if there are any const field in the base type and then add it as a const also in the new type

[Datseris]

ok great!

[Datseris]

why isn't it possible to do things the native julia way and allow prefacing with const a constant field? Why do we have to have this extra weird consts declaration? what if a user wanted a field named consts?

[Tortar]

yes, a field name consts is disallowed, there is no way to have const inside the agent macro unfortunately due to parsing errors I think, see JuliaLang/julia#45024

[Tortar]

it's a bit hacky yes but it's a way not to redefine the agent macro as was done in #728, we could follow that path alternatevely but this means that every code will break :D

[Datseris]

but the julia issue you cite is fixed...?

[Datseris]

Okay, I see, there is no way to have const x = 1. Okay. I think your solution with the extra field is the only option then.

But, rename it to constants. and document it appropriately.

@Tortar this is a big change. please do not tag anything after mergfing the const and keyword improvements. i want to read the final docstring and changelog and give the ok for a new tag.

[Tortar]

yes, but the fix is disallowing the construct :D Il Sab 22 Lug 2023, 11:59 George Datseris ***@***.***> ha scritto:

…

but the julia issue you cite is fixed...? — Reply to this email directly, view it on GitHub <#831 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQH6VX2MLVPWHNC7I3KC4ELXROP6JANCNFSM6AAAAAA2KZPRPI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Tortar]

sure, I actually have another important change which I would like to finish up, we should probably tag them all together

[Tortar]

@Datseris I completed the higlighting of the changes reverting some of the changed lines in https://github.com/JuliaDynamics/Agents.jl/pull/732/files , I also declared the id field of NoSpaceAgent as constant

[Datseris]

Sounds great. Simple comments to address and then please merge (BUT NNOT TAG!!!)

[Tortar]

Merging! (Not tagging :D)