Add optional distance range to function

Question

I have a simple brute-force search function:

function run_query(search_data::Vector{Float64}, query::Vector{Float64}, search_f::Function)::Tuple{Float64, Int}
    current_best = Inf
    loc = -1
    q_len = length(query)

    for d_i in 1:(length(data) - q_len)

        dist = search_f(data[d_i:d_i+q_len-1], query)
        if dist < current_best
            current_best = dist
            loc = d_i
        end
    end

    return current_best, loc
end

I have an alternative version of the function which also returns the maximum distance found while searching to data:

function run_query_max(search_data::Vector{Float64}, query::Vector{Float64}, search_f::Function)::Tuple{Float64, Int, Float64}
    current_best = Inf
    loc = -1
    q_len = length(query)
    max_dist = 0.

    for d_i in 1:(length(data) - q_len)

        dist = search_f(data[d_i:d_i+q_len-1], query)

        if dist > max_dist
            max_dist = dist
        end

        if dist < current_best
            current_best = dist
            loc = d_i
        end
    end

    return current_best, loc, max_dist
end

This can be tested as follows:

function simple_dist(data::Vector{Float64}, query::Vector{Float64})::Float64
    return sum((data .- query) .^ 2)
end

data = [0., 0., 1., 2.,
        3., 1.1, 3., 3.,
        1., 0., 0., 1.,
        0., 0., 1., 4.5,
        2., 1., 0., 0.]

query = [1., 2., 3., 1.]
val, idx = run_query(data, query, simple_dist)
@assert (idx == 3)
@assert isapprox(val, 0.01, atol=0.001)

val, idx, max_dist = run_query_max(data, query, simple_dist)
@assert (idx == 3)
@assert isapprox(val, 0.01, atol=0.001)
@assert isapprox(max_dist, 21.25, atol=0.001)

This pattern of:

1. Basic search function
2. Add keeping track of maximum distance in search function

Occurs in multiple places in my code. What is a reasonable way to refactor it out?

Answer 1

I think this depends a lot on your motivation for separating these into two methods.

One simple option I would consider first is to just have 1 version of each of these functions to avoid the duplication in the first place. The caller can just ignore the max_dist value if they don't need it.

If performance is an issue and max_distcalculation could be a bottleneck, you could add a Boolean argument to the single function to indicate if max_dist should be calculated or not.

If the logic for max_dist is complex enough to hurt the readability of the basic search, I think you may be better off going with the duplication instead of doing something more complex. It could be done, but not in a way I would consider reasonable in most cases.