Closest Point Blueprint

2026-01-21 17:08 UTC gpt-5-2-thinking Open in ChatGPT ↗

You

I have 2 arrays of vectors (positions) in Unreal and I need to find the closest point of the 2nd array from each point of the 1st array. I need to use blueprints. I need the least ugly way

ChatGPT

In Blueprints, the cleanest (and fastest-enough) pattern is: for each point in ArrayA, scan ArrayB once, keep the best (smallest) distance, output the winner — but wrap it so your graph stays readable.

Least-ugly Blueprint approach (pure BP, no plugins)

1) Use Distance Squared (avoid Vector Length / Distance)

Comparing squared distances avoids a sqrt every iteration.

• Use Vector - Vector → Vector Length Squared
• Or VSizeSquared equivalent nodes (depends on version)

2) Put the logic in a Blueprint Function (or Macro)

Create a function like:

FindClosestPoint

• Inputs
  • Target (Vector)
  • Candidates (Array of Vector)
• Outputs
  • Closest (Vector)
  • ClosestIndex (Int)
  • bFound (Bool)

Function graph logic

• If Candidates.Length == 0 → bFound=false, return
• Set BestDistSq = very large (e.g. 1e30)
• Set BestIndex = -1
• ForEachLoop over Candidates
  • DistSq = (Candidates[i] - Target) -> LengthSquared
  • If DistSq < BestDistSq
    • BestDistSq = DistSq
    • BestIndex = i
• After loop:
  • bFound = (BestIndex != -1)
  • Closest = Candidates[BestIndex]

This keeps your “ugly” loop hidden inside one neat function node.

3) Main usage: loop ArrayA, call the function

• ForEachLoop over ArrayA
  • Call FindClosestPoint(ArrayA[i], ArrayB)
  • Store results into:
    • ClosestPoints (Array of Vector) (same size as A), and/or
    • ClosestIndices (Array of Int)

---

Making it even cleaner

Option A: Return a struct

Make a struct ClosestResult:

• ClosestPoint (Vector)
• Index (Int)
• DistSq (Float)

Then the function returns one struct, so wiring is minimal.

Option B: Also support “max search radius”

If you don’t want matches past a threshold:

• Input MaxDist (Float)
• Compare to MaxDistSq
• Only accept if BestDistSq <= MaxDistSq

---

Performance notes (so you don’t build something you regret)

• This is O(N*M). Fine for a few hundred or a few thousand points occasionally.
• If you’re doing this every tick with large arrays, you’ll want spatial partitioning (grid / kd-tree) — which is awkward in BP.
• A solid “still BP-friendly” optimization: early-out with radius or chunking (run over several frames).

---

If you want the absolute cleanest graph

Tell me:

• Are these points Actors or just Vectors?
• Rough sizes of A and B (like 50? 5000?)
• How often you run it (once, on demand, every tick?)

…and I’ll suggest the nicest Blueprint layout (Function vs Macro vs BP Function Library) for your case.

You

how to pick one or the other value based on a boolean ?

ChatGPT

In Blueprints you’ve basically got 3 clean options.

1) Select node (the nicest)

Use Select and drive it with your boolean.

• Right-click → search Select
• Pick the typed one you need (e.g. Select Float, Select Vector, Select Int, etc.)
• Plug:
  • Index / Pick A = your Bool
  • A = value when false
  • B = value when true

This is the Blueprint equivalent of b ? B : A.

2) Branch (when you need execution flow)

If choosing the value also changes what you do next (not just which value), use a Branch:

• Condition = Bool
• True path sets/uses value A
• False path sets/uses value B

3) Inline with Make Array → Get (not recommended)

You can do bool → int and index an array of 2 values, but it’s less readable than Select.

---

Tip: If you’re picking between two vectors/floats purely as a value, Select is almost always the least ugly.

If you tell me what type you’re selecting (Vector? Float? Actor?), I can name the exact Select node to use.