Feasibility of supporting L-shaped available space (Pre-loaded containers / Obstacles)

[jfrascon]

Hello,

First of all, thank you for this library. I am currently evaluating xflp for a warehouse project involving truck loading.

I have a critical requirement regarding the initial state of the container. In many operational scenarios, the truck is not empty when the loading process begins; it already contains cargo from a previous stop.

This effectively means the available free space is not a perfect rectangular cuboid, but often an irregular shape (e.g., an L-shape).

The Scenario:
Imagine a standard trailer (13.6m×2.4m×2.5m).

• There is pre-existing cargo occupying a block at the front-left corner (e.g., 4m×1.2m×2.5m).
• This cargo is immovable (fixed).
• The solver must pack new items into the remaining space (the resulting "L" shape) without overlapping the pre-existing cargo.

My Question:
Given the current architecture of xflp, how complex would it be to support this scenario?

Is there an existing mechanism to:

• Add "Fixed Items": Inject items at specific coordinates (x,y,z) before the packing starts, treating them as immutable obstacles that consume free space?
• Define Custom Free Spaces: Instead of initializing the container as one single large free rectangle, initialize it with a collection of free rectangles that form the L-shape?

I would appreciate your insight on whether this is achievable with the current codebase or if it would require a major rewrite of the core packing heuristic.

Thanks!

[hschneid]

Surely this is possible to think about having preconditions, like already loaded boxes. But as always, preconditions are risky thing: The preconditions must hold the standards of internal conditions like fobidden overlapped boxes.

lets say every added item could have a preset for x,y,z. So, every run of algorithm must first place these fixed boxes. Surely these fixed boxes potentially overlap with themself or are bigger than the truck. Or bearing is too high. This needs to be checked in advance and the algorithm must give a proper feedback to user, when failing.

So, yes, surely it is possible. But it needs a bit more than just fixing. ;-)

I put in on my list and inform you, when added.