The array modifier is a very powerful tool, especially when stacked multiple times and/or paired with other modifiers such as Curve. The start and end cap feature of the array modifier is crucial to it's ability to maintain non-destructive modeling, but the current implementation makes it a hassle.

Issue:

The array modifier currently references object-type data blocks in order to create start and end caps. However, the process of separating the cap mesh into a new object and placing it so that the modifier connects it properly with the original mesh is challenging and unintuitive.

Suppose I am trying to achieve this model using a simple repeated base-mesh.

In order to do this currently, the body object and its start/end caps must be positioned like this.

Setting this up correctly is troublesome and unintuitive, and makes it difficult to edit the junction between body and cap later on.

Proposal:

I propose that the array modifier be changed to reference vertex groups for its start and end caps. This would mean that no objects need to be separated/moved, and the junction between body and cap can be edited freely and as expected. Vertex groups would look like this:

And the end result would look like this: (Excuse the crappy photoshop)

Ideally the modifier would remove the parts of the mesh included in each vertex group while duplicating, then reattach them as caps afterward. I am not familiar enough with Blender's code to know if this is feasible, but I think it would really help. Please let me know if this is something that I could try to tackle as someone with just a little experience with Python and no experience coding Blender.