Troubles defining Float32 or empty phantoms with KomaMRI 0.9

[Tooine]

What happened?

Hi,

I switched to KomaMRI 0.9 and observed some performance improvements with simulations, which is great 👍

Yet, I had some troubles (see below) with the way I define Float32 or empty phantoms with the new version.
I am not sure if this is a bug or just the new expected behavior since 0.9, as for both situations, I found a (imo, less esthetic) way to define it as I want.

Note that, if this is not a bug, there might be parts of the documentation to modify, like in the $T_2^*$ section of this page, where the following empty phantom definition does not work anymore (at least for me)

# (3.1.1) Create an empty phantom
obj_t2star = Phantom{Float64}(x=[])

Issue 1: Float32 vs Float64 phantom
Creating a phantom encoded in Float32 like this leads to an error

Phantom{Float32}(x=[1.0])
→ ERROR: MethodError: Cannot `convert` an object of type NoMotion{Float64} to an object of type KomaMRIBase.AbstractMotion{Float32}

but this works fine and produces a phantom encoded in Float32 (including the motion!)

Phantom(x=Float32[1.0])
→ Phantom{Float32}
  name: String "spins"
  x: Array{Float32}((1,)) Float32[1.0]
  y: Array{Float32}((1,)) Float32[0.0]
  z: Array{Float32}((1,)) Float32[0.0]
  ρ: Array{Float32}((1,)) Float32[1.0]
  T1: Array{Float32}((1,)) Float32[1.0f6]
  T2: Array{Float32}((1,)) Float32[1.0f6]
  T2s: Array{Float32}((1,)) Float32[1.0f6]
  Δw: Array{Float32}((1,)) Float32[0.0]
  Dλ1: Array{Float32}((1,)) Float32[0.0]
  Dλ2: Array{Float32}((1,)) Float32[0.0]
  Dθ: Array{Float32}((1,)) Float32[0.0]
  motion: NoMotion{Float32} NoMotion{Float32}()

Issue 2: Empty phantom
Those commands do not work anymore

Phantom(x=[])
Phantom{Float32}(x=[])
Phantom{Float64}(x=[])
→ ERROR: MethodError: no method matching zero(::Type{Any})

But this works fine:

Phantom(x=Float32[])

Environment

OS x86_64-linux-gnu
Julia 1.10.2
KomaMRIPlots 0.9.0
KomaMRIFiles 0.9.0
KomaMRI 0.9.0
KomaMRICore 0.9.0
KomaMRIBase 0.9.0

[pvillacorta]

Thank you for posting this! This is indeed a bug, and is related to the new motion field in this new version.
See #499
Cheers