`-array_expr` fails with "Matrices are the only arrays that can be raised to a power"

[1-Bart-1]

When negating a symbolic array expression that is a sum of scalar * array terms, the unary negation operator produces an invalid result that cannot be scalarized. The expression incorrectly collapses to -1 - 1 instead of remaining a 3-element array.

Interestingly, -1 * array_expr works correctly - only the unary - operator is broken.

MWE:

using Symbolics
using LinearAlgebra: ×

@variables t
@variables group_chord(t)[1:3, 1:4]
@variables group_y_airf(t)[1:3, 1:4]
@variables twist_angle(t)[1:4]

# Build symbolic vectors
gc = collect(group_chord[:, 1])
gc_norm = sqrt(sum(abs2.(gc)))
x_airf = gc / gc_norm

gy = collect(group_y_airf[:, 1])
init_z_airf = x_airf × gy

# Sum of scalar * array terms - this is a valid 3-element symbolic array
z_airf = sin(twist_angle[1]) * x_airf + cos(twist_angle[1]) * init_z_airf

# Scalarizing z_airf works fine
println("scalarize(z_airf) works:")
println(Symbolics.scalarize(z_airf))

# Workaround: -1*z_airf works
println("\nscalarize(-1*z_airf) works:")
println(Symbolics.scalarize(-1*z_airf))

# BUG: Scalarizing -z_airf (unary negation) fails
# First, look at what -z_airf looks like (use show to avoid scalarization in print)
print("\n-z_airf = "); show(stdout, -z_airf); println()
# Note: It shows as "-1 - 1" which is bizarre - a 3-element array became scalar -2!

println("\nscalarize(-z_airf) fails:")
Symbolics.scalarize(-z_airf)

Error:

ArgumentError: Matrices are the only arrays that can be raised to a power. Found array of shape UnitRange{Int64}[1:3].

Observed behavior:

• show(stdout, -z_airf) displays -1 - 1 instead of a 3-element array
• scalarize(-z_airf) throws the power error

Expected behavior:

• -z_airf should be a 3-element symbolic array (same as -1 * z_airf)
• scalarize(-z_airf) should work and return element-wise negation

Workaround:

Use -1 * z_airf instead of -z_airf.

Version info:

• Julia: v1.12.4
• Symbolics.jl: [0c5d862f] Symbolics v7.8.0

[AayushSabharwal]

Thanks for the bug report. Something seems to be seriously wrong here. I believe the -1 -1 bug is fixed in SymbolicUtils@4.15.1. I'll look into the printing issue.