Fix missing metadata when operation and resulting BasicSymbolic subtype mismatch

[bowenszhu]

This PR addresses an issue in the basicsymbolic function where it incorrectly assumes that the + and * operations always result in Add and Mul objects, respectively.

SymbolicUtils.jl/src/types.jl

Lines 589 to 600 in 2ec473e

	if f === (+)
	res = +(args...)
	if isadd(res)
	@set! res.metadata = metadata
	end
	res
	elseif f == (*)
	res = *(args...)
	if ismul(res)
	@set! res.metadata = metadata
	end
	res

Currently, when arguments possess metadata, the operations might return a Term object instead, as seen in

SymbolicUtils.jl/src/types.jl

Lines 1082 to 1083 in 2ec473e

	function +(a::SN, b::SN)
	!issafecanon(+, a,b) && return term(+, a, b) # Don't flatten if args have metadata

SymbolicUtils.jl/src/types.jl

Lines 1136 to 1138 in 2ec473e

	function *(a::SN, b::SN)
	# Always make sure Div wraps Mul
	!issafecanon(*, a, b) && return term(*, a, b)

This inconsistency was causing problems in Vaibhavdixit02/SymbolicAnalysis.jl#43 (comment)

This PR corrects this problem by explicitly handling the Term case. This ensures the maketerm function behaves consistently and correctly handles arguments with metadata.

[github-actions]

Benchmark Results

	master	d8327c0...	master/d8327c054f981d...
overhead/acrule/a+2	0.833 ± 0.021 μs	0.733 ± 0.02 μs	1.14
overhead/acrule/a+2+b	0.816 ± 0.022 μs	0.722 ± 0.016 μs	1.13
overhead/acrule/a+b	0.264 ± 0.0083 μs	0.267 ± 0.011 μs	0.989
overhead/acrule/noop:Int	25.9 ± 0.91 ns	26.2 ± 0.92 ns	0.988
overhead/acrule/noop:Sym	0.0369 ± 0.0055 μs	0.0367 ± 0.0061 μs	1
overhead/rule/noop:Int	0.0444 ± 0.00051 μs	0.0443 ± 0.0011 μs	1
overhead/rule/noop:Sym	0.0559 ± 0.0026 μs	0.0546 ± 0.0026 μs	1.02
overhead/rule/noop:Term	0.0558 ± 0.0025 μs	0.055 ± 0.0027 μs	1.02
overhead/ruleset/noop:Int	0.129 ± 0.0026 μs	0.129 ± 0.004 μs	0.999
overhead/ruleset/noop:Sym	0.157 ± 0.0055 μs	0.152 ± 0.0042 μs	1.03
overhead/ruleset/noop:Term	3.27 ± 0.16 μs	3.03 ± 0.11 μs	1.08
overhead/simplify/noop:Int	0.138 ± 0.0023 μs	0.14 ± 0.0028 μs	0.984
overhead/simplify/noop:Sym	0.153 ± 0.0066 μs	0.153 ± 0.006 μs	0.997
overhead/simplify/noop:Term	0.0369 ± 0.0023 ms	0.0359 ± 0.0019 ms	1.03
overhead/simplify/randterm (+, *):serial	0.0904 ± 0.0012 s	0.087 ± 0.0012 s	1.04
overhead/simplify/randterm (+, *):thread	0.0522 ± 0.028 s	0.0502 ± 0.031 s	1.04
overhead/simplify/randterm (/, *):serial	0.218 ± 0.0078 ms	0.212 ± 0.0072 ms	1.03
overhead/simplify/randterm (/, *):thread	0.247 ± 0.009 ms	0.242 ± 0.0081 ms	1.02
overhead/substitute/a	0.0636 ± 0.0016 ms	0.0583 ± 0.0016 ms	1.09
overhead/substitute/a,b	0.0559 ± 0.0015 ms	0.0503 ± 0.0014 ms	1.11
overhead/substitute/a,b,c	18.7 ± 0.75 μs	16.9 ± 0.68 μs	1.1
polyform/easy_iszero	30.2 ± 2 μs	28.5 ± 2 μs	1.06
polyform/isone	2.79 ± 0.01 ns	3.1 ± 0.01 ns	0.903
polyform/iszero	1.17 ± 0.032 ms	1.11 ± 0.031 ms	1.05
polyform/simplify_fractions	1.67 ± 0.045 ms	1.59 ± 0.039 ms	1.05
time_to_load	4.48 ± 0.021 s	4.46 ± 0.04 s	1.01

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[Vaibhavdixit02]

Thank you @bowenszhu!!

[Vaibhavdixit02]

Works for me

[YingboMa]

Yeah, this is the correct fix. Thanks!

[YingboMa]

@bowenszhu could you merge and release this fix when you feel comfortable?