Notes from the README

[odow]

XpressAPI.jl/README.md

Lines 9 to 10 in f20e039

	The goal of this package is *not* to provide a full-fledged Julia API or even
	a modeling API. These things can be built on top of this package.

Is this still true?

XpressAPI.jl/README.md

Line 45 in f20e039

See [MOI_FEATURES.md](MOI_FEATURES.md) for a comprehensive list of:

Does not exist

XpressAPI.jl/README.md

Line 258 in f20e039

**Custom User-Defined Operators**: Supports JuMP's `@operator` macro for user-defined functions with provided derivatives.

Niiiiice! Did it all work properly? Any issues?

XpressAPI.jl/README.md

Lines 271 to 272 in f20e039

	| `MOI.VectorQuadraticFunction` | `MOI.Indicator{MOI.ACTIVATE_ON_ONE, S}` | Quadratic indicator |
	| `MOI.VectorNonlinearFunction` | `MOI.Indicator{MOI.ACTIVATE_ON_ONE, S}` | Nonlinear indicator |

Nice! These are missing from Xpress.jl.

XpressAPI.jl/README.md

Lines 363 to 381 in f20e039

	### Custom Objective Bridge
	XpressAPI provides a **custom objective slack bridge** (`StrictObjectiveSlackBridge`) that reformulates non-native objective types:
	**Transformation**:
	```
	min/max F(x)
	```
	becomes:
	```
	min/max c
	s.t. F(x) - c == 0
	```
	This bridge enables Xpress to solve problems with objective functions that would otherwise require reformulation.
	**Automatically applied for**:
	- Nonlinear objectives (`MOI.ScalarNonlinearFunction`)
	- Any other objective function type not natively supported by Xpress

We should revisit this. It really shouldn't be needed.

XpressAPI.jl/README.md

Lines 383 to 394 in f20e039

	### MOI Bridge Support
	XpressAPI works seamlessly with MOI's automatic bridging system, which can transform:
	- Conic constraints → Quadratic constraints (e.g., `MOI.SecondOrderCone`)
	- Interval constraints → Separate bounds
	- And many more transformations
	To use bridges with XpressAPI:
	```julia
	using JuMP, XpressAPI
	model = Model(() -> MOI.Bridges.full_bridge_optimizer(XpressAPI.Optimizer(), Float64))
	```

This is not needed. JuMP adds bridges by default. Users need model = Model(XpressAPI.Optimizer).

[AdamFDC]

Thanks for all the remarks! I just had time to come back to the package, I'll address everything that you raised.
We are still indeed required to make the claim of not delivering a full API, even tough development is active.

For custom user defined operators; they work well ! In the backend we create a callback that is treated as a grey-box callback by Xpress. You have to provide evaluation for you callback; First and second order derivatives are optional. Using Forwardiff.jl is very handy for this.

I will indeed do an overall of by custom bridges and NL objective support.