## ProductInputs


Everything [cubegenpy.writer.build_hdulist()](build_hdulist.md#cubegenpy.build_hdulist) needs, as one object.


Usage

``` python
ProductInputs(
    cube,
    raw_counts,
    cal_factor,
    wavelength,
    header,
    geometry,
    kernels,
    bodies,
    resolved_bodies,
    dims,
    rings_in_fov=False,
    edge_on=False
)
```


Field names and meaning mirror [build_hdulist](build_hdulist.md#cubegenpy.build_hdulist) one-to-one. `cube` is the only field a `~cubegenpy.sources.Source` may leave as `None` -- the `~cubegenpy.calibrate.Calibrator` computes it from `raw_counts` and `cal_factor` and the builder fills it in via `dataclasses.replace()`.


## Parameter Attributes


`cube: np.ndarray | None`  

`raw_counts: np.ndarray`  

`cal_factor: np.ndarray`  

`wavelength: np.ndarray`  

`header: Mapping[str, object]`  

`geometry: Mapping[str, Mapping[str, object]]`  

`kernels: Sequence[tuple[str, str]]`  

`bodies: Sequence[str]`  

`resolved_bodies: Sequence[str]`  

`dims: Dims`  

`rings_in_fov: bool = ``False`  

`edge_on: bool = ``False`  


## Array Conventions

These are the contract between this package and `pyuvis`, and they are conventional rather than enforced by any import -- which is exactly why they are written down here. Everything on this class is in **caller order**; the reversal to the proposal's FITS order happens once, in `cubegenpy.writer._to_fits_order()`, and is undone once, in [cubegenpy.fitsio.read_product()](read_product.md#cubegenpy.read_product).

================== =========================================================== `cube` `(NX, NY, NZ)` float32. NaN marks undefined. `raw_counts` `(NX, NY, NZ)` **uint16**, PDS3 convention: 65535 is the null, not a count. Written to FITS as signed int16 (or int32 when a real value exceeds 32767) with the null recoded to -1 and declared via `BLANK`. `cal_factor` `(NX, NY)` float32. `wavelength` `(NX,)` float32, **Angstrom** -- note `pyuvis` returns nm, so the production path must convert. geometry cells `(nrows, *reversed(tdim))`: a `5,NY,NZ,NT` backplane arrives as `(nrows, NT, NZ, NY, 5)`. ================== ===========================================================

`pyuvis` reads PDS3 qubes with `reshape(CORE_ITEMS, order="F")`, so its arrays are Fortran-ordered with shape `(NX, NY, NZ)` -- which is what makes the writer's transpose a no-copy view and lets the FITS data section reproduce the PDS3 byte sequence.


## Methods

| Name | Description |
|----|----|
| [as_writer_kwargs()](#as_writer_kwargs) | Return the keyword dict to splat into [writer.build_hdulist()](build_hdulist.md#cubegenpy.build_hdulist). |

------------------------------------------------------------------------


#### as_writer_kwargs()


Return the keyword dict to splat into [writer.build_hdulist()](build_hdulist.md#cubegenpy.build_hdulist).


Usage

``` python
as_writer_kwargs()
```