# `cube_collect()`
**Category:** Grid geometry
**Status:** stable

## Purpose
Materialise a lazy cube in memory.

## When to use
Use it when its declared input and output match the workflow. Prefer a narrower selection before any operation that may materialise values.

## When not to use
Do not use it to calculate `spatind` indicators or to bypass coordinate, CRS, unit, and provenance checks.

## Real signature
```r
cube_collect(x)
```

## Arguments
- `x`: see the installed Rd manual for validation and defaults.

## Input and output
Input: **ocean_cube**. Output: **ocean_cube**.

## Backend support
Memory: **TRUE**. NetCDF: **TRUE**. Backend operations retain read-only semantics; materialisation is explicit.

## Dimensional effect
May change shape or return a non-cube analytical object; inspect dimensions/rows.

## Metadata
Units, coordinates, provenance, masks, and QA metadata are retained or reported where meaningful.

## Minimal executable check
```r
library(oceancube)
stopifnot(is.function(oceancube::cube_collect))
names(formals(oceancube::cube_collect))
```

## Example in a workflow
Typical position: `read_nc` → `cube_collect()` → `cube_slice/cube_crop`. See the workflow scripts for a checked end-to-end example.

## Frequent errors and limitations
Invalid dimensions, selectors, units, CRS, source-file identity, optional dependencies, or unsupported grids are rejected explicitly. Read the function Rd page for exact diagnostics.

## Related functions
Previous: `read_nc`. Next: `cube_slice/cube_crop`.