Build the variable definitions for one level of a nested dataset

Each sampling level becomes a netCDF-4 group. The link to the parent is an explicit index variable, not repetition of the parent's columns — that is the whole reason for netCDF-4 here rather than a flat table: tow effort is stored once, and a length-frequency bin points at the tow it came from. Flattening instead repeats each net's volume_sampled onto every one of its size bins, which turned 76,512 real ichthyo values into 369,978 repeated ones and inflated any naive SUM() of effort by ~5x.

Usage

nc_level_vars(
  group,
  df,
  dim,
  parent_dim = NULL,
  parent_index = NULL,
  var_meta = list(),
  strlen = 64L
)

Arguments

group

Group name, e.g. "tow", "occurrence", "length_bin". Pass "" to write at the file root instead of in a group, which is what a CF featureType=point collection is — one flat dimension with no nesting.

df

data.frame for this level, ordered so children are contiguous.

dim

The ncdim4 for this level.

parent_dim

Parent level's ncdim4, or NULL at the root.

parent_index

1-based index into the parent level, one per row of df.

var_meta

Named list from measurement_var_meta().

strlen

Fixed character length for string variables.

Value

Named list of ncvar4 objects to pass to ncdf4::nc_create(). The parent link, when present, is the element named __parent_index.