Introduction
============

The proposed `CF aggregation rules
<https://cf-pcmdi.llnl.gov/trac/ticket/78>`_ allow for the aggregation
of `CF <http://cf-pcmdi.llnl.gov>`_ fields across multiple
dimensions. These rules are based solely on the fields' metadata,
therefore allowing such aggregations to be reliably automated. This
raises the possibility of large amounts of aggregations being created
and therefore storing these collections is clearly desirable.

This document describes how the data arrays of such aggregations could
be stored in the memory of an application and proposes a convention --
the NCA (`netCDF <http://www.unidata.ucar.edu/software/netcdf/>`_
aggregate) convention -- for their efficient file storage.

The key element of the NCA files is that they are CF-like netCDF
files, which only require extra processing to realise their aggregated
data arrays.

These ideas have been implemented in `cf-python
<http://cfpython.bitbucket.org>`_. In particular, its `Large Amounts
of Massive Arrays (LAMA)
<http://cfpython.bitbucket.org/docs/0.9.6/build/lama.html>`_
functionality stores arrays in this fashion and it can read and write
files according the NCA convention as described here.

This style of aggregation storage is a generalization of `NetCDF
Markup Language (NcML) storage
<http://www.unidata.ucar.edu/software/netcdf/ncml/v2.2/Aggregation.html>`_,
which has long-standing use in the community, in that it allows:

* Simultaneous aggregation across more than one dimension.

* Aggregation accounting for different but equivalent variable
  properties (order of dimensions, number of size 1 dimensions, senses
  in which dimensions run, units of the data values, missing data
  values)

* Storage of changes to arbitrary (not necessarily contiguous) parts
  of the aggregated data array.

* Aggregation of arrays stored in a mixture of formats (in-memory,
  netCDF, PP, *etc.*).