⇐ ⇒

[CF-metadata] time

From: Russ Rew <russ>
Date: Sat, 02 Dec 2006 11:16:20 -0700

Jonathan,

> Yes, I agree, ragged arrays (in netCDF-4) would be a good solution to some
> 2D time problems.

That's what I used to think also, but details of how the underlying
HDF5 variable-length types work make me less convinced that we can use
ragged arrays in this context.

The limitation is that a row of a ragged array must be accessed
atomically: the entire row is returned in allocated memory space,
along with its actual length. This is fine if the granularity of the
elements is small, such as for a row of floating-point values or small
data structures, but may be impractical for a ragged array of larger
elements such as time slices of model forecasts. This is a
characteristic of the underlying HDF5 variable-length types, so may
not be easy to work-around short of using lots of memory.

--Russ
Received on Sat Dec 02 2006 - 11:16:20 GMT

This archive was generated by hypermail 2.3.0 : Tue Sep 13 2022 - 23:02:40 BST

⇐ ⇒