[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 20010827: Add/Remove layer in netCDF

>To: address@hidden
>From: Jean-Marc SINQUIN <address@hidden>
>Subject: Add/Remove layer
>Organization: IFREMER / Centre de Brest
>Keywords: 200108271029.f7RATH106729 netCDF rewrite

Hi,

> We use 3.5 netCDF interface, for managing multi-grid datas. Thus, we
> need to remove and add grids in a one file.
> 
> If we change a layer (with a different size), we must copy all the
> layers before and then add the one.
> 
> Is it possible to perform the process (without copying the layers) ?

Sorry, but no, not in general.  But if you make use of the "UNLIMITED
dimension", you can add data along the UNLIMITED dimension without
copying all the previous data.

If the layers don't differ much in size, you could allocate the
maximum size for each grid and use the UNLIMITED dimension for adding
grids without copying previous grids, but it would waste space for
grids less than the maximum size.

Also, there is no easy way in netCDF to remove data, for example a
grid, and recover the space of the removed data without rewriting the
file.  About the best you can do is keep information in a separate
array that specifies whether a block of data has been removed and
reuse it if it happens to fit a new block of data you want to write.

NetCDF is not a database system that uses sophisticated storage
allocation schemes; in particular, netCDF has no implementation of
"garbage collection".  Instead it supports efficient access to
multidimensional arrays by making it easy to compute the location of
any data in the file from the static shapes of declared variables.

If you need to efficiently change more than one dimension in a file or
add/delete variables, you may want to consider using another file
format such as HDF-5 that permits such flexibility.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu