> Russ: > > Thank you very much for your answers. They are very helpful. But some new > questions raise up. Regarding the first question, a variable has a fixed size > or shape, which is defined by its dimensions. That's right except that the shape can grow along a dimension declared as "unlimited". So a variable could have a shape such as (time, lon, lat) where lon and lat are fixed size to represent a latitude-longitude grid, but time could be declared "unlimited", so that more data can be added along the time dimension for later time steps. A netCDF-3 file can have only one unlimited dimension, but a netCDF-4 file may have multiple unlimited dimensions. > No matter how many variables stored in a netCDF file, they have to have the > same size or shape or dimensions if they are stored in the same netCDF file. > Is my understanding right? No, that's wrong. Dimensions of different sizes may be used in declaring variables of different shapes in the same netCDF file. Variables that share dimensions can be defined on the same grid, but variables on different grids using completely different dimensions can also be defined in the same file. > Then, How many dimensions can a variable have? Any limitation? A variable may have up to 1024 dimensions, although typically variables have 0, 1, 2, 3, or 4 dimensions. 1024 different dimensions may be defined in a file, so you could have lots of variables defined on different grids. > As for the second question, can I consider the structure of netCDF as the > file system under Windows? Hierarchical, root, branch, multiple folders or > directories containing a variety of files? No, you can't mount a netCDF file as a Windows file system. However a netCDF-4 file is similar conceptually, in that groups are organized hierarchically and each group has its own name space or scope for naming dimensions, variables, and attributes. The top-level unnamed group is analogous to the root directory in a file system. Each group can have any number of subgroups. However, you can't store a file in a group, only netCDF metadata and data objects. > Can I install NcML on Windows platform? NcML creates binary files different > from netCDF? Yes, NcML is part of netCDF-Java. It is pure Java, so it works fine on Windows: http://www.unidata.ucar.edu/software/netcdf-java/ NcML doesn't "create" netCDF files, it "wraps" them to provide additional metadata or to aggregate multiple netCDF files and make them appear as if they are one netCDF file. NcML can wrap files in other formats as well that are supported by the Unidata Common Data Model (CDM). This provides a way to leave files in their original format, but make them appear to the netCDF-Java library as if they are netCDF files or conform to netCDF metadata conventions. NetCDF-Java can read or write netCDF-3 files and read netCDF-4 files, but it cannot yet write netCDF-4 files. --Russ > ________________________________ > From: Unidata netCDF Support <address@hidden> > To: address@hidden > Cc: address@hidden > Sent: Thursday, May 3, 2012 6:00 PM > Subject: [netCDF #MKO-964114]: some questions > > Hi Doug, > > > I am a new user of netCDF and wish you could help me with my questions. > > 1. Within one netCDF file, can variables have different dimensions? In my > > imagination, netCDF stores one variable (array) in one layer. Each layer > > may not have relation with others. In this way, variables in one netCDF can > > have different dimensions. > > No, a single variable is defined to be a specific shape, determined by the > list of dimensions > it uses. > > However, netCDF-4 files have named "groups", and each group may have > dimensions and variables > with the same names but different sizes or shapes as dimensions and variables > in other groups. > It is analogous to the situation with files in directories, where the same > name can refer to > different files in different directories of a hierarchical file system. > > So maybe what you are calling "layers" are like netCDF-4 "groups". > > > 2. When we create a netCDF file, do we have to put all data in the CDL > > file? It may not be complex but time consuming. Can we link a well-defined > > external file to the CDL file? For example, indicate the external fiel > > location on the hard drive. > > No, you don't need to use CDL files (which are human-readable text files).Â > NetCDF files > (which are binary files) may be created directly with custom programs or > utilities. > Typically, you either use a CDL file to create metadata, then the ncgen > utility to turn > that into a binary netCDF file.Â Or you use a program to create e netCDF > file, then use > ncdump to see that as a CDL file.Â But you don't need to manually modify a > CDL file to > keep it consistent with a netCDF file (which I think you are asking about, if > I understand > your question). > > There is no standard or convention for including data by reference or linking > in a binary > netCDF file, although you could store file names as values of string type > attributes or > variables.Â However, netCDF Java supports NcML files, which can store links > to external > data in one or more other files, which can be netCDF files.Â For more > information about > NcML, and especially the use of NcML for aggregation, here: > > Â http://www.unidata.ucar.edu/software/netcdf/ncml/ > > --Russ > > Russ RewÂ Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â UCAR > Unidata Program > address@hidden Â Â Â Â Â Â Â Â Â Â http://www.unidata.ucar.edu > > > > Ticket Details > =================== > Ticket ID: MKO-964114 > Department: Support netCDF > Priority: Normal > Status: Closed > Russ Rew UCAR Unidata Program address@hidden http://www.unidata.ucar.edu Ticket Details =================== Ticket ID: MKO-964114 Department: Support netCDF Priority: Normal Status: Closed
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata inquiry tracking system and then made publicly available through the web. If you do not want to have your interactions made available in this way, you must let us know in each email you send to us.