set

<handle> set ?-vtype char|short|int|long|longlong|float|double|string|compound? ?-fields {<name> <subtype> <name> <subtype>...}? ?-compress 0-9? ?-dimensions {<nr> ?<nc>? ?<nd>?}? <name> <value> ?<name> <value>?

Creates or updates a dataset. name must be the full path name of the dataset (for example: /group1/group2/datasename).

value is a vector. If it is already a intarray or doublearray tcl_obj, the type is inferred from it. Otherwise, it has to be specified with option -vtype.
If -vtype is 'string' then value must be a list of strings.
-dimensions specify the number of dimensions of the array. Must be a list of one, two or three integers that specify the size or each dimension.
by default it is assumed that the number of dimensions is 1, and the size of this dimension is the amount of values
nr is the number of rows.
nd is the number of 'depths'
nc is the number of columns to represent a bidimensional array, by default is a single vector (nc==1).
The array value must contain the values ordered by rows (C-like), not by columns (FORTRAN-like), and off course the amount of data must be a multiple of nc.
The first size nr could be set to 0 to mean that this number must be according of the amount of provided values.
Instead of -dimensions {0 nc} it is possible to use (for back compatibility) the deprecated syntax -ncolumns nc (that was replaced because only allow 2 dimensional arrays).
If -vtype is 'compound' then is necessary to specify the struct components with -fields followed by a list of paired 'field name' and 'field subtype'.
'field subtype' could be: char|uchar|short|ushort|int|uint|long|ulong|llong|ullong|float|double (unsigned integers starting by 'u', and llong meaning 'long long')