Creating-Destroying

We provide several different levels of sophistication for creating sep_90 types:

• sep_create

  sep_create provides the user the most flexibility in creating a sep_90 type. It can be used to create

  • a simple data sub-space described entirely by its dimensions
  • a SEPlib77 simple dataset
  • a SEPlib90 irregular dataset with headers and traces
  • numerous mutations of the above (irregular without headers, only header information, etc.)
  sep_create, like most of the SEPF90 library takes full advantage of Fortran 90's overloading and optional abilities. As the user desires higher levels of complexity the call to sep_create expands:

  add sepdata,n

  To create a sep_90 dataset where o, d, and label are extraneous. Where sepdata is a type(sep_90), pointer and n is an integer array describing its dimensions. To creating a dataset that is analogous to a SEPlib77 dataset.

  add o,d,label.

  To create a SEPlib90-like dataset

  add n_keys, keyname, keytype, keyformat, mask (or tr_num)

  . The user has to additionally supply information about the number (n_keys, integer), name (keyname, character array), type (keytype, character array), format (keyformat, character array), and potentially information about data irregularity by passing a logical or integer array describing where traces exist (mask, logical array or tr_num, integer array).

• sep_90_alloc and sep_90_duplicate

  These two routines will create a new sep_90 dataset from another sep_90 type. The first simply copies the shape, the second copies not only the shape but values. The interface to both takes the form:

		         call sep_90_alloc(old,new)

• sep_reed

  The final, and easiest, method to create a sep_90 dataset is to simply read it off disk using the sep_reed routine described below.

All of these methods must be preceded a call to init_vars which allocates the sep_90(s) and sets up some initial defaults.

Freeing a sep_90 dataset is accomplished through a call to cleanup_sep

 call cleanup_sep(sepdata)