IBM Visualization Data Explorer Programmer's Reference

[ Bottom of Page | Previous Page | Next Page | Table of Contents | Partial Table of Contents | Index ]


4.3 Data Explorer Execution Model

The executive executes your module by calling the C function that you provide and passing it two arguments: one pointer to an Array for the input Object(s); another to an Array for the output Object(s) the module creates. In each case, the size of the Array is defined by the corresponding mdf file and the number of input and output statements found there.

If a parameter is not specified by the user when the module is called, the corresponding element of the input Array is set to NULL.

As noted earlier, Data Explorer modules are generally required to be pure functions, producing the same results from the same inputs. The reason for this requirement is that the executive caches the results of module execution and does not reexecute the module (given the same input) if the previously computed result is still in the cache.

In particular, modules must not maintain state (e.g., by saving values in global or static variables). Indeed, it may be impossible for a module to maintain state in a multiprocessor environment, since it may execute on a different processor each time. However, a module may use the cache to maintain information that makes succeeding invocations more efficient, provided that it maintains pure-function semantics. (For more information, see 13.5 , "Cache".)

Outboard modules whose PERSISTENT flag is set (see page reference #2) may maintain state, but they are still required to maintain pure-function semantics: the executive might not execute a module if its inputs have not changed and the results are still in the cache. Modules that have no outputs are executed every time a visual program using them is run, regardless of whether or not their inputs have changed. See also Chapter 4. "Data Explorer Execution Model" in IBM Visualization Data Explorer User's Guide.
For Future Reference

Although you must supply a C function for Data Explorer to call, it is still possible to write the bulk of a module in FORTRAN: Write a C "wrapper" that (1) extracts the data from the input Object (using the programming interface described in this manual) and (2) passes the data to the FORTRAN subroutine.

You should be aware of the following when writing a module that interfaces to a FORTRAN routine:

  • FORTRAN variables are always called by reference. For example, if you are using the Module Builder you need to modify the worker routine so that it passes the addresses of the parameters.
  • Since FORTRAN routines do not return a value, the ERROR (or OK) return value must be a parameter.
  • Depending on the compiler, it may be necessary to affix an underscore to the name of the FORTRAN routine and to lowercase the name of the worker routine.
  • Passing strings from C to a FORTRAN routine may require passing both the string length and the string pointer. Consult the appropriate compiler manual.


[ Top of Page | Previous Page | Next Page | Table of Contents | Partial Table of Contents | Index ]
[Data Explorer Documentation | QuickStart Guide | User's Guide | User's Reference | Programmer's Reference | Installation and Configuration Guide ]

[Data Explorer Home Page | Contact Data Explorer | Same document on Data Explorer Home Page ]


[IBM Home Page | Order | Search | Contact IBM | Legal ]