Data Access

Historically Asterix applications have used only one data access method for bulk data handling, namely file mapping. This has three disadvantages,

• It encourages the mode of programming where all the data is read in, manipulated, and then perhaps written back out. While this method is sometimes required, it is very inefficient for large datasets. For many applications a single pass with a modest buffer would suffice.
• Mapping returns a pointer to a place in the process' address space as an integer which must be dereferenced using the %VAL function before the data can be used at a low level. This inevitably forces programs to be split into two levels. The combination of the mapping process and the resulting requirement to split the code makes programming noddy analysis procedures unnecessarily complicated.

For these reasons ADI supports two data access modes in addition to mapping. The first of these is a simple get/put interface where the entire data object is read into or written from Fortran variables in one fell swoop. This makes the life of the amateur programmer considerably simpler.

The third access mode is buffered access. ADI buffers can be defined to scan over a data object or objects with buffer sizes under either user or automatic control. Access to buffered objects can be synchronised to (for example) step over multiple data members of a class at the same rate.