Array length: Difference between revisions

Early fortrans offered no protocol for ascertaining the length (or dimensionality) of arrays, though the compiler might havehas this information. Thus a statement such as <code>PRINT A</code> would print all the elements of a variable <code>A</code> according to its definition. A subprogram that received a parameter would have no access to these details, so its parameter might be declared as <code>A(12345)</code> simply to signify that it was an array (rather than an ordinary variable) and the programmer would rely on other data to know the boundsupper bound to employ, for instance via an additional parameter. ''Any mistakes would cause crashes!'' On the other hand, with heavy computational tasks, it was common to take advantage of the opportunities. Thus, a subprogram might regard its array parameter as one-dimensional even though the actual parameter was not. Carefully-programmed routines might thusly process a sequence of elements via 1-D indexing, far faster than the 2-D or higher order indexing of the original. Success at this game required understanding how array elements were arranged in multidimensional arrays.

Later fortrans allowed <code>A(*)</code> to signify an array parameter of unstated boundsupper bound, but there was still a problem with higher dimensions. All but the last dimension has to be stated correctly if a multi-dimensional array parameter is to be indexed correctly.

With Fortran 90, a new protocol was introduced, whereby the parameter might be declared as <code>A(:)</code> signifying an array of one dimension, of bounds unstated. A 2-D array would have <code>A(:,:)</code> and so on. Further, arrays could have arbitrary lower bounds as well, as in <code>A(-7:12)</code> but if no colon appeared for a dimension, the lower bound would be assumed to be one so <code>A(2)</code> means an array of two elements, as before. And as before, a bound could be explicitly stated, perhaps via an explicit parameter such as <code>N</code>, but now, the compiler is passing secret additional parameters to the subprogram giving the bounds of the array, and these can be accessed via the library functions LBOUND and UBOUND. For multi-dimensional arrays there are multiple bounds, and an invocation might be <code>UBOUND(A,DIM = 2)</code> but in the example only a one-dimensional array is involved. These facilities are available only if the new MODULE protocol is employed.

CHARACTER*6 ARRAY(2-1:1)

ARRAY(-1) = "Apple"

ARRAY(20) = "Orange"

Lower bound 1 , Upper bound 23