EWD249 - 72

this type, again we abstain from introducing a new identifier (wisely or not).

Then, "print", "clear", and "mark pos", being operations that were understood as operating on an "image" are translated in algorithms expressed in terms of operations on a line. In the code of these algorithms, the (true) actual parameter specifies which line; at the end of the description we give the instruction list, indicating that the actions operate on "a line"; we have given the type, but not the parameter.

This level introduces some new features. To start with (as in explaining "image") we treat the structural refinement of a data type on a footing very similar to the algorithmic refinements (as applied to "print", "clear" and "mark pos"). Before this level, our approach could have been regarded as an effort to establish a discipline for "subroutinization" -if the reader will excuse this horrible term!-, now we observe that that characterization of our effort covers only half of what we are trying to do, as we are trying to apply a similar technique to data structures as well. Secondly, our previous machine explained just one entity (instruction or data type) in contrast to "LINER", which explains a whole bunch of them. The point is that we try to associate with each level a separate design decision; the decision take here is to understand the image from now onwards in terms if lines, and therefore all operations dealing with an image as such have to be translated in terms of operations dealing with its lines. The image has been "explained away", the only unusual type we still have to deal with is the type "line" and that is what we are going to do now. I draw your attention to the fact that in the level to come, we have to deal with lines: that lines are used to compose images from is no longer relevant!

To represent a line we have many different possibilities, e.g. a list of the x-co-ordinates of the positions where a mark should be printed (possibly sorted in order of increasing x-value), a boolean array of 100 elements, each element indicating whether the corresponding position on the line of the picture should be marked, or an integer array of 100 elements, each element having the value "mark" or "space" of the PRSYM-parameter for the corresponding printable position. The last representation caters for extension when different curves (with different marks) have to printed in the same picture; therefore we select the last one. This leads to