The SLI does not meet several of the requirements in our Statement of Requirements. We do not consider this to be a problem however as from the outset we have known that our SLI implementation was unlikely to fulfil the ambitious requirements of our investigation.
There are several areas we can identify as being deficient in our implementation. The cell referencing is not fully implemented. The interface does not provide any editing facilities further than modifying the formula line and selecting a single cell. If a formula such as '1+1' is entered it will not parse, it should be parsed as a string. However '=1+1' is parsed correctly as an expression. There is a memory leak in the Interpreter, if the same image is requested twice then one will be stored in the image cache but the other will not be disposed of. Bad referencing has not been tested as insertion and deletion of rows and columns is not implemented. Strings cannot start with a number. The interface does not implement widgets. A formula containing nested image processing functions will not be evaluated correctly. If a dependency cycle is detected then all cells are treated as being in the cycle.
The interpreter in particular has proved to be very difficult to implement.
The SLI can however accept and process numeric formulae, load images into cells from the formula line and display larger images of cells.
The most positive aspect of the SLI is that the design can support most of the aspects we have considered. The design is very extendible and provides the framework for a much fuller application with minimal redesign. There are several features that are partially implemented such as cell referencing and priorities for large images.