When the SPLIT option is chosen from the above menu, the user is called upon to design a polygon which demarcates the intended division. That part of the initial grid which lies inside the splitting polygon will eventually be output as an independent grid, and the remaining part outside the polygon will be output as another grid. Each of the output grids will be in NEIGH format and will have its nodes numbered consecutively from 1 upwards. Note: A current bug in the Splitter makes it inadvisable to place a polygon vertex directly at a boundary node of the grid: a triangle may consequently be dropped from the sub-grid lying inside the polygon when the split takes place. Avoid this by finding the coordinates of the boundary node using the INFO option before choosing SPLIT and extrapolate the required splitting line to find a suitable vertex position outside the grid.
The first prompt which appears after SPLIT is chosen is:
Normally, the polygon vertices will be placed by means of the cursor, option C, but the X option is useful occasionally, e.g. when repeating a previous run, for instance when practising with the demonstration data. Coordinates are always given in problem length units. On exit from the Splitter, a file of polygon vertex coordinates with the filename POLY.DAT is output automatically to facilitate such repeat runs. The next prompt asks the user to indicate the desired location for the first vertex of the polygon, after which the following prompt appears:
Vertices should be entered either in clockwise order or counterclockwise order, since they are connected and displayed in order of entry. Choice of location by cursor or coordinates is offered for each vertex. Currently, up to ten vertices are allowed. After the last vertex has been positioned, the user can either confirm or cancel the displayed polygon. Assuming that a satisfactory polygon has been drawn, the next step is to decide which side of the polygon to work on, i.e. to choose one of the sides of the polygon which actually intersects the grid, and then move nearby nodes to this line, so that both sub-grids produced will have boundary nodes lying on the splitting polygon. The relevant prompts are:
At this point, in response to the prompt:
there is a choice of methods for moving nodes to the splitting line. Manual movement of nodes to the splitting polygon requires choosing both the node and its new position by cursor. Under option A - automatic, so far as internal nodes are concerned, the user merely has to indicate with the cursor which node to move and the node is then moved to the current working side of the splitting polygon along a perpendicular through its original position. However, boundary nodes are still moved manually, as the automatic move option might result in a node being moved away from the original boundary, particularly when the boundary and the splitting line are not approximately perpendicular to one another. Whether automatic or manual moving of nodes is chosen, each move is displayed for confirmation or cancellation.
Note that in choosing nodes to move to the splitting line, it is important that consecutive nodes moved should be connected to one another, i.e. be neighbours. An example of an improper sequence of nodes is A-B-C-D in Figure 4.xx (not yet available). In this case, the connection EF still crosses the splitting line after nodes A,B,C and D have been moved. On the other hand A-B-F-D or A-E-F-D would be acceptable sequences. Where one of the sub-grids output by the Splitter is going to be discarded, for example when some unwanted extension of a grid is being removed, it is normally better to move unwanted nodes to the splitting line, rather than choosing nodes which will be retained in any case. It is then easier to maintain well-shaped triangles in the remaining sub-grid. Prior to moving a node to the splitting line, the user has a choice of the following:
Option W is worth considering before moving any nodes; it brings up the display control menu:
Note that menu selection is carried out with the mouse. The functions in this menu provide control over windowing (zoom) and refreshing the display. Up to 6 levels of windowing are allowed. When moving nodes to the splitting polygon, it is advisable to zoom in until nodes to be moved are at least 0.5 cm apart on the screen. This may involve displaying only part of the current polygon side; opportunity to zoom out again is offered at appropriate times. When ready to move a node to the current working side of the splitting polygon, choose option M (move node) from . This will bring up further prompts appropriate to automatic or manual node moving. After the node has been moved, the user will be asked to confirm or cancel the move. One situation in which a move may have to be cancelled is if the wrong node is selected due to insufficient sensitivity of the cursor. This will be corrected automatically by the program if the move is cancelled and the selection is repeated. There is a bug in the Splitter at present which results in occasional failure to restore the previous display when a move is cancelled; this can be ignored, since the grid file is in fact handled correctly by the program. Continue moving nodes to the splitting line until connections between nodes coincide with the line along its whole length. Note that a node should be "moved" even if it originally lies on the line; otherwise the program is not informed that such a node is now a boundary node. If in doubt whether a node has been moved or not, check the colour of lines connecting it to its neighbours: all of these will be in the "modify colour" if the node has been moved. There is no harm in moving a node twice, if in doubt. The completion of a side is signalled by using option D in , but before doing so, it is necessary to window out, using option W, to the extent that the whole length of the current working side is visible in the display. Then when D is subsequently entered, it will always be possible to see which node prompt , below, refers to. When moving nodes to the current working side of the splitting polygon is complete and this has been signalled to the Splitter program by entering D in answer to prompt , a marker is placed at each moved node in turn and the following prompt appears:
The code entered by the user will be assigned to the node in question in both sub-grid files output by the splitter. Water depth at the new location of a moved node is evaluated by linear interpolation, if it is originally an internal node. For a moved boundary node, the user has the option of leaving the depth equal to its value before the node was moved or entering a new value of depth. When codes have been entered for all nodes moved to the current working side, the following prompt appears:
If more than one side of the splitting polygon intersects the grid, and not all such sides have had nodes moved to them, option M should be chosen. The first subsequent step is to window up, if necessary, to make sure that the whole of the next side to be worked on is visible. The user will then be lead through the same procedure for the next working side as for the previous one. Eventually, when nodes have been moved to all sides of the splitting polygon which intersect the grid, choose option F, which initiates the actual splitting process.
The user is then led through a series of steps concerning display and output of the two sub-grids produced. When these have been carried out, splitting is complete and control returns to the top menu, where the EXIT option should be chosen.