When the user selects the `node' command in the `Elements' menu, the tool displays a node-addition cursor under the current mouse position on the dataflow canvas, as shown in Figure 3.

To create a node, the user positions the cursor at the desired placement location, then presses and releases the left mouse button at that location. In response, the tool draws the node, with the upper left corner of the smallest enclosing rectangle at the pressed location, as shown in Figure 4.

As the user types, the text remains centered within the node. If the user presses the Enter key while typing, a new line of text begins, with the text remaining centered within the node. If as the user types, the horizontal or vertical size of the text becomes larger than the diameter of the node, the tool dynamically resizes the node so that the text does not extend beyond the node boundary. The minimum distance between any portion of the text and the node boundary is the value of the node-name margin, which by default is 1 em in the current node-name font. The the user can change the default node- name margin with an option setting, as described in ???.

If the user deletes text to make the size of the name smaller, the system dynamically contracts the size of the node, such that the distance between the text and node boundary is no larger than the current node-name margin value. The tool stops node contraction whenever the text reaches a size where it fits within a node of the node diameter setting, which by default is 10 em of the current font. The the user can change the default node diameter with an option setting, as described in ???.

The user signals the completion of node-name typing by clicking the mouse some place other than within the text of the name. If the user performs such a click without having typed any text for the node name, the tool types a default name of the form "NodeN", where N is a positive integer ... (see the phrasing in the cal tool rqmts).

Node names may contain only the following characters:

• lowercase and uppercase letters
• digits
• underscores
• blank and newline characters (referred to as "whitespace")

Formally, the name of a node does not include any whitespace characters. Therefore, the formal name of a node consists of the concatenation of all of its characters with all whitespace removed. So, for example, the following are the same name formally:

"ActivateAutopilot"

"Activate Autopilot"

"Activate
   Auto
  pilot"

Uppercase and lowercase letters are considered distinct within node names. So, for example, "ActivateAutopilot" and "ActivateAutoPilot" are two different names, due to the uppercase versus lowercase "p" character.

2.2.1. Sizing New Nodes by Click and Drag

When the user specifies the placement of a new node with a single mouse click, the tool places a node of the current default diameter. If the user places a node with a click and drag operation, the position and size of the node are determined by the location of the initial click plus the length and direction of the drag.

2.2.2. Using Already-Defined Nodes

When entering the text for a node name, the user may type the name of a new node that has not yet been defined, or the name of an already-defined node. Sketch of the remainder:

• Allow the normal kind of emacs completion thing here.
• Explain how nodes can be defined by loading an fmsl spec.
• Describe how existing nodes have ins and outs displayed as little iconic ports, shown as ingoing and outgoing arrow heads (illustrated in sketch figure below).
• Probably provide some form of "show port names" command, that displays names of ports, probably in roll-over form; i.e., when "show port names" is on, moving the cursor over a port shows its (type) name; or maybe this'll happen when Elements->Edge is selected, and to see output ports otherwise, use Element->Properties. YES, I think I like the idea of port names appearing roll-over style during edge creation; see the edges section for details.

The idea I have right now is that the default appearance of port icons will be as follows:

• the icon is a small black right-pointing arrow
• the presumed normal positioning for output icons is on the right of nodes, inputs on the left; "presumed normal" means that the relative ordering of input ports is clockwise from 0 degrees and the relative ordering of outputs is counterclockwise from 0 degrees; that said, there is no restriction on the positioning of ports around the circumference of nodes, including intermixing of input and output ports; the relative ordering rules apply separately to input and output ports, as illustrated in Figure 6
  
  

  ---

  

  Figure 6: DRAFT -- Example to clarify port ordering (and other aspects of node display).

  ---

  
  
• an output icon is positioned such that the invisible line for its edge is a radius from the center of the circle (ellipse), making the base of the arrow head parallel to a chord on the edge of the circle, with the center point of the arrow head base exactly on the circumference of the circle (ellipse)
• inputs have the tip of the arrow head exactly on the circumference, and are rotated in an analogous way to output arrow heads
• a single input or output port is positioned on the horizontal diameter of the circle (ellipse)
• multiple outputs are positioned some number of degrees apart
• with an even number of ports, there is nothing on the horizontal axis, just equi-distance above and below it
• we need to come up with a reasonable way to deal with ops that have more ports than can be displayed for a given size node; some kind of ellipses-labeled drop-down menu, as either the only icon, or the bottommost icon seems like it should (could) work pretty well.
• we need to make sure that there's always enough space between ports to allow new ports to be created, between any two extant ports; AND, it would seem, that the ordinal position of the click to create a new port defines a new input or output in the relative position clicked on; e.g., if the user clicks between ports 3 and 4, then a new parameter is added at position 4; two ways to ensure ample between-port spacing are:
  • define (as a settable option) the minimum distance between two ports, and auto- enlarge the diameter of a node as necessary when new ports are added;
  • some form of overflow menu; open reflection, this seems to be an unworkable solution, since it makes it hard to draw all edges clearly
• to get down and dirty with the precision, it would seem the the degrees of separation have to be based on the fixed base width of the port arrow icons and the diameter of the node

Other port-related functionality:

• When `Options->Show Ports' is on, the edge arrowheads are off, and vice versa. Ports can be moved around the circumference of the circle; we may want an option to allow/disallow crossing lines

Nodes may not overlap. When the node cursor moves over any part of an existing node, it turns grey.

2.2.5. Resizing Nodes

Clicking anywhere within label causes cursor to become vertical text-edit bar. This can happen at any time, including during initial creation. Can change the name to something other than an existing node. Can change the name to that of an existing node, as long as it has the same signature of the node with the changing name. In the latter case, the completion list contains only the names of other same-signature nodes. (I think this is pretty cool, actually.)

2.2.8. Repositioning the Label