Another feature I implemented is proper multi-line input handling. In the UI listener, if you type the start of an incomplete expression, for example
: foo, pressing Return automatically inserts a newline. What the UI listener does is attempt to parse the input, and if it does not parse completely, it simply retains the contents of the input area, instead of passing it to the evaluator. This is very nice for entering multi-line expressions.
Factor 0.86 and previous versions already supported multi-line input in the UI listener but in a more awkward way; if you pressed Return with a partial expression entered, it would go to the evaluator and disappear from the input area, at which point you could still continue entering the remainder of the expression, but you could not edit what had already been parsed. You could also explicitly insert a newline with Shift+Return. I think the new approach is far more usable, especially together with the improved parse error handling as described below.
If you get a parse error, instead of simply dumping the parse error to the output area and forcing you to pick the previous input from the history, fix it, and try again, the listener now positions the caret at the location of the parse error and displays the parse error in a popup (which idiotically obscures the input field right now, but that's going to get fixed).
Finally, if there are any restarts, for instance offering to add a vocabulary to the vocabulary search path so that a word can be found, they are presented in a keyboard-navigatable list. No more trying to type
0 :res, mistyping it as
0 ;res, and losing your restarts, forcing a trip to the history, and another attempt at entering the input! Instead you use the Up/Down arrow keys and the Return key to pick the relevant restart.
I'd post some screenshots, but the appearance is not finalized and its hard to get a feel for the interactive behavior of the new listener in screenshots. I think when 0.87 is out, I'll try my hand at making a screencast instead.
I think the UI listener is now rather sophisticated, offering better usability than most other interactive evaluators I've seen. The big hole in the functionality right now is that you cannot select and copy text from the output area. This is not trivial to implement, since the output area actually consists of gadgets, one for every piece of styled text or presentation which was printed. So to be able to select and copy I would need to implement some kind of cross-gadget text selection protocol. However McCLIM already does something like this, its not rocket science, and it shouldn't be too hard to implement.