Re: [Rosegarden-devel] Latest ruler commits

[Chris F. <chr...@go...>]

On Sat, 2009-09-12 at 19:46 -0400, D. Michael McIntyre wrote:
> On the horizontal axis, it all seems a bit more dubious ...
> but if you shove them to the wall and 
> leave them there, it's eminently possible to wind up with a big stack of 
> different values at exactly the same time.  This is a pretty meaningless 
> situation from a MIDI controller perspective, and it might be cleaner to just 
> clip at these boundaries, and throw away any data points that get released on 
> the wrong side.
> 
> I think clipping would probably be better, but this whole observation is a 
> quibble mixed with some thoughtful speculation, and not a demand at all.
I agree. Deleting points that are dragged to the boundary and left there
sounds a good way of doing it. My only reservation is that you may want to
drag a single even to the start of a segment. On that basis, a compromise might
be to delete all but one of a selection if it's dragged to the edge of a
segment. Slightly counter-intuitive but workable.

> > * Mimic MatrixTool return of FollowMode in ControlTool classes
> 
> Oh really, and did you do that with or without extra splorzleblats?  (No idea 
> what you're talking about on that one.)
Fairly irrelevant code comment, especially if you didn't write the Matrix code.
Just ignore it.

> > * Quick value readout tryout
> First, if I know I'm trying to insert a value of 1492 for some reason, it's 
> pretty much impossible to do this in one step unless I just randomly get 
> lucky.  First I have to click something into the big unlabeled blank empty 
> space, and then I get to see what number I picked.  Oh.  8157.  No cigar.  
> Then I can drag it to where I was trying to aim in the first place.
I'm thinking of drawing a scale at either end of the ruler in light grey so it's 
unobtrusive but at least tells the user what they're drawing. A hint
written by the + cursor for the pencil tool would be pretty easy as
well. I'll look into that.

> Second, this works well for data points that have a wide horizontal 
> separation, but it's extremely easy to glob a bunch of these value labels 
> together into an indecipherable blue mess when they're closely spaced.
That needs fixing. Overlapped text looks terrible.

> > * No scaling to (eg) difference from default yet
> In the interim, however, it seems like representing 8191 as 0 and showing the 
> displayed values as if they were a signed int would be more intuitive, and 
> consistent with how I vaguely half remember doing this in the past.  Don't 
> sequencers normally present "no pitch bend" as 0 and show + or - numbers from 
> there?
I think I'll code a custom translation for each different control.
Shouldn't be a huge job.

Chris