>>I agree, but question is what should be displayed<<
The point has to be on one trace or the other (logically). The display should show the units associated with that trace. There should be no mechanism that allows the specification of a point as the intersection of two traces with different units.
Note that the use of traces with different units is implicitly usinf multiple y axes (whether that is explicitly provided for or not, it should be). The y value for any point in the graph (no matter how chosen or calculated) has to be relative to one of the y axes. You need to be able to make that relationship explicitly. __________________ � � � � Tom Gutman
For some reason I can't see any of the images in your message.
I'm trying to decide why I would want to see units in a marker on a plot? I would typically want to see the units in the title (whether I type it myself or PRIME inserts it) so I don't necessarily need to see it on a marker.
Could be I'm missing part of the story without seeing your examples.
If you're proposing that the user can specify the unit used for each axis for each trace (that's what we do now by "dividing out" the units when plotting), then you could just display that user-specified unit in the TRACE view.
When you were talking about "markers" I was hoping you meant the user could specifiy an x- or y-coordinate and have a special marker placed on the graph with the corresponding trace values. For example, if I'm always interested in the value of Y given X=1, then designating a marker at X=1 would show the corresponding value of Y, to be updated and re-displayed as the calculations change. That would be nice for doing presentations or final engineering calculations.
Thinking more about the display units ... I think it should default to whatever the user specified on the graph, but give the user the option to change the display unit just like we can on evaluated expressions. Let me fill in that "black box" with whatever display unit I desire, and scale the value (and any "leftover units" accordingly. Could be pi, meters, pounds, etc. :+)
Your first example really shows why we need to have quantity/units as an option, not just quantity (units). An axis label of X(kg) is common enough, and although a lot of people / books / journals will not accept it (some of my professors at university would not even accept it!) it can at least be argued that's it's a convention. I do not ever recall seeing X(PI) though. That would always be written as X/PI. X(PI) is confusing and looks terrible.
On 2/19/2010 5:40:58 PM, Mona Zeftel wrote: >Here are the images. They >didn't come through in the >email.
The real (underlying) question is how did the user get to that position?
If we are using the trace option, then we have a disambiguation problem of either: how to select curve 1 from curve 2, or, whether the trace option should list all curve values (which then begs the question "in which order - nearest first?")
If we posit the case that the user was actually clicking not at the intersection point, but nearby, and had then navigated to that point (visual inspection being the user choice), then clearly the single reported value would be the value, with units, of the displayed variable.
Clarity of how unit scaling is applied is still a potential issue. In current practice users are encouraged to evaluate say (height/ft), which becomes unitless, but actually would want the marker to show the value to show in ft. - an awkward transition issue for old sheets.
An issue I have is when I am displaying real data with transients (e.g. 50,000 points) and want to positively locate the displayed transient. Clearly I can't click on the exact point, even on a 5,000 pixel screen even if I could get one, so there is a user interaction issue while searching for that elusive 'peak' of the transient.
While doing the trace, I'd like some form of automatic zoom/magnification (just like some disability support magnifiers) so that I can see the fine detail of where I want to place my marker.
Summary. The original question needed clarification. Curve selection in dense regions is an issue. Transition from old to new graph unit practices will be an issue point selection on dense curves is an issue.
>In Mathcad Prime 1.0, you can >have units on plots, so the >question arose >about showing units on the >markers. We want to show them >but what if two >or more traces have different >unit scaling? This would be a >conflict if >you have a marker where two or >more traces intersect.
I think we need to back up a bit here and establish what it is one actually plots on a graph. The ONLY thing you can plot on a graph is a number. You cannot plot, for example, 5 tonnes as a point. That is why the quantity must always be divided by an appropriate unit before plotting. I can plot 5*tonnes/tonnes, which shows as a value of 5, or 5*tonnes/kg, which shows as a value of 5000, but I can't plot 5*tonnes or 5000*kg. Incidentally, that's why the strictly correct way of labeling an axis is quantity/units, not quantity (units), and you should allow people to use this form of labeling.
>What is the point below for a >marker, 4.9? 4.9 kg or 4.9 m?
The point is at 4.9. Like anything else on a graph, a point can only have purely numerical coordinates. Your apparent conflict occurs because, as Tom has already pointed out, you are asking a nonsensical question to begin with. You might as well ask "how many 1kg masses do I need to make a 3 meter length?". I understand what you want, but the reason you will never be able to come up with a satisfactory way to do it is because there is no way to do it.
It appears to me that your graphics showed points of intersection between two functions by coincidence only. Is the intent to allow the user to "Trace" along the x-axis and return values for each of the graphed functions? If that's the case then I hope the user also has the ability to explicitly key in values for x- coorindates. (This would enhance the current TRACE feature too.)
On 2/20/2010 9:56:12 PM, Mona Zeftel wrote: >Richard and Tom, > >I'll raise this in Monday's >discussion of the issue. It >seemed to us >that it would be useful to >include the units in the point >marker. > >Mona
I'll confess to actually being a little confused about what you proposing. You talk about markers, but in a way that implies they show or return some value. Are they markers in the sense we have now, or something closer to the current trace dialog?
If you really mean markers, then the user enters the value (or values, for an x-y marker). If that's the case then the only thing they can enter is a number or an expression that evaluates to a number, exactly the same as now. It is not possible to plot 4.9kg as a vertical line. You can't plot quantities on graphs, only numbers.
If you mean something like the trace dialog, where the user designates a point and gets the coordinates then you could do something with units. It should only return units if you have selected "Track data points" though, so that it is locked to one specific curve. If you turn that off then it should return only numbers. If you do implement some sort of trace with units then please add an option so we can turn the units off!