Chris, thanks for responding. I appreciate the suggestion of different plotting
method. I don't yet know what the advantage is, but I'll try it. Maybe my head
will hurt less too.
See **** comments below.
Sharon
On Apr 27, 2:59pm, Chris Pelkie wrote:
> Subject: Re: [opendx-users] validity of points in x/y plots
> Here's what I think should happen. Reality may differ.
>
> First, some basic review of expectations.
>
> Field P1 has "data"(1), "positions"(1), "invalid positions"(1)
> Field P2 has "data"(2), "positions"(2), "invalid positions"(2)
>
>
> You Replace "data"(1) into P2 as "positions" yielding:
> Field P3 with "data"(2), "positions"(data1), "invalid positions"(2)
>
> Plot sees only "invalid positions"(2) as invalid. Therefore the points that
> are not plotted correspond to those that were invalid in P2 originally.
> Note that Replace only carries over the named component (it does not drag
> along the "invalid positions" for free). Note also that "invalid positions"
> knows or cares nothing about "data": it only masks off designated positions
> by their index. If you slipped in some different numbers behind its back
> (as we did above), it doesn't know or care about that either. Still masks
> the same positions by index. (I don't mean by ref, but because our invalids
> are position-dep, they are implicitly associated in linear order with the
> corresponding list of positions.)
>
> That's not what you wanted, so let's go to Experiment 2.
****
Right, but what is your result if you plot the field at this step. Does the
range of your axes cover the out of range value (-9, e+38, or whatever) ? Mine
doesn't, which make me wonder why these values or not plotted. Since the new
field contains only the invalids carried over from P2 (but not P1), it seems to
me that the "former" P1 invalids should show up in the plot. Maybe your
Glyph/autoaxes method as mentioned below works differently than Plot in this
regard. Or, maybe my invalid value is just too darn high (e+38), that Plot
ignores it, regardless of its invalidity-tagged status. Dunno. I suppose that I
should forget about it since it's all inconsequential when I follow through
with the extra steps (experiment #2), but it's bothersome to me.
> Extract(P1, "invalid positions"(1)) and Extract(P2, "invalid positions(2)")
> (this could be done with Marks and Unmarks as well)
>
> Next, we combine the invalids in Compute:
> byte("invalid positions"(1)) || byte("invalid positions"(2))
>
> which yields an array we'll call "invalid positions"(1or2).
>
> Replace "invalid positions"(1or2) into P2 as "invalid positions" to yield:
> Field P3 with "data"(2), "positions"(2), "invalid positions"(1or2)
>
> Now, Replace "data"(1) into P3 as "positions" to yield:
> Field P4 with "data"(2), "positions"(data1), "invalid positions"(1or2)
>
> Then you Plot P4.
>
> My simple tests all work as expected. Any point for which there is an
> invalid entry in "invalid positions"(1or2) does not appear in the Plot.
>
>
> BTW, normally I steer clear of Plot cause it makes my head hurt. I'd
> probably do it like this:
> Do the same thing with combining and replacing the "invalid
> positions"(1or2) to make Field P3.
> Then, feed P3 and P1 to Compute: ["data"(2), "data"(1)]
> Unmark("positions")
>
> This yields:
> Field P4 with "data"(2), "positions"([data1,data2]), "invalid
positions"(1or2)
> then Glyph to get "data"(2) glyphs plotted in [data2, data1] space. Turn on
> AutoAxes if you want to see numbers on the axes. By feeding Compute P3 on
> the left, P4 inherits the structure of P3 including the "invalid
> positions"(1or2).
>
> This is obviously extensible to 3D unlike Plot.
>
> Chris Pelkie
> Vice President/Scientific Visualization Producer
> Conceptual Reality Presentations, Inc.
> 30 West Meadow Drive
> Ithaca, NY 14850
> [EMAIL PROTECTED]
>
>-- End of excerpt from Chris Pelkie
