pdm/graphics bug?
Philip Massey wrote on May 17, 2006
Mike,
I've encountered something that might be a general bug in graphics, or maybe not. The specifics of what I'm doing is running the astutil task "pdm" to obtain a periodogram (significance vs frequency plot) for some light-curve data. Because the data extend over 3 years, in order for the periodogram to mean anything you have to use A LOT of points (high resolution) to not smear out phases.
So, if I run pdm with ntheta=200000 and hit the "i" key I get
the expected periodogram, with 200,000 points. But, I cannot expand a particular section using the
standard "E" key strokes. If I try expanding a section, what I get on the screen is actually way way to the left of what I expanded---often part of the graph where there are no data. So, I can't actually look at it in any detail.
I get similar (but not as extreme) behavior at 100,000, and as far as I can tell,
normal behavior at about 50,000.
If this is too obscure a description, I can give you a sample data file and
the specific parameters: retreive
http://www.lowell.edu/users/massey/R136mg0021crf
Then do an unlearn on pdm. Run it as
pdm R136mg0021crf fip+ minp=1 maxp=40 ntheta=200000
So far, so good? Now hit the "i" key to see a plot of theta vs frequency.
Looks great? Try expanding the region around the big dip at x=0.1 using
the "E" key. At least on my screen, I get a big blank. Hit capital A and you
can see that it THINKS it grabbed the right section---the plot axis values are
right---but there isn't any data. And there should be, which you can confirm
by just doing an "i" to get back the whole plot.
I'm running this in X11 on a G5.
---phil
I've encountered something that might be a general bug in graphics, or maybe not. The specifics of what I'm doing is running the astutil task "pdm" to obtain a periodogram (significance vs frequency plot) for some light-curve data. Because the data extend over 3 years, in order for the periodogram to mean anything you have to use A LOT of points (high resolution) to not smear out phases.
So, if I run pdm with ntheta=200000 and hit the "i" key I get
the expected periodogram, with 200,000 points. But, I cannot expand a particular section using the
standard "E" key strokes. If I try expanding a section, what I get on the screen is actually way way to the left of what I expanded---often part of the graph where there are no data. So, I can't actually look at it in any detail.
I get similar (but not as extreme) behavior at 100,000, and as far as I can tell,
normal behavior at about 50,000.
If this is too obscure a description, I can give you a sample data file and
the specific parameters: retreive
http://www.lowell.edu/users/massey/R136mg0021crf
Then do an unlearn on pdm. Run it as
pdm R136mg0021crf fip+ minp=1 maxp=40 ntheta=200000
So far, so good? Now hit the "i" key to see a plot of theta vs frequency.
Looks great? Try expanding the region around the big dip at x=0.1 using
the "E" key. At least on my screen, I get a big blank. Hit capital A and you
can see that it THINKS it grabbed the right section---the plot axis values are
right---but there isn't any data. And there should be, which you can confirm
by just doing an "i" to get back the whole plot.
I'm running this in X11 on a G5.
---phil
Mike Fitzpatrick wrote on May 17, 2006
Hi Phil,
The default graphics buffer apparently isn't large enough for a plot of that size and so some of the data are getting lost. Try
or change the value in your login.cl and the points will come back as expected.
Cheers,
-Mike
The default graphics buffer apparently isn't large enough for a plot of that size and so some of the data are getting lost. Try
cl> reset cmbuflen = 256000
cl> flpr 0 # zero, not oh
cl> gflushor change the value in your login.cl and the points will come back as expected.
Cheers,
-Mike
Last post on May 17, 2006