Making Randoms within Mask

I spent the weekend trying to make random catalogs inside the masks. They are taking forever to run, so I probably need to optimize better. I didn't spend too much time doing this before because I figured we would only need to make the random catalogs once and then we could just use them over and over again.

However, I've been running the code to generate them for > 13 hours and they still haven't finished, so there is a problem somewhere. The issue is that the mask represents a small region of the sky, but I am generating random objects over the whole sky, so 99.99% of the objects are rejected and thus the code is very inefficient.

Update: I found a bug in this code while trying to optimize explains why it was taking so long to run. Here is some testing about how it scales with number of randoms generated:

photo randoms:
500 randoms
Mon Dec 13 22:28:44 2010
Mon Dec 13 22:29:13 2010
29 seconds

5000 randoms
Mon Dec 13 22:34:00 2010
Mon Dec 13 22:31:02 2010
~3 minutes

50,000 randoms
Mon Dec 13 22:36:04 2010
Mon Dec 13 23:02:00 2010
~28 minutes

spectro randoms:
500 randoms
Mon Dec 13 22:32:10 2010
Mon Dec 13 22:31:45 2010
25 seconds

5000 randoms
Mon Dec 13 22:37:06 2010
Mon Dec 13 22:34:38 2010
~2.5 minutes

50,000 randoms
Mon Dec 13 22:38:51 2010
Mon Dec 13 23:01:42 2010
~23 minutes

The way the code works in it's current (inefficient) form:

For Spectroscopic Data:
1) Read in the two masks.
2) Create random ra and dec that are 5X as big as the spectro data set
3) Assign a redshift from the redshifts of the spectro data set (to keep same redshift distribution)
4) See if ra/dec is in the both masks
5) If it is in both masks, look at the weight in the BOSS mask. Assign a random weight to the object (between 0 and 1), if it is less than or equal to the weight of the BOSS mask then it is "observed", if it is greater than the BOSS weight it "isn't observed"
6) Repeat above until you reached the right number of randoms

For Photometric Data:
1) Read in the two masks.
2) Create random ra and dec that are 5X as big as the photo data set
3) See if ra/dec is in the both masks
4) Repeat above until you reached the right number of randoms

We don't need to assign a redshift to the photo data or "observe or not observe" because the photometric data doesn't have weights.

The code is here:

../pcode/photorandoms.pro
../pcode/spectrorandoms.pro

The code to run this as a qsub is here:
../pcode/makeRphoto.script
../pcode/makeRspec.script

You run it with the following code:
qsub makeR[photo,spec].script

The randoms have the same distribution as the data, shown in the following histograms, white is the data, red is the randoms:



Sweet!