How to make a loop over tracts and patches to split the DC2 dataset

Support Data Preview 0
#1

Hello,

I would like to explore the whole DC2 dataset by using the data organisation according tracts and patches.
I mean I want to retrieve all the objects belonging to a tract and a patch from the dp01_dc2_catalogs.object table.
I am able to get from a query to the TAP service on dp01_dc2_catalogs.object table the list of tract numbers.

But then I would like to have access to the information associated to each tract and patch, ie its center, its corners (in radec). I have seen that the butler is aware of the skymap collection.

I am able to access from the butler to a skymap object

       repo = 's3://butler-us-central1-dp01'
       collection = '2.2i/runs/DP0.1'
       butler = dafButler.Butler(repo,collections=collection)
       skymap = butler.get('skyMap')

      <lsst.skymap.ringsSkyMap.RingsSkyMap object at 0x7f370d3d4f10>

     help(skymap)...

But it is not clear how to get the list of tracts and patches from skymap, because skymap methods ask for (ra,dec) argument.

I don’t want to provide the (ra,dec) argument because precisely I want to explore the tract-patch structure independently.

Is there an example of code doing this ?

You can think as if I want to draw a plot of the tract-patch structure in a 2D figure (in ra,dec) coordinates.

After this I would like to retrieve all the object in dp01_dc2_catalogs.object table within given patch, probably by providing the radec corners of the patch.
Thanks for your help.

Sylvie

Visualizing a mosaic of patches in coadd
#2

Hello Sylvie,

I use a query like this to grab all the tracts in the DC2 data set:

%%time
now0=datetime.now()

table_full_name = 'dp01_dc2_catalogs.object'

query = """SELECT DISTINCT tract FROM %s ORDER BY tract""" % (table_full_name)

results = service.search(query)

df = results.to_table().to_pandas()
tract_list = df['tract'].tolist()

now1=datetime.now()
print("Total time:", now1-now0)

print(tract_list)

Be warned, however, about downloading the whole DC2 tract-by-tract (or patch-by-patch). I’ve tried this with asynchronous TAP queries, and it still takes a long time (16+ hours). I’ve been told it is better – when possible – to do analyses within the database as much as possible (instead of downloading the data into a pandas dataframe and doing the analysis on the pandas dataframe).

2 Likes
#3

Hi @sylvielsstfr

This might not be quite what you are after but you could query and group by tract,patch and use aggregate functions.

from lsst.rsp import get_tap_service, retrieve_query
service = get_tap_service()
results = service.search("select tract,patch,min(ra) as minRA, max(ra) as maxRA, min(dec) as minDec, max(dec) as maxDec,count(*) as num \
                          from dp01_dc2_catalogs.object group by tract,patch")
results_tab = results.to_table()
import matplotlib
from matplotlib.patches import Polygon
from matplotlib.collections import PatchCollection
ll=results_tab['minRA','minDec'].as_array()
ul=results_tab['minRA','maxDec'].as_array()
ur=results_tab['maxRA','maxDec'].as_array()
lr=results_tab['maxRA','minDec'].as_array()
counts=results_tab['num']
fig, ax = plt.subplots()
patches = []
num_polygons = len(results_tab)
num_sides = 4
for i in range(num_polygons):
    corners = np.array([ ( ll[i][0], ll[i][1]), (ul[i][0], ul[i][1] ), (ur[i][0], ur[i][1]), (lr[i][0], lr[i][1])  ])
    polygon = Polygon(corners,True)
    patches.append(polygon)
p = PatchCollection(patches, cmap=matplotlib.cm.jet, alpha=0.4)
p.set_array(np.array(counts))
ax.add_collection(p)  
plt.xlim([47, 75])
plt.ylim([-47, -25])

jup
jup915×593 106 KB

It would not work well across the 0h boundary if there were data there.

Apologies if my python offends anyone :grinning:

Interested to know if the plot even looks right as I might have messed something up.

Mike

4 Likes
#4

Many thanks, that’s look very nice. We can see the patches within the tract. However it is obtained from the objects table, not by a pre-defined configuration of tracts and patches. I wonder what kind of information a skymap object includes about the definition of the geometry of the survey.
To explain : where is stored the static information allowing the DM-stack pipeline to decide such object belong to this tract and this patch. I am sure such kind of database exists.

#5

Thanks Douglas,
I am pretty sure that it is bad to download the whole objects of all the tracts. However if one ask for a restricted number of features like ra, dec, redshift for one tract, I don’t have to wait a lot.

#6

Hi @sylvielsstfr

Sorry I thought you were just after a way of seeing the tracts/patches structure.

How about


numTracts=0
patches = []
for tractID in tract_list:
    for tractInfo in skymap:
        if tractInfo.getId() == tractID:
           # print(tractID)
            for patch in tractInfo:
                numTracts+=1
                #print(patch)
                ibb=patch.getInnerBBox()
                tWCS=tractInfo.getWcs()
                corners = []
                for corner in ibb.getCorners():
                    p = geom.Point2D(corner.getX(), corner.getY())
                    coord = tWCS.pixelToSky(p)
                    corners.append([coord.getRa().asDegrees(), coord.getDec().asDegrees()])
                #print(corners)
                polygon = Polygon(corners,True)
                patches.append(polygon)
print(numTracts)
fig, ax = plt.subplots()
p = PatchCollection(patches, cmap=matplotlib.cm.jet, alpha=0.4)
ax.add_collection(p)  
plt.xlim([47, 77])
plt.ylim([-47, -25])

where tract_list is a list of tracts which as you say you can obtain by a querying Object (distinct tractID). For some reason this finds whole tracts as I think skymap is everything rather than what’s just in DP0? You could probably refine the plot by also querying for distinct tractID,patch and then filtering on patch as well as tract.

HTH

Mike

jup1
jup11089×691 106 KB

1 Like
#7

Thanks you Mike, that is exactly what I was looking for.
Best regards.

Sylvie