Running the AP pipeline

Setup

Pick up where you left off in Getting Started. This means you already have a repository of ingested DECam data and have setup the LSST Science Pipelines stack as well as ap_pipe and ap_association.

Your directory structure should look something like

repo/   <-- where you will be running ap_pipe
   registry.sqlite3   <--- created during image ingestion
   image1.fits   <--- linked here by default during image ingestion
   image2.fits
   ...
   ref_cats/   <--- copied or linked here manually by you
      gaia/
         shard1
         shard2
         ...
      pan-starrs/
         shard1
         shard2
         ...
   calibs/
      calibRegistry.sqlite3   <--- created/updated during calib/defect ingestion
      cpBIAS/
         bias1.fits   <--- linked here by default during calib ingestion
         bias2.fits
         ...
      cpFLAT/
         flat1.fits   <--- linked here by default during calib ingestion
         flat2.fits
         ...
      defects/   <--- put here manually by you prior to defect ingestion
         defect1.fits
         defect2.fits
         ...
   templates/   <--- copied or linked here manually by you
      repositoryCfg.yaml
      deepCoadd/
         g/
            0/
               psfMatched-0,0.fits
               psfMatched-0,1.fits
               ...

AP pipeline on the command line

The executable to run for the AP Pipeline (ApPipeTask) is in ap_pipe/bin/ap_pipe.py.

To process your ingested data, run

mkdir apdb/
make_apdb.py -c apdb.isolation_level=READ_UNCOMMITTED -c apdb.db_url="sqlite:///apdb/association.db"
ap_pipe.py repo --calib repo/calibs --rerun processed -c apdb.isolation_level=READ_UNCOMMITTED -c apdb.db_url="sqlite:///apdb/association.db" --id visit=123456 ccdnum=42 filter=g --template templates

In this case, a processed directory will be created within repo/rerun and the results will be written there. See Setting up the Alert Production Database for ap_pipe for more information on make_apdb.py.

This example command only processes observations that have a dataId corresponding to visit 123456 and ccdnum 42 in with a filter called g.

lsst.ap.pipe supports dataId parsing, e.g., ccdnum=3^6..12 will process ccdnums 3, 6, 7, 8, 9, 10, 11, and 12.

Note

Until a resolution for DM-12672 is found, you should include a filter in the dataId string for ap_pipe to run successfully.

If you prefer to have a standalone output repository, you may instead run

ap_pipe.py repo --calib repo/calibs --output path/to/put/processed/data/in -c apdb.isolation_level=READ_UNCOMMITTED -c apdb.db_url="sqlite:///apdb/association.db" --id visit=123456 ccdnum=42 filter=g --template path/to/templates

In this case, the output directory will be created if it does not already exist. If you omit the --template flag, ap_pipe will assume the templates are somewhere in repo.

Note

If you are using the default (SQLite) association database, you must configure the database location, or ap_pipe will not run. The location is a path to a new or existing database file to be used for source associations (including associations with previously known objects, if the database already exists). In the examples above, it is configured with the -c option, but a personal config file may be more convenient if you intend to run ap_pipe many times.

Expected outputs

If you used the rerun option above, most of the output from ap_pipe should be written out in the repo/rerun/processed directory,. The exception is the source association database, which will be written to the location you configure. The result from running ap_pipe should look something like

apdb/
   association.db   <--- the Alert Production Database with DIAObjects
repo/
   rerun/
      processed/
         repositoryCfg.yaml
         deepDiff/
            v123456/   <--- difference images and DIASource tables are in here
         123456/   <--- all other processed data products are in here (calexps etc.)

This is one example, and your rerun or output directory structure may differ. Of course, to inspect this data with the Butler, you don’t need to know where it lives on disk. You should instead instantiate a Butler within python in the processed directory and access the data products that way.

For example, in python

import lsst.daf.persistence as dafPersist
butler = dafPersist.Butler('repo/rerun/processed')
dataId = {'visit': 123456, 'ccdnum': 42, 'filter': 'g'}
calexp = butler.get('calexp', dataId=dataId)
diffim = butler.get('deepDiff_differenceExp', dataId=dataId)
diaSourceTable = butler.get('deepDiff_diaSrc', dataId=dataId)

Calexp template mode

By default, ap_pipe assumes you would like to use PSF-matched coadd images as templates for difference imaging. However, the pipeline also supports using calibrated exposures (calexps) as templates instead. A configuration file config/calexpTemplates.py is included witha ap_pipe to enable this.

To use ap_pipe in calexp template mode, point to the config file with the --configfile (-C) flag and additionally specify the dataId of the template with the --templateId flag, e.g.,

-C $AP_PIPE_DIR/config/calexpTemplates.py --templateId visit=234567

Be sure to also specify the location of the repo containing the calexp templates with the --template flag if they are not in the main repo. A full command looks like

ap_pipe.py repo --calib repo/calibs --rerun processed -C $AP_PIPE_DIR/config/calexpTemplates.py -c apdb.isolation_level=READ_UNCOMMITTED -c apdb.db_url="sqlite:///apdb/association.db" --id visit=123456 ccdnum=42 filter=g --template /path/to/calexp/templates --templateId visit=234567

Supplemental information

Previewing dataIds

So far, we have implicitly assumed that you know reasonable values to choose for the dataId values (i.e., visit, ccdnum, and filter for DECam). While it is your responsibility to ensure the data you want to process and your templates do indeed overlap with each other, ap_pipe supports the --show data flag.

To get a list of all the dataIds available in repo in lieu of actually running ap_pipe, try

ap_pipe.py repo --calib repo/calibs --rerun processed --id visit=123456 ccdnum=42 filter=g --show data

Running on other cameras

Running ap_pipe on cameras other than DECam works much the same way: you need to provide a raw repo and either a rerun or an output repo, and you may need to provide calib or template repos. The calexp configuration file will work with any camera.

You will need to use a dataId formatted appropriately for the camera; check the camera’s obs package documentation or consult the –show data flag.

Common errors

  • ‘No locations for get’: This means you are trying to access a data product which the Butler cannot find. It is common to encounter this if you do not have all of the calibration products in the right spot or a template image cannot be accessed.

Interpreting the results

Warning

The format of the ap_association Alert Production Database is rapidly evolving. For the latest information on how to interface with it, see lsst.ap.association.

Try these python commands to make some initial plots of your newly processed data. You can also use the Butler to display calibrated exposures, difference images, inspect DIAObjects and/or DIASources, etc.

import os
from copy import deepcopy
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import sqlite3
import lsst.daf.persistence as dafPersist

workingDir = 'repo/rerun/processed'
butler = dafPersist.Butler(os.path.join(workingDir))

# Open and read all data from the association database
sqliteFile = os.path.join('apdb', 'association.db')
connection = sqlite3.connect(sqliteFile)
tables = {'obj': 'dia_objects', 'src': 'dia_sources', 'con': 'dia_objects_to_dia_sources'}
conTable = pd.read_sql_query('select * from {0};'.format(tables['con']), connection)
objTable = pd.read_sql_query('select * from {0};'.format(tables['obj']), connection)
srcTable = pd.read_sql_query('select * from {0};'.format(tables['src']), connection)
connection.close()

# Plot how many sourceIDs are attached to any given objectID
obj_id = objTable['id'].values  # object ids from the objTable
con_obj_id = conTable['obj_id'].values  # object ids from the conTable
con_obj_id.sort()
lowerIndex = np.searchsorted(con_obj_id, obj_id, side='left')
upperIndex = np.searchsorted(con_obj_id, obj_id, side='right')
count = upperIndex - lowerIndex
plt.hist(count, bins=50)
plt.yscale('log')
plt.xlabel('Number of DIASources per DIAObject')
plt.ylabel('DIAObject Count')
plt.show()

# Plot all the DIAObjects on the sky
plt.hexbin(objTable['coord_ra'], objTable['coord_dec'],
                cmap='cubehelix', bins='log', gridsize=500, mincnt=1)
plt.title('DIA Objects', loc='right')
plt.xlabel('RA')
plt.ylabel('Dec')
plt.show()