Using Jobs to collect and serialize measurements

The lsst.validate.base framework works in tandem with the SQUASH metric monitoring service. This page describes how to use the lsst.validate.base.Job class to collect measurements and blobs, and build a JSON document that is accepted by SQUASH’s RESTful API.

Collecting measurements in a Job

A Job is a container for measurements and their referenced data, including blobs.

You can add multiple measurements to a Job when you create it:

from lsst.validate.base import Job

# ...create measurements meas1, meas2, ...

job = Job(measurements=[meas1, meas2])

Alternatively, you can add measurements one at a time with the Job.register_measurement method:

from lsst.validate.base import Job

# ...create measurements meas1, meas2, ...

job = Job()

A convenient way of adding measurements to a Job is in a measurement’s __init__ method, like this:

import os
import astropy.units as u
from lsst.utils import getPackageDir
from lsst.validate.base import MeasurementBase, Job

class PA1Measurement(MeasurementBase):

    def __init__(self, yaml_path, job=None):

        self.metric = Metric.from_yaml('PA1', yaml_path=yaml_path)

        # measurement code
        # ...

        # This is the scalar measurement value; always as an astropy.unit.Quantity
        self.quantity = 2. * u.mmag

        if job is not None:

job = Job()
yaml_path = os.path.join(getPackageDir('validate_drp'),
PA1Measurement(yaml_path, job=job)

Getting measurements from a Job

Not only can you put measurements into a Job, you can also get measurements out of a Job. Your application might use this functionality to get measurements to pass as parameters of other measurements, or to provide a pass/fail printout of all measured metrics.

If a job contains a measurement of the 'PA1' metric, you can get that measurement using the get_measurement method:


Some measurements correspond to a particular specification level or filter. You can pass this information to get_measurement to disambiguate several measurements of a given metric:

pa2_design = job.get_measurement('PA2', spec_name='design')
pa2_min = job.get_measurement('PA2', spec_name='minimum')

A RuntimeError is raised if Job.get_measurement does not have sufficient information (like spec_name or filter_name) to retrieve a single measurement.

Serializing to JSON

JSON objects

Once all measurements and blobs are registered in a Job, you can generate a JSON serialization of that dataset:

from lsst.validate.base import Job

job = Job()
# .. register measurements and blobs
json_doc = job.json

json_doc is a dict wrapping json-serializable objects.


All lsst.validate.base classes (Datum, MeasurementBase, BlobBase, Metric, Specification and Job) have a json property. That property lets you get a json-serializable object for a specific object. Job.json simply calls the json properties of every object it contains.

Writing a JSON file

Job also provides a convenience method, write_json, for writing the JSON document directly to the filesystem:

from lsst.validate.base import Job

job = Job()
# .. register measurements and blobs

Uploading lsst.validate.base’s JSON to SQUASH

At the moment the SQUASH API does not directly accept the JSON produced Job.json. Instead, it’s shimmed with a package called post-qa. We expect to standardize the process for uploading measurements to SQUASH in the near future. In the meantime, please contact the DM SQuaRE team on or the #dm-square Slack channel and we’ll help integrate your package with SQUASH.