Overview of lsst.pex.config¶
The lsst.pex.config
module provides a configuration system for the LSST Science Pipelines.
Key concepts: configuration objects and field attributes¶
Configurations are hierarchical trees of parameters used to control the execution of code. They should not be confused with input data.
Configurations are stored in instances of a “config” object, which are subclasses of the Config
class.
Different configuration sets are associated with different subclasses of Config
.
For example, in the task framework each task is associated with a specific Config
subclass.
Configuration objects have fields that are discrete settings.
These fields are attributes on config classes.
Fields are themselves objects, usually an instance of the Field
class, or a subclass of Field
.
Fields have hooks for validation and documentation, and can even have their values recorded for provenance.
Field
and Field
subclasses are specialized for storing different types of configuration values, from simple integers to lists and mappings, and even other configuration objects or configurable objects.
See Types of configuration fields for more information.
Note
Fields are attributes of config classes, not instances.
When you access a field on a config instance, you don’t get the Field
instance itself, but rather an object that represents the value of that field, like an int
or a specialized container (like lsst.pex.config.List
).
Think of field attributes like property
attributes of classes, which implement custom getters and setters on otherwise simple attributes.
For more background on the architecture of fields, see the Notes section of the Field
class reference.
Example config class and usage¶
Here is a config class that includes five fields:
import lsst.pex.config as pexConfig
class IsrTaskConfig(pexConfig.Config):
doWrite = pexConfig.Field(
doc="Write output?",
dtype=bool,
default=True)
fwhm = pexConfig.Field(
doc="FWHM of PSF (arcsec)",
dtype=float,
default=1.0)
saturatedMaskName = pexConfig.Field(
doc="Name of the mask plane to use in saturation detection.",
dtype=str,
default="SAT")
flatScalingType = pexConfig.ChoiceField(
doc="The method for scaling the flat on the fly.",
dtype=str,
default='USER',
allowed={
"USER": "User defined scaling",
"MEAN": "Scale by the inverse of the mean",
"MEDIAN": "Scale by the inverse of the median",
})
keysToRemoveFromAssembledCcd = pexConfig.ListField(
doc="Fields to remove from the metadata of the assembled ccd.",
dtype=str,
default=[])
Fields are attributes on the config class.
The doWrite
, fwhm
, and saturatedMaskName
fields are of Field
type, which supports values with simple types like int
, float
, str
, and bool
.
The flatScalingType
field is a ChoiceField
type.
In this case, a user can only set flatScalingType
to a value of "USER"
, "MEAN"
, OR "MEDIAN"
.
In code, a config object might be used like this:
def doIsrTask(ccd, configOverrideFilename=None):
config = IsrTaskConfig()
if configOverrideFilename is not None:
config.load(configOverrideFilename)
# Note: config override files are Python code with .py extensions
# by convention
config.validate()
config.freeze()
detectSaturation(ccd, config.fwhm, config.saturatedMaskName)
# Note: methods typically do not need the entire config; they should be
# passed only relevant parameters
for k in config.keysToRemoveFromAssembledCcd:
ccd.metadata.remove(k)
if config.doWrite:
ccd.write()
Notice how configuration field values are accessible as attributes on the config
instance.
Also notice the load
method.
This is a way of loading configuration values from a file.
A configuration override file for IsrTaskConfig
might look like this:
config.doWrite = False
config.fwhm = 0.8
config.saturatedMaskName = 'SATUR'
config.flatScalingType = 'MEAN'
config.keysToRemoveFromAssembledCcd = ['AMPNAME']
This override file looks like Python code because it is.
The root
variable refers to the config instance that called its load
method, which is the config
variable in the doIsrTask
example.
In more advanced cases a configuration field’s value can itself be a config instance, so there will be a hierarchical namespace of configurations, like:
config.configField.fieldOnConfigField = 'value'
Principles for using lsst.pex.config¶
lsst.pex.config arose from a desire to have a configuration object holding key-value pairs that also allows for (arbitrarily simple or complex) validation of configuration values.
To configure code using lsst.pex.config, you create a subclass of the Config
class.
The subclass specifies the available Field
attributes, their default values (if any), and their validation, if necessary.
Config
configuration objects are hierarchical (see ConfigField
), so calling code can embed the configuration definitions of called code.
Configurations are not input data.
They should not be used in place of function or method arguments, nor are they intended to replace ordinary dictionary data structures.
A good rule of thumb is that if a particular parameter does not have a useful default, it is probably an input rather than a
configuration parameter.
Another rule of thumb is that configuration parameters should generally not be set in algorithmic code, only in initialization or user interface code.
In fact, changing configuration after a configurable object (such as Task
) has been initialized can lead to incorrect behavior.
You create a configuration object by instantiating the Config
subclass.
If any default Field
values need to be overridden, you can assign new values to the configuration object’s Field
attributes.
For example: config.param1 = 3.14
).
Often you can override defaults of either a config base class or a nested config in the Config.setDefaults
method, or by loading an external file with the Config.load
method.
Overrides should never be used to set already-existing default values
You code then uses the configuration values by accessing the object’s Field
attributes.
For example, x = config.param1
.
A Config
instance can also be frozen so that any attempt to change the values of field raises an exception.
This is useful to expose bugs that change configuration values after none should happen.
Finally, the contents of Config
objects may easily be dumped, for provenance or debugging purposes.
See Inspecting configurations for details.
Referencing Other Configuration Files¶
In some cases you would like to load other config files from a config file.
You can do this by using the __file__
special variable just like any other Python code.
This variable contains the path to this config file and therefore the enclosing directory can be located by using os.path
:
import os.path
configDir = os.path.dirname(__file__)
config.load(os.path.join(configDir, "otherconfig.py"))
Alternatively, if you wish to locate another configuration file using LSST infrastructure you can do the following:
import os.path
from lsst.utils import getPackageDir
config.load(os.path.join(getPackageDir("product_x"), "config", "otherconfig.py"))
Specialized Config subclasses¶
There exists a subclass of Config
which is designed to be configurable like a standard config, but have a runtime call interface.
These specialized subclasses are named ConfigurableAction
s, or actions for short.
These actions are not intended to replace other runtime components, but compliment them.
They provide configuration time mechanics to a simple runtime function.
This interface allows for both configuration of an action as well as making which action to run configurable.
These configurations are serialized out in a standard way, and thus allow complete functional states to be completely restored.
The selection (thus configuration) of which ConfigurableAction
s to run is made possible through the use of special Field
s named ConfigurableActionField
and ConfigurableActionStructField
.
See Types of configuration fields for more details and examples of both ConfigurableAction
s and the corresponding fields.