`MeasurementConfig`#

Describing a measurement stream by what it is, rather than by wiring it up. A stream is identified by its tracking file, its measurement type and its station; the type supplies the model class, the data block and the units, and the config supplies what every stream in a scenario shares.

`MeasurementRegistry`#

class MeasurementRegistry#

Bases: object

The table of measurement types a scenario can be built from.

Maps a short type name such as "SeqRange" onto the MeasurementType describing it, so a stream can be written as (file, meas_type, station) and everything else looked up.

Every method is a static method: there is one table per process, not one per instance, and get() and register() both work on it.

Methods

`all`()	The whole table, keyed by type name.
`assign_dataset_names`(streams[, registry])	Name every stream, keeping the names unique.
`get`(name)	Look up one measurement type.
`register`(name, spec)	Add a measurement type, or replace one of the same name.

See also

scarabaeus.MeasurementConfig: Settings shared across streams.
scarabaeus.MeasurementStream: One stream, named by type.
scarabaeus.MeasurementSpec: Consumes the registry to build a filter-ready spec.

Notes

The table is built on first use rather than at import, because building it means importing every model class and this module is imported during package initialisation.

Examples

import scarabaeus as scb

scb.MeasurementRegistry.register(
    "MyRange",
    scb.MeasurementType(MyRangeModel, "meas_sim", km, "Range",
                        real=False),
)
spec = scb.MeasurementSpec.build([(file, "MyRange", station)])

static all() → dict#

The whole table, keyed by type name.

Returns:

dict: The live table, not a copy: mutating it changes what later lookups see. Prefer register() for that.

static assign_dataset_names(streams, registry: dict = None) → list#

Name every stream, keeping the names unique.

Streams that share a station and a type would otherwise collide, which matters because solutions are keyed by dataset name. Repeats get a counter appended, added after the label so that the substrings Plotting classifies on stay intact.

Parameters:

streamssequence of MeasurementStream: The streams to name. A stream that already carries a name keeps it, and is still checked against the generated names for collisions.
registrydict, optional: Table to resolve each stream’s type against. Defaults to all().

Returns:

list of str: One name per stream, in the order given.

static get(name: str) → MeasurementType#

Look up one measurement type.

Parameters:

namestr: A registered type name.

Returns:

MeasurementType: The specification registered under name.

Raises:

KeyError: If name is not registered. The message lists the names that are, since a typo here is otherwise reported much later as a missing model.

static register(name: str, spec: MeasurementType) → None#

Add a measurement type, or replace one of the same name.

Lets a new measurement type be used by build() without editing this module.

Parameters:

namestr: Short name streams will refer to the type by.
specMeasurementType: What the name implies: model class, data block, units, label.

`MeasurementType`#

class MeasurementType(model_class: Any, meas_name: str, units: Any, label: str, real: bool = True)#

Bases: object

What a measurement type implies.

Attributes:

model_classtype: Measurement model to build.
meas_namestr: Block within the tracking json holding this measurement.
unitsUnits: Units the observations are read in. Keeping these with the type is what stops a sigma being built in one unit and the observations in another, which is easy to do when both are written out by hand.
labelstr: Human-readable suffix used when naming datasets. Must keep the substrings Plotting classifies on (“Range Rate” before “Range”).
realbool: True for real tracking data, False for simulated.

`MeasurementConfig`#

class MeasurementConfig(media_folder: str | None = None, tropo_seasonal_file_path: str | None = None, sc_transponder_delay_RU: float = 0.0, frame: Any = None, sigma: dict = <factory>, bias: dict = <factory>, extra: dict = <factory>)#

Bases: object

Settings shared by every stream in a scenario.

Anything here can also be overridden per stream. Extra keys can be carried in extra so this can grow without changing the signature.

Attributes:

media_folderstr, optional

Folder holding the tropospheric and ionospheric correction files.

tropo_seasonal_file_pathstr, optional

Seasonal troposphere file.

sc_transponder_delay_RUfloat

Spacecraft transponder delay in range units (RU), applied to the real radiometric reads (DopplerReal, SequentialRangingReal). Defaults to 0.0, which takes the delay each tracking block carries, or no delay when it has none.

A plain float on purpose, not an ArrayWUnits: a range unit is not a fixed unit of time. One RU lasts 1 / (range_unit * reference_frequency) seconds, which depends on the band and on the exciter scalars of each pass, so it is only known once a pass is read and is not something Units can convert. The reader converts it per pass with transponder_delay_seconds(). RU is also how missions quote the delay.

frameFrame, optional

Frame the observations are expressed in. None leaves each model on its own default, which is J2000. Set it and build() passes it to every stream’s read.

sigmadict, optional

Default measurement sigma per type name or per dataset name, used for any stream that does not carry its own.

A value is needed for every stream, including real data. The sigma refit during outlier rejection runs on the residuals, which come from the dataset, which cannot be built without a sigma to weight them – so the refit revises an initial value rather than standing in for one. A stream left without one reports what is missing (see _require_sigma()).

biasdict, optional

Measurement bias per type name or dataset name. Defaults to 0.

extradict

Anything further, offered to every model constructor.

Offered, not passed: build() filters each model’s keyword arguments against its signature, so a key that only one model accepts reaches only that model instead of raising TypeError on the others. The cost is that a key no model accepts – a typo, or one meant for a different scenario – would simply do nothing, so build warns about any key it could not give to at least one model.

Methods

`get_bias`(type_name[, dataset_name])	Return the bias of a stream.
`get_sigma`(type_name[, dataset_name])	Return the sigma of a stream.

`MeasurementStream`#

class MeasurementStream(file: str, meas_type: str, station: Any, name: str | None = None, sigma: Any = None, bias: Any = None, media: Any = None)#

Bases: object

One measurement stream.

Normally written as a plain (file, meas_type, station) tuple and normalised into this; use the class directly when a stream needs something the defaults do not cover.

Methods

`coerce`(item)	Return `item` as a MeasurementStream.
`default_name`(type_spec)	Return the default dataset name of this stream.

MeasurementConfig#

MeasurementRegistry#

MeasurementType#

MeasurementConfig#

MeasurementStream#

`MeasurementConfig`#

`MeasurementRegistry`#

`MeasurementType`#

`MeasurementConfig`#

`MeasurementStream`#