Inferring Missing Values

Expressions

LinkML supports the use of expressions to define how slot values can be derived from other slot values. The slot equals_expression specifies how to calculate the value of a slot, as demonstrated here:

class:
  Person:
    attributes:
      age_in_years:
        range: decimal
        minimum_value: 0
        maximum_value: 999
        equals_expression: "{age_in_months} / 12"
      age_in_months:
        range: decimal
        equals_expression: "{age_in_years} * 12"
      is_juvenile:
        range: boolean
        equals_expression: "{age_in_years} < 18"

The following code will populate missing values:

from linkml_runtime.utils.inference_utils import infer_all_slot_values
from .personinfo infer Person

p = Person(age_in_years=30)
infer_all_slot_values(p, schemaview=sv)
assert p.age_in_months == 360
assert not p.juvenile

You can also use the linkml-convert script with the --infer flag

String serialization

For simple string based expressions, string_serialization can be used:

Address:
  attributes:
    street:
    city:
    full_address:
      string_serialization: |-
        {street}
        {city}

The following code will populate the full address:

from linkml_runtime.utils.inference_utils import infer_all_slot_values
from .personinfo infer Address

a = Address(street="1 Oak street", city="Oaktown")
infer_all_slot_values(p, schemaview=sv)
print(a.full_address)

Configuration

class linkml_runtime.utils.inference_utils.Config(use_string_serialization: bool = <factory>, parse_string_serialization: bool = <factory>, use_rules: bool = <factory>, use_expressions: bool = <factory>, resolve_function: ~typing.Callable[[str, ~typing.Any], ~typing.Any] = None)

Controls which inferences are performed

• slot.string_serialization

• slot.equals_expression

class linkml_runtime.utils.inference_utils.Policy(value, names=<not given>, *values, module=None, qualname=None, type=None, start=1, boundary=None)

Policy for when inferred values differ from already set values

Code

linkml_runtime.utils.inference_utils.infer_slot_value(obj: YAMLRoot, slot_name: str | SlotDefinitionName, schemaview: SchemaView, class_name: str | ClassDefinitionName = None, policy: Policy = Policy.STRICT, config: Config = Config(use_string_serialization=True, parse_string_serialization=False, use_rules=False, use_expressions=False, resolve_function=None))

Infer the value of a slot for an object

Parameters:

• obj – mutable object to be transformed

• slot_name –

• schemaview –

• class_name –

• policy –

• config –

linkml_runtime.utils.inference_utils.infer_all_slot_values(obj: YAMLRoot, schemaview: SchemaView, class_name: str | ClassDefinitionName = None, policy: Policy = Policy.STRICT, config: Config = Config(use_string_serialization=True, parse_string_serialization=False, use_rules=False, use_expressions=False, resolve_function=None))

Walks object tree inferring all slot values.

• if a slot has a string_serialization metaslot, apply this

• if a slot has a equals_expression metaslot, apply this. See eval_expr()

Parameters:

• obj –

• schemaview –

• class_name –

• policy – default is STRICT

• config –

Returns: