lamindb.base.dtypes

Data types.

Most data types in LaminDB are string-serializable representations of standard Python types, commonly used for validating data in libraries such as pydantic. Categorical data types are absent from built-in Python types and are needed to anchor the validation of a categorical variable in a LaminDB registry.

Simple

In the table below, the first column shows the object that can be passed to the dtype argument of Feature() or Schema() and the second the string serialization that’s used in the database.

dtype	string serialization	pandas
int	"int"	int64 | int32 | int16 | int8 | uint | ...
float	"float"	float64 | float32 | float16 | float8 | ...
str	"str"	object
bool	"bool"	boolean | bool
datetime	"datetime"	datetime
"datetime64[ns, UTC]"	"datetime64[ns, UTC]"	datetime64[ns, UTC]
date	"date"	object (pandera requires an ISO-format string, convert with df["date"] = df["date"].dt.date)
dict	"dict"	object
"num"	"num"	int | float (“num” is a convenience type for int | float)
"path"	"path"	str (pandas does not have a dedicated path type, validated as str)
"url"	"url"	str (pandas does not have a dedicated url type, validated as str)

Categorical/ relational

For any categorical, you can restrict permissible values to the values defined in a registry. This establishes a relationship.

dtype	string serialization
ln.ULabel	"cat[ULabel]"
bt.CellType	"cat[bionty.CellType]"
bt.Disease	"cat[bionty.Disease]"
ln.Artifact	"cat[Artifact]"

You can restrict permissible values to instances of ULabel or Record types, i.e., to dynamic registries.

dtype	string serialization
ulabel_type (a ULabel with is_type=True)	"cat[ULabel[<uid_of_ulabel_type>]]"
record_type (a Record with is_type=True)	"cat[Record[<uid_of_record_type>]]"

You can restrict permissible values by filtering the categorical on fields of its registry.

dtype	cat_filters	string serialization
bt.Disease	{"source": source}	"cat[bionty.Disease[source__uid='<uid_of_source>']]"
ln.Artifact	{"schema": schema}	"cat[Artifact[schema__uid='<uid_of_schema>']]"

Lists

dtype	string serialization
list[bt.CellType]	"list[cat[bionty.CellType]]"
list[float]	"list[float]"

Unions

Unions are currently only supported for static registries.

dtype	string serialization
[bt.Tissue.ontology_id, bt.CellType.ontology_id]	"cat[bionty.Tissue.ontology_id|bionty.CellType.ontology_id]"

Usage

In features, you can pass dtypes to the dtype argument. See Feature for more details.

In function signatures, you can use dtypes to type annotate and validate arguments. See Track notebooks, scripts & workflows or flow() for examples.