Export metadata as a Frictionless Table Schema.
export_metadata(
source,
destination=None,
format="frictionless",
**kwargs,
)
The export_metadata() function is the reverse of import_metadata(): it writes a MetadataImport out in a standard format. The only format currently supported is the Frictionless Table Schema, a JSON description of a table’s fields that many data tools can read. Because the metadata can come from any of the formats that import_metadata() reads, this is also a way to convert between them, for example to get a Table Schema from a CSVW file or from the variable definitions in a Define-XML document.
Each variable becomes a field, with its dtype mapped to a Frictionless type ("integer", "number", "string", "date" and so on) and its label and description used as the field’s title and description. The constraints required, unique, minimum, maximum, minLength, maxLength, pattern and enum are written for the variables that have them, and declared missing values become the field’s missingValues. Variables that are both required and unique make up the schema’s primaryKey, and the dataset label and description become the schema’s title and description.
Parameters
source: MetadataImport-
The metadata to export. To export one dataset from a MetadataPackage, select it first (as in package["DM"]).
destination: str | Path | None = None-
A file path to write the schema to, as indented JSON. Any missing parent directories are created. If None (the default), nothing is written. The schema is returned in both cases.
format: str = "frictionless"-
The output format. Only "frictionless" is supported ("table_schema" is accepted as another name for it).
**kwargs: Any-
Options for the exporter.
include_constraints=False leaves out the constraints of every field, giving a schema with only the field names, types, titles and missing values.
Returns
dict-
The Table Schema as a dictionary.
Raises
ValueError-
If the format isn’t supported.
Examples
Let’s import the bundled CSVW metadata for a table of weather observations and export it as a Frictionless Table Schema:
import json
import pointblank as pb
meta = pb.import_metadata(pb.load_metadata_example("weather_csvw.json"))
schema = pb.export_metadata(meta)
print(json.dumps(schema["fields"][:3], indent=2))
[
{
"name": "station_id",
"type": "string",
"title": "Station ID",
"constraints": {
"required": true,
"unique": true
}
},
{
"name": "timestamp",
"type": "datetime",
"title": "Observation Time",
"constraints": {
"required": true,
"unique": true
}
},
{
"name": "temperature_c",
"type": "number",
"title": "Temperature (Celsius)",
"constraints": {
"required": true,
"minimum": -50.0,
"maximum": 60.0
}
}
]
The first three fields show how the CSVW column definitions were carried over: the titles, the types (a CSVW decimal is a Frictionless number), and the constraints. The remaining keys of the schema describe the table as a whole:
print(schema["primaryKey"])
print(schema["title"])
['station_id', 'timestamp']
Weather Station Observations
The two required and unique columns form the primary key, and the title comes from the dataset label in the CSVW file.
Give a destination= to write the schema to a file. Here it’s written to a temporary directory and then imported again:
import tempfile
from pathlib import Path
with tempfile.TemporaryDirectory() as tmp:
path = Path(tmp) / "weather_schema.json"
pb.export_metadata(meta, destination=path, include_constraints=False)
reimported = pb.import_metadata(path)
print(reimported.variable_names)
print(reimported.get_variable("temperature_c").min_val)
['station_id', 'timestamp', 'temperature_c', 'humidity_pct', 'wind_speed_kmh', 'precipitation_mm', 'condition']
None
The file is a standalone Table Schema, so import_metadata() reads it back with the same variables. With include_constraints=False the range of temperature_c wasn’t written, so its min_val is now None. That’s useful when you want to share the structure of a table without the rules for its values.