pycldf.dataset

The core object of the API, bundling most access to CLDF data, is the Dataset . In the following we’ll describe its attributes and methods, bundled into thematic groups.

Dataset initialization

class pycldf.dataset.Dataset(tablegroup)[source]

API to access a CLDF dataset.

Parameters:: tablegroup (csvw.metadata.TableGroup) –

__init__(tablegroup)[source]

A Dataset is initialized passing a TableGroup. The following factory methods obviate the need to instantiate such a TableGroup instance yourself:

in_dir()
from_metadata()
from_data()

Parameters:: tablegroup (csvw.metadata.TableGroup) –

classmethod from_data(fname)[source]

Initialize a Dataset from a single CLDF data file.

See https://github.com/cldf/cldf#metadata-free-conformance

Return type:: Dataset
Parameters:: fname (typing.Union[str, pathlib.Path]) –

classmethod from_metadata(fname)[source]

Initialize a Dataset with the metadata found at fname.

Parameters:: fname (typing.Union[str, pathlib.Path]) – A URL (str) or a local path (str or pathlib.Path). If fname points to a directory, the default metadata for the respective module will be read.
Return type:: Dataset

classmethod in_dir(d, empty_tables=False)[source]

Create a Dataset in a (possibly empty or even non-existing) directory.

The dataset will be initialized with the default metadata for the respective module.

Return type:

Dataset

Parameters:

d (typing.Union[str, pathlib.Path]) –
empty_tables (bool) –

Accessing dataset metadata

property Dataset.directory: str | Path

Returns:: The location of the metadata file. Either a local directory as pathlib.Path or a URL as str.

property Dataset.module: str

Returns:: The name of the CLDF module of the dataset.

property Dataset.version: str: The CLDF version.

property Dataset.metadata_dict: dict: The TableGroup instance as dict.

property Dataset.properties: dict

Returns:: Common properties of the CSVW TableGroup of the dataset.

property Dataset.bibpath: str | Path

Returns:: Location of the sources BibTeX file. Either a URL (str) or a local path (pathlib.Path).

property Dataset.bibname: str

Returns:: Filename of the sources BibTeX file.

Accessing schema objects: components, tables, columns, etc.

Similar to capability checks in programming languages that use duck typing, it is often necessary to access a datasets schema, i.e. its tables and columns, to figure out whether the dataset fits a certain purpose. This is supported via a mapping-like interface provided by Dataset, where the keys are table specifiers or pairs (table specifier, column specifier). A table specifier can be a table’s component name or its url, a column specifier can be a column name or its propertyUrl.

check existence with in:

if 'ValueTable' in dataset: ...
if ('ValueTable', 'Language_ID') in dataset: ...

retrieve a schema object with item access:

table = dataset['ValueTable']
column = dataset['ValueTable', 'Language_ID']

retrieve a schema object or a default with Dataset.get():

table_or_none = dataset.get('ValueTableX')
column_or_none = dataset.get(('ValueTable', 'Language_ID'))

remove a schema object with del:

del dataset['ValueTable', 'Language_ID']
del dataset['ValueTable']

Note

Adding schema objects is not supported via key assignment, but with a set of specialized methods described in Editing metadata and schema.

property Dataset.tables: list[csvw.metadata.Table]

Returns:: All tables defined in the dataset.

property Dataset.components: OrderedDict[str, Table]

Returns:: Mapping of component name to table objects as defined in the dataset.

Dataset.__getitem__(item)[source]

Access to tables and columns.

If a pair (table-spec, column-spec) is passed as item, a csvw.Column will be returned, otherwise item is assumed to be a table-spec, and a csvw.Table is returned.

A table-spec may be

a CLDF ontology URI matching the dc:conformsTo property of a table
the local name of a CLDF ontology URI, where the complete URI matches the the dc:conformsTo property of a table
a filename matching the url property of a table.

A column-spec may be

a CLDF ontology URI matching the propertyUrl of a column
the local name of a CLDF ontology URI, where the complete URI matches the propertyUrl of a column
the name of a column.

Parameters:: item (typing.Union[str, csvw.metadata.Link, csvw.metadata.Table, tuple[typing.Union[str, csvw.metadata.Link, csvw.metadata.Table], typing.Union[str, csvw.metadata.Column]]]) – A schema object spec.
Raises:: SchemaError – If no matching table or column is found.
Return type:: typing.Union[csvw.metadata.Table, csvw.metadata.Column]

Dataset.__delitem__(item)[source]

Remove a table or column from the datasets’ schema.

Parameters:: item (typing.Union[str, csvw.metadata.Link, csvw.metadata.Table, tuple[typing.Union[str, csvw.metadata.Link, csvw.metadata.Table], typing.Union[str, csvw.metadata.Column]]]) – See __getitem__()

Dataset.__contains__(item)[source]

Check whether a dataset specifies a table or column.

Parameters:: item (typing.Union[str, csvw.metadata.Link, csvw.metadata.Table, tuple[typing.Union[str, csvw.metadata.Link, csvw.metadata.Table], typing.Union[str, csvw.metadata.Column]]]) – See __getitem__()
Return type:: bool

Dataset.get(item, default=None)[source]

Acts like dict.get.

Parameters:: item (typing.Union[str, csvw.metadata.Link, csvw.metadata.Table, tuple[typing.Union[str, csvw.metadata.Link, csvw.metadata.Table], typing.Union[str, csvw.metadata.Column]]]) – See __getitem__()
Return type:: typing.Union[csvw.metadata.Table, csvw.metadata.Column, None]

Dataset.get_foreign_key_reference(table, column)[source]

Retrieve the reference of a foreign key constraint for the specified column.

Parameters:

table (typing.Union[str, csvw.metadata.Table]) – Source table, specified by filename, component name or as Table instance.
column (typing.Union[str, csvw.metadata.Column]) – Source column, specified by column name, CLDF term or as Column instance.

Return type:

typing.Optional[tuple[csvw.metadata.Table, csvw.metadata.Column]]

Returns:

A pair (Table, Column) specifying the reference column - or None.

property Dataset.column_names: SimpleNamespace

In-direction layer, mapping ontology terms to local column names (or None).

Note that this property is computed each time it is accessed (because the dataset schema may have changed). So when accessing a dataset for reading only, calling code should use readonly_column_names.

Returns:: an types.SimpleNamespace object, with attributes <object>s for each component <Object>Table defined in the ontology. Each such attribute evaluates to None if the dataset does not contain the component. Otherwise, it’s an types.SimpleNamespace object mapping each property defined in the ontology to None - if no such column is specified in the component - and the local column name if it is.

property Dataset.readonly_column_names: SimpleNamespace

Returns:: types.SimpleNamespace with component names as attributes.

Editing metadata and schema

In many cases, editing the metadata of a dataset is as simple as editing Dataset.properties(), but for the somewhat complex formatting of provenance data, we provide the shortcut Dataset.add_provenance().

Likewise, csvw.Table and csvw.Column objects in the dataset’s schema can be edited “in place”, by setting their attributes or adding to/editing their common_props dictionary. Thus, the methods listed below are concerned with adding and removing tables and columns.

Dataset.add_table(url, *cols, **kw)[source]

Add a table description to the Dataset.

Parameters:

url (str) – The url property of the table.
cols (typing.Union[str, dict, csvw.metadata.Column]) – Column specifications; anything accepted by pycldf.dataset.make_column().
kw (typing.Any) – Recognized keywords: - primaryKey: specify the column(s) constituting the primary key of the table. - description: a description of the table.

Return type:

csvw.metadata.Table

Returns:

The new table.

Dataset.remove_table(table)[source]

Removes the table specified by table from the dataset.

Parameters:: table (typing.Union[str, csvw.metadata.Table]) –
Return type:: None

Dataset.add_component(component, *cols, **kw)[source]

Add a CLDF component to a dataset.

Parameters:

component (typing.Union[str, dict]) – A component specified by name or as dict representing the JSON description of the component.
kw – Recognized keywords: - url: a url property for the table; - description: a description of the table.
cols (typing.Union[str, dict, csvw.metadata.Column]) –

Return type:

csvw.metadata.Table

Dataset.add_columns(table, *cols)[source]

Add columns specified by cols to the table specified by table.

Parameters:

table (typing.Union[str, csvw.metadata.Table]) –
cols (typing.Union[str, dict, csvw.metadata.Column]) –

Return type:

None

Dataset.remove_columns(table, *cols)[source]

Remove cols from table’s schema.

Note

Foreign keys pointing to any of the removed columns are removed as well.

Parameters:

table (typing.Union[str, csvw.metadata.Table]) –
cols (typing.Union[str, csvw.metadata.Column]) –

Return type:

None

Dataset.rename_column(table, col, name)[source]

Assign a new name to an existing column, cascading this change to foreign keys.

This functionality can be used to change the names of columns added automatically by Dataset.add_component()

Parameters:

table (typing.Union[str, csvw.metadata.Table]) –
col (typing.Union[str, csvw.metadata.Column]) –
name (str) –

Return type:

None

Dataset.add_foreign_key(foreign_t, foreign_c, primary_t, primary_c=None)[source]

Add a foreign key constraint.

..note:: Composite keys are not supported yet.

Parameters:

foreign_t (typing.Union[str, csvw.metadata.Table]) – Table reference for the linking table.
foreign_c (typing.Union[str, csvw.metadata.Column]) – Column reference for the link.
primary_t (typing.Union[str, csvw.metadata.Table]) – Table reference for the linked table.
primary_c (typing.Union[str, csvw.metadata.Column, None]) – Column reference for the linked column - or None, in which case the primary key of the linked table is assumed.

Return type:

None

Dataset.add_provenance(**kw)[source]

Add metadata about the dataset’s provenance.

Parameters:: kw (typing.Any) – Key-value pairs, where keys are local names of properties in the PROV ontology for describing entities (see https://www.w3.org/TR/2013/REC-prov-o-20130430/#Entity).
Return type:: None

Adding data

The main method to persist data as CLDF dataset is Dataset.write(), which accepts data for all CLDF data files as input. This does not include sources, though. These must be added using Dataset.add_sources().

Dataset.add_sources(*sources, **kw)[source]

Add sources to the dataset.

Parameters:: sources (typing.Union[str, pycldf.sources.Source]) – Anything accepted by pycldf.sources.Sources.add().
Return type:: None

Reading data

Reading rows from CLDF data files, honoring the datatypes specified in the schema, is already implemented by csvw. Thus, the simplest way to read data is iterating over the csvw.Table objects. However, this will ignore the semantic layer provided by CLDF. E.g. a CLDF languageReference linking a value to a language will be appear in the dict returned for a row under the local column name. Thus, we provide several more convenient methods to read data.

Dataset.iter_rows(table, *cols)[source]

Iterate rows in a table, resolving CLDF property names to local column names.

Parameters:

table (typing.Union[str, csvw.metadata.Table]) – Table name.
cols (str) – List of CLDF property terms which must be resolved in resulting dict s. I.e. the row dicts will be augmented with copies of the values keyed with CLDF property terms.

Return type:

collections.abc.Generator[collections.OrderedDict[str, typing.Any], None, None]

Dataset.get_row(table, id_)[source]

Retrieve a row specified by table and CLDF id.

Raises:: ValueError – If no matching row is found.
Parameters:: table (typing.Union[str, csvw.metadata.Table]) –
Return type:: collections.OrderedDict[str, typing.Any]

Dataset.get_row_url(table, row)[source]

Get a URL associated with a row. Tables can specify associated row URLs by

listing one column with datatype anyURI or
specfying a valueUrl property for their ID column.

For rows representing objects in web applications, this may be the objects URL. For rows representing media files, it may be a URL locating the file on a media server.

Parameters:

table (typing.Union[str, csvw.metadata.Table]) – Table specified in a way that __getitem__ understands.
row (typing.Union[collections.OrderedDict[str, typing.Any], str]) – A row specified by ID or as dict as returned when iterating over a table.

Return type:

typing.Optional[str]

Returns:

a str representing a URL or None.

Dataset.objects(table, cls=None)[source]

Read data of a CLDF component as pycldf.orm.Object instances.

Parameters:

table (str) – table to read, specified as component name.
cls (typing.Optional[typing.Type]) – pycldf.orm.Object subclass to instantiate objects with.

Return type:

pycldf.util.DictTuple

Returns:

Dataset.get_object(table, id_, cls=None, pk=False)[source]

Get a row of a component as pycldf.orm.Object instance.

Parameters:

table (str) –
id_ (str) –

Return type:

pycldf.orm.Object

Writing (meta)data

Dataset.write(fname=None, zipped=None, **table_items)[source]

Write metadata, sources and data. Metadata will be written to fname (as interpreted in pycldf.dataset.Dataset.write_metadata()); data files will be written to the file specified by csvw.Table.url of the corresponding table, interpreted as path relative to directory().

Parameters:

zipped (typing.Optional[collections.abc.Iterable]) – Iterable listing keys of table_items for which the table file should be zipped.
table_items (list[collections.OrderedDict[str, typing.Any]]) – Mapping of table specifications to lists of row dicts.
fname (typing.Optional[pathlib.Path]) –

Return type:

pathlib.Path

Returns:

Path of the CLDF metadata file as written to disk.

Dataset.write_metadata(fname=None)[source]

Write the CLDF metadata to a JSON file.

Fname:: Path of a file to write to, or None to use the default name and write to directory().
Parameters:: fname (typing.Union[str, pathlib.Path, None]) –
Return type:: pathlib.Path

Dataset.write_sources(zipped=False)[source]

Write the sources BibTeX file to bibpath()

Return type:: typing.Optional[pathlib.Path]
Returns:: None, if no BibTeX file was written (because no source items were added), pathlib.Path of the written BibTeX file otherwise. Note that this path does not need to exist, because the content may have been added to a zip archive.
Parameters:: zipped (bool) –

Reporting

Dataset.validate(log=None, validators=None, ontology_path=None)[source]

Validate schema and data of a Dataset:

Make sure the schema follows the CLDF specification and
make sure the data is consistent with the schema.

Parameters:

log (logging.Logger) – a logging.Logger to write ERRORs and WARNINGs to. If None, an exception will be raised at the first problem.
validators (list[tuple[typing.Optional[str], str, typing.Callable[[pycldf.dataset.Dataset, csvw.metadata.Table, csvw.metadata.Column, collections.OrderedDict[str, typing.Any]], None]]]) – Custom validation rules, i.e. triples (tablespec, columnspec, attrs validator)
ontology_path (typing.Union[str, pathlib.Path, None]) –

Raises:

ValueError – if a validation error is encountered (and log is None).

Return type:

bool

Returns:

Flag signaling whether schema and data are valid.

Dataset.stats(exact=False)[source]

Compute summary statistics for the dataset.

Return type:: list[tuple[str, str, int]]
Returns:: List of triples (filename, component, rowcount).
Parameters:: exact (bool) –

Dataset discovery

We provide two functions to make it easier to discover CLDF datasets in the file system. This is useful, e.g., when downloading archived datasets from Zenodo, where it may not be known in advance where in a zip archive the metadata file may reside.

pycldf.sniff(p)[source]

Determine whether a file contains CLDF metadata.

Parameters:: p (pathlib.Path) – pathlib.Path object for an existing file.
Return type:: bool
Returns:: True if the file contains CLDF metadata, False otherwise.

pycldf.iter_datasets(d)[source]

Discover CLDF datasets - by identifying metadata files - in a directory.

Parameters:: d (typing.Union[str, pathlib.Path]) – directory in which to look for CLDF datasets (recursively).
Return type:: collections.abc.Generator[pycldf.dataset.Dataset, None, None]
Returns:: generator of Dataset instances.

Sources

When constructing sources for a CLDF dataset in Python code, you may pass pycldf.Source instances into Dataset.add_sources(), or use pycldf.Reference.__str__() to format a row’s source value properly.

Direct access to pycldf.dataset.Sources is rarely necessary (hence it is not available as import from pycldf directly), because each pycldf.Dataset provides access to an apprpriately initialized instance in its sources attribute.

class pycldf.Source(genre, id_, *args, _check_id=True, _lowercase=False, _strip_tex=None, **kw)[source]

A bibliograhical record, specifying a source for some data in a CLDF dataset.

Parameters:

genre (str) –
id_ (str) –
_check_id (bool) –
_lowercase (bool) –
_strip_tex (typing.Optional[collections.abc.Iterable[str]]) –

property entry: Entry: Converts Source to a pybtex Entry.

classmethod from_entry(key, entry, **_kw)[source]

Create a cls instance from a simplepybtex entry object.

Parameters:

key (str) – BibTeX citation key of the entry
entry (simplepybtex.database.Entry) – simplepybtex.database.Entry instance
_kw – Non-bib-metadata keywords to be passed for cls instantiation

Returns:

cls instance

static persons(s)[source]

Yields persons encoded in an author names string.

Parameters:: s (str) –
Return type:: collections.abc.Generator[simplepybtex.database.Person, None, None]

refkey(year_brackets='round')[source]

Compute an author-year type reference key for the item.

Parameters:: year_brackets (typing.Optional[typing.Literal['round', 'square', 'curly']]) –

class pycldf.Reference(source, desc)[source]

A reference connects a piece of data with a Source, typically adding some citation context often page numbers, or similar.

Parameters:

source (pycldf.sources.Source) –
desc (typing.Optional[str]) –

__str__()[source]: String representation of a reference according to the CLDF specification.

See also

https://github.com/cldf/cldf#sources

class pycldf.dataset.Sources[source]

A dict like container for all sources linked to data in a CLDF dataset.

add(*entries, **kw)[source]

Add a source, either specified as BibTeX string or as Source.

Parameters:: entries (typing.Union[str, pycldf.sources.Source]) –
Return type:: None

expand_refs(refs, **kw)[source]

Turn a list of string references into proper Reference instances, looking up sources in self.

This can be used from a pycldf.Dataset as follows:

>>> for row in dataset.iter_rows('ValueTable', 'source'):
...     for ref in dataset.sources.expand_refs(row['source']):
...         print(ref.source)

Parameters:: refs (collections.abc.Iterable[str]) –
Return type:: collections.abc.Iterable[pycldf.sources.Reference]

classmethod from_file(fname)[source]

Instantiate an instance from the data in a BibTeX file.

Parameters:: fname (typing.Union[str, pathlib.Path]) –
Return type:: pycldf.sources.Sources

static parse(ref)[source]

Parse the string representation of a reference into source ID and context.

Raises:: ValueError – if the reference does not match the expected format.
Parameters:: ref (str) –
Return type:: tuple[str, str]

read(fname, zipped=False, **kw)[source]

Read sources from a BibTex file (possibly specified via URL).

Parameters:: fname (typing.Union[str, pathlib.Path]) –

validate(refs)[source]

Make sure refs are valid. If not, raises Exceptions.

Parameters:: refs (typing.Union[str, list[str]]) –
Return type:: None

write(fname, ids=None, zipped=False, **_)[source]

Write sources to a file (if there are any).

Parameters:: fname (typing.Union[str, pathlib.Path]) –
Return type:: typing.Optional[pathlib.Path]

Subclasses supporting specific CLDF modules

Note

Most functionality provided through properties and methods described below is implemented via the pycldf.orm module, and thus subject to the limitations listed at ./orm.html

class pycldf.Generic(tablegroup)[source]

Generic datasets have no primary table.

Parameters:: tablegroup (csvw.metadata.TableGroup) –

property primary_table: None: Returns the primary table for the dataset.

class pycldf.Wordlist(tablegroup)[source]

Wordlists have support for segment slice notation.

Parameters:: tablegroup (csvw.metadata.TableGroup) –

get_segments(row, table='FormTable')[source]

Retrieve the list of segments of a form.

Parameters:: row (collections.OrderedDict[str, typing.Any]) –
Return type:: list[str]

get_subsequence(cognate, form=None)[source]

Compute the subsequence of the morphemes of a form which is specified in a partial cognate assignment.

Parameters:

cognate (collections.OrderedDict[str, typing.Any]) – A dict holding the data of a row from a CognateTable.
form (typing.Optional[str]) –

Return type:

list[str]

property primary_table: str: Returns the primary table for the dataset.

class pycldf.StructureDataset(tablegroup)[source]

Parameters in StructureDataset are often called “features”.

Parameters:: tablegroup (csvw.metadata.TableGroup) –

property features: Just an alias for the parameters.

property primary_table: str: Returns the primary table for the dataset.

class pycldf.TextCorpus(tablegroup)[source]

In a TextCorpus, contributions and examples have specialized roles:

Contributions are understood as individual texts of the corpus.
Examples are interpreted as the sentences of the corpus.
Alternative translations are provided by linking “light-weight” examples to “full”, main examples.
The order of sentences may be defined using a position property.

>>> crp = TextCorpus.from_metadata('tests/data/textcorpus/metadata.json')
>>> crp.texts[0].sentences[0].cldf.primaryText
'first line'
>>> crp.texts[0].sentences[0].alternative_translations
[<pycldf.orm.Example id="e2-alt">]

Parameters:: tablegroup (csvw.metadata.TableGroup) –

get_text(tid)[source]

Retrieve a text by ID.

Parameters:: tid (str) –
Return type:: typing.Optional[pycldf.orm.Object]

property primary_table: str: Returns the primary table for the dataset.

property sentences: list[pycldf.orm.Example]: Sentences of the corpus.

property texts: DictTuple | None: Retrieve texts.