Changelogs » Agraph-python




Raise error when attemping to federate certain sessions

A session started by AllegroGraphServer.openSession() cannot
be federated because this type of session may not be for a
single repository and may in fact already be a session over
a federation of repositories.  To federate repositories
pass a set of Repository.getConnection() objects.
With this change an exception will be raised when attempting to
federate a RepositoryConnection created by AllegroGraphServer.openSession().



Non-RDF document import

JSON and CSV documents can be imported into AllegroGraph using
the new 'transform' service introduced in AG 6.5.0. This can
now be accessed from the Python client by calling the
``addDocumentFile()`` method of the connection object.
Document data from strings or Python dictionaries can be added
with ``addDocumentData()``.

Update dependencies which have security vulnerabilities

Bumped urllib3 from 1.22 to 1.23 and requests from 2.18.4 to 2.20.0
according to recommendations made by Github.

Ensure sessions are closed properly

Sessions created via the ``openFederation()`` or ``openSession()``
methods of franz.openrdf.sail.allegrographserver.AllegroGraphServer
were not closed when ``close()`` was called on the resulting
connection.  This has been fixed.



JSON-LD support

A new RDF format (`RDFFormat.JSONLD`) has been added. When importing
documents from strings or files it is now possible to specify
a few JSON-LD specific arguments:

- `json_ld_context`: a JSON object defining the mapping between
JSON-LD terms and triples.
- `json_ld_store_source`: a flag that can be used to persist
the whole input document in a single triple.
- `allow_external_references` a flag that must be explicitly
set to True to allow the JSON-LD importer to retrieve
external resources referenced in the input document or in
the `json_ld_context` argument.
- `external_reference_timeout`: HTTP timeout that will be
used by the server when retrieving external resources.

In addition the `addData()` method can import a JSON-LD document
in the form of a dictionary.

URI object changes

URI objects are now canonicalized by default. This means that
two equal URIs are guaranteed to be the same object. This behavior
can be disabled by passing `canonical=False` when creating a URI.

URI objects no longer carry an attribute dictionary. This means
it is no longer possible to add arbitrary fields to an URI object.

rfe15690: warmup support

Add `RepositoryConnection.warmup()` to allow the user to
warm up the string table and/or indices.

User data access

AllegroGraph allows each user to store arbitrary key-value data
on the server. This storage can now be accessed from Python by
using new ``AllegroGraphServer`` methods:

- ``listUserData()``
- ``getUserData()``
- ``setUserData()``
- ``deleteUserData()``



Pandas support

It is now possible to turn a query result into a Pandas DataFrame
by calling the ``toPandas()`` method of the result object. Note that
Pandas must be installed separately for this to work.



bug25281: Proxy settings are ignored
Proxy settings used to be ignored when the requests backend was used.
This has been corrected.

Thanks to Iván Darío Ramos Vacca for reporting the bug and providing
a fix.



A bugfix release that adds some missing dependencies that are needed
when using Python < 3.5.



Triple attributes

Added support for triple attributes (requires AG >= 6.1). Specifically
it is now possible to:

- Set and retrieve the static attribute filter using
``conn.setAttributeFilter()`` and ``conn.getAttributeFilter()``
- Set and retrieve user attributes (that will be sent with each
request) using ``conn.setUserAttributes()`` and
- Manage attribute definitions using various methods in the
connection class..
- Add triples with attributes - a new keyword parameter named
'attributes' has been added to methods that add triples,
such as ``addData()``. It is also possible to pass five-element
tuples to ``addTriples()``, where the fifth element is
a dictionary of attribute values.

Distributed transaction settings

It is now possible to configure distributed transaction parameters
in multiple ways:

- By passing arguments to the ``commit()`` method
- By calling ``setTransactionSettings()`` on the connection object.
- By using a context manager returned by the
``temporaryTransactionSettings()`` method.

In all cases the settings can be passed either in a single
``TransactionSettings`` object or as individual keyword arguments.

Enhanced namespace objects

Namespace objects can now create URIs when indexed or called like a
function. This makes it easier to create URIs where the local name is
not a valid attribute name:

>>> from franz.openrdf.connect import ag_connect
>>> conn = ag_connect('repo')
>>> ex = conn.namespace('')
>>> ex('is')
>>> ex['def']



Jupyter-friendly stdout

The ``output_to`` context manager (used internally when writing output
to stdout) has been modified to work better in environments that
hijack the ``sys.stdout`` value, such as Jupyter notebooks or IDLE.



Resolved issues with running unit tests from a wheel

Some unit tests used to fail when the module was installed
from a binary wheel. This has been corrected.

bug25081: The 'context' argument to addTriples() is broken

Using the ``addTriples()`` method with the ``context`` parameter
set to a non-default value used to produce errors::

>>> conn.addTriples([(s, p, o)], context=g)
400 MALFORMED DATA: Invalid graph name: (<ex://g>)

This has been corrected. Context can now be set to a single URI
or a list of URIs. Both URI objects and strings are supported.

bug25079: Statement objects not created from strings are broken

Statement objects that were created in user code were not fully
functional. In particular attempts to convert such statements to
strings or to pass them to addTriples() would fail.

This has been corrected.

Namespace objects

Namespace objects can be used to create URIs, as in the following

>>> from franz.openrdf.connect import ag_connect
>>> conn = ag_connect('repo')
>>> ex = conn.namespace('')



New query methods

Four new methods have been added to the RepositoryConnection class:

- executeTupleQuery()
- executeGraphQuery()
- executeBooleanQuery()
- executeUpdate()

These can be used to prepare and evaluate a SPARQL query in a single

New tutorial

The tutorial has been updated and restyled using Sphinx.

Finalizers for query results

All result objects are now closed automatically when garbage collected.
This makes it possible to write simple loops like the one below::

for stmt in conn.executeTupleQuery('...'):

without having to use the ``with`` statement, since reference counting
will ensure that the query result is closed at the right time. Note that
this should not be relied upon in more complex scenarios, where circular
references might occur and prevent the result object from being closed.

Connection parameters can now be passed in environment variables

The following environment variables are now used when connecting
to the server:

- ``AGRAPH_HOST`` - server address, the default is ''
- ``AGRAPH_PORT`` - port number (default: 10035 for HTTP connections,
10036 for HTTPS).
- ``AGRAPH_USER`` - Username, no default.
- ``AGRAPH_PASSWORD`` - Password, no default.

Note that parameters passed to ``ag_connect()`` or  ``AllegroGraphServer()``
will override these variables.

Various fixes related to data export

Specifically the following adjustments have been done:

- Changed the default RDF export format to N-Quads.
- Fixed a bug where errors returned during export
caused an encoding error.
- Provided a default format (CSV) for tuple queries.
- Value of the output parameter can now be True (stdout)
or a file descriptor.



bug24892: Time parsing fixes

The Python client used to fail when trying to retrieve a
datetimeValue() of a literal that contained time zone
information. This has been corrected.

All datetime objects created by the Python API are now timezone-aware.

rfe15005: duplicate suppression control API

It is now possible to set and query the duplicate suppression policy of
a repository from Python, using three new methods of the connection

- getDuplicateSuppressionPolicy()
- setDuplicateSuppressionPolicy()
- disableDuplicateSuppression()

New export methods

A new mechanism for exporting data has been added. It utilizes a new
``output`` parameter that has been added to the following methods:

- RepositoryConnection.getStatements()
- RepositoryConnection.getStatementsById()
- TupleQuery.evaluate()
- GraphQuery.evaluate()

Setting the new parameter to a file name or a file-like object
will cause the data that would normally be returned by the call
to be saved to the specified file instead. Serialization format
can be controlled by setting another new parameter,



New versioning scheme

Client versions no longer match the server version. Major version
number has been bumped to 100 to avoid confusion.

bug24819: Circular import

Importing com.franz.openrdf.query.query failed due to a circular
import. Thanks to Maximilien de Bayser for reporting this.

bug24826: removeStatement uses context instead of object

The removeStatement method of RepositoryConnection was broken.
Patch by Maximilien de Bayser.


bug24728: Incorrect conversion between boolean literals and Python values

The booleanValue() method of the Literal class used to work
incorrectly.  It would return True for any literal that is not empty,
including the "false"^^xsd:boolean literal.  This has been corrected -
the function will now return expected values for literals of type
xsd:boolean.  Result for other types remains undefined.


bug24680: to_native_string is broken on Python 2

The Python client sometimes failed while processing values with
non-ascii characters, showing the following error message:

UnicodeEncodeError: 'ascii' codec can't encode characters in position ??: ordinal not in range(128)

This has been corrected.


Released with AllegroGraph 6.2.2. Change log for this and all previous
Python client releases can be found in AllegroGraph release notes: