Releases: audeering/audb
Releases · audeering/audb
Release v1.11.0
- Added: a sampling rate of 24000 Hz as flavor
- Added: a sampling rate of 22050 Hz as flavor
- Added: support for Python 3.13
(without Artifactory backend) - Added: support for Python 3.12
(without Artifactory backend) - Changed: switch default repository
toaudb-public
,
hosted on S3 - Changed:
audb.Repository.create_backend_interface()
andaudb.publish()
now raise aValueError
for a repository with non-registered backends,
or an Artifactory backend under Python>=3.12 - Changed: skip non-registered backends
without raising an error
in all functions with read-only access to repositories - Changed: simplify quickstart section
of the documentation - Changed: depend on
audbackend>=2.2.1
- Changed: depend on
audeer>=2.2.0
- Deprecated: a sampling rate of 22500 Hz as flavor
- Removed: file-system repository from default configuration
- Fixed: handle an empty configuration file
Release v1.10.2
- Fixed: remove extra
"/"
at end of dataset names
inaudb.available()
for S3 and Minio backends
Release v1.10.1
- Added:
"s3"
as a registered backend name - Changed: depend on
audbackend>=2.2.0
- Changed: make Artifactory backend optional,
to allow importingaudb
in Python>=3.12 - Fixed: speedup
audb.available()
for S3 and Minio backends
Release v1.10.0
- Added: support for repositories
on S3 and MinIO servers,
using theminio
backend
ofaudbackend
- Changed: depend on
audbackend>=2.1.0
Release v1.9.0
- Added: pseudo-streaming support with
audb.stream()
,
which returns the newaudb.DatabaseIterator
object.
In each iteration
it will load a few rows from a requested table
and downloads corresponding media files - Added:
map
argument toaudb.load_table()
,
which behaves identical to themap
argument
ofaudformat.Database.get()
- Added:
pickle_tables
argument
toaudb.load()
,
audb.load_to()
andaudb.load_table()
with default value ofTrue
.
It can be used to disable
storing tables as pickle files
in cache/root folder - Fixed:
audb.load_table()
now only loads additional misc tables,
that are used as scheme labels inside the requested table,
and not in the whole database
Release v1.8.0
- Added: support for publishing tables as parquet files
- Changed: depend on
audeer >=2.1.0
- Changed: depend on
audformat >=1.2.0
- Changed: depend on
pandas >=2.1.0
- Fixed: update progress bar at least every second
inaudb.load()
,
audb.load_attachment()
,
audb.load_media()
,
audb.load_tables()
,
audb.load_to()
,
audb.publish()
- Removed: support for Python 3.8
Release v1.7.4
- Fixed: ensure correct data types
in dependency table
when loaded from a version in cache,
stored byaudb<=1.6.3
Release v1.7.3
- Fixed: ensure correct data types
in dependency table
when loaded from cache - Fixed: publishing an update of a database
when the previous version
was stored in cache
by an older version ofaudb
- Fixed: loading of database attachments
whenaudb.config.CACHE_ROOT
andaudb.config.SHARED_CACHE_ROOT
point to the same folder - Fixed: ensure
audb.versions()
does not fail
when database is not available
in a repository - Fixed: loading of dependency table from cache
when the previous version
was stored in cache
by a differentpandas
version
Release v1.7.2
- Fixed: loading of dependency table from cache
under Python 3.8,
when stored by an older version ofaudb
Release v1.7.1
- Fixed: require
pandas>=2.0.1
forpyarrow
based data types