4.10.2

Fixed

• Audio visualizer for .wav types was not playing the file due to a MIME type
  mixup (#815)
• Updated to graphANNIS bugfix version 2.4.3

3.7.1 (legacy version)

This is a maintenance release for the legacy version of ANNIS 3. Use ANNIS 4 whenever possible.

[Fixed Bugs]
#813 Missing join on corpus reference when multiple corpora are selected

v4.10.1

Fixed

• The GridExporter could randomly fail to include certain matches. This was
  caused by a buggy logic, where a result was ommitted when any node in the
  result was not connected to a token. Instead, just exclude this single node
  and also check the other nodes. Only fail when no node is connected to any
  token. There is now also a warning in the console messages when this occurs.
• Long running exports could fail when you are logged in and the internal token
  expired. The communication with the REST webservice has been refactored to
  always get the newest token (automatically refreshed by spring) before
  executing the call.

v4.10.0

Added

• Show informative error message when segmentation in export does not exist
  instead of a generic stack trace

Fixed

• Allow execution on macOS with Apple Silicon (M1/M2) using the Rosetta
  emulation when starting the graphANNIS service process

v4.9.9

Fixed

• Fixed issue with corpus names containing special characters when login is
  active. There was a "Forbidden" error when trying to access the corpus
  configuration.

v4.9.8

Fixed

• Fix display of match path in results that match a document or corpus (metadata
  search) and the corpus has a special character like an umlaut.
• Do not overwrite existing service.toml file (which might get the file
  corrupted in certain cases) but always write the updated configuration to a
  temporary TOML-file.
• Update Apache commons-text dependency to >= 1.0 because of of CVE-2022-42889

v4.9.7

Fixed

• Fixed handling of virtual tokenization in exporters when relANNIS corpora had
  an explicit mapping defined. ANNIS3 had the virtual_tokenization_mapping and
  virtual_tokenization_from_namespace fields in the
  ExtData/corpus.properties file which allowed to configure which span
  annotation belonged to which segmentation. In the exporters, this was used to
  reconstruct a Salt graph where the virtual token are replace with timeline
  items and the segmentation spans become actual SToken. There are several
  corpora that depend on this functionality and where the exporters failed to
  generated proper text spans because of this.
• The context defined in the exporter panel was not taken into account.

v4.9.6

Fixed

• No Row outside dataProvider size error message when the corpus list is
  empty.
• The segmentation context was automatically reset to its default value from
  configuration when changed manually.
• Update to graphANNIS 2.3.0 which has various bug fixes, including fixing
  broken result ordering for certain queries.
• Fixed order of subgraphs when there are gaps in the result
• Update jsoup dependency to 1.15.3

v4.9.5

Fixed

• Corpus selection was not reliably shown when executing a corpus query or
  corpus selecting by using the URL.
• Update to graphANNIS 2.2.2 to avoid out of memory errors for results which a
  large number of matches.

v4.9.4

Fixed

• Ignore terminal namespace when null in tree visualizers. The tree visualizer
  used to fail, when the terminal namespace mapping is not configured, thus
  null. Null was then matched against the terminal node's actual namespace in
  ANNIS4. In ANNIS3, though, the namespace for terminals was ignored when the
  mapping was not configured (i.e. null). This behavior is now restored.