Skip to content

CHANGELOG.md

Latest commit

 

History

History
89 lines (69 loc) · 5.42 KB

CHANGELOG.md

File metadata and controls

89 lines (69 loc) · 5.42 KB
Raw

Version history

1.4.1

  • Updates recommended RKorAPClient version to 1.4.1
  • Requires Python >= 3.10 (support for Python 3.7–3.9 dropped, Python 3.14 added)
  • libv8 is no longer needed as a system dependency
  • Python dicts are now converted to named R vectors (or named lists), e.g. vc={"Germany": "pubPlaceKey=DE", "Switzerland": "pubPlaceKey=CH"} to label virtual corpora
  • Documentation and tests updated for the new features listed below, including labelling virtual corpora (label column) and caching results with cacheAs
  • follows the DeReKo-KorAP-2026-II release: corpusQuery() now fetches the new dmozDomain and wikiDomain metadata fields, which replaced textClass (RKorAPClient #31)
  • long running searches are no longer aborted after 10 minutes; only the connection's timeout applies
  • corpusQuery() and frequencyQuery() now warn about failed or incomplete (timeExceeded) searches instead of silently reporting 0 or too few hits; their results carry a new queryDuration column
  • clearer verbose output: series of queries or virtual corpora are logged as an aligned table, with more accurate ETAs
  • cacheAs parameter for transparent result caching, now supported by collocationAnalysis(), frequencyQuery(), corpusStats(), collocationScoreQuery() and textMetadata()
  • names of a named vc vector are consistently used as labels by frequencyQuery(), corpusStats() and collocationScoreQuery()
  • changed logDice values: now computed as defined by Rychlý (2008), so values are comparable to Sketch Engine and other tools
  • changed ll() values: contingency table now scales the sample size and both marginals with the window size (Evert 2004)
  • collocationAnalysis() discards collocates occurring less often than expected by chance (new minObservedExpectedRatio parameter, default 1; set to 0 for the previous behaviour)
  • experimental support for comparing collocation analyses across multiple (optionally labelled) vcs
  • collocationScoreQuery() accepts a vector of collocates
  • fixed collocation analysis dropping snippets with certain markup shapes, e.g. sentence-bounded contains(<base/s=s>, ...) queries (RKorAPClient #14)
  • fixed score threshold in recursive collocation analysis

1.2.1

  • Updates recommended RKorAPClient version to 1.2.1
  • fetchAnnotations() method added to KorAPQuery class, to fetch annotations for all collected matches

1.1.0

  • Updates recommended RKorAPClient version to 1.1.0
  • fixed bug with fetching result pages with an offset >= 10,000 (=1e+05 ...) issue #25
  • timed out corpus queries are no longer cached (see issue #7)
  • improved stability of ci function
  • improved error handling
  • improved logging
  • added ETAs to logging in verboose mode

1.0.1

  • Fixed NULL warning in fetch functions

1.0.0

  • Simplified authorization process for accessing restricted data via the new auth() function
  • Fixed issues with tokenized matches in corpusQuery results
  • Fixed smoothing constant in mergeDuplicateCollocates function
  • Fixed chainability of fetch methods in corpusQuery

0.9.0

  • Updates recommended RKorAPClient version to 0.9.0
  • Added matchStart and matchEnd columns to corpusQuery results, containing the start and end positions of the match in the text
  • Added mergeDuplicateCollocates function to merge collocation analysis results for different context positions
  • Added a query column to collocation analysis results
  • Improved documentation for span parameter in collocationAnalysis functions
  • Updated textMetadata method to use new metadata fields API, if available, to retrieve custom metadata for a text based on its sigle
  • Added new unit tests to cover the new features and changes

0.8.1

  • Updates recommended RKorAPClient version to 0.8.1
  • fixed rare frequencyQuery incompatibility with outdated KorAP instances (see KorAP/Kustvakt#668)

0.8.0

  • Updates recommended RKorAPClient version to 0.8.0

  • Added textMetadata KorAPConnection method to retrieve all metadata for a text based on its sigle

  • Added webUiRequestUrl column also to corpusStats results, so that also virtual corpus definitions can be linked to / tested directly in the KorAP UI

  • Uses server side tokenized matches in collocation analysis, if supported by KorAP server

  • Unless metadataOnly is set, also tokenized snippets are now retrieved in corpus queries (stored in res.slots['collectedMatches']['tokens.left'], res.slots['collectedMatches']['tokens.match'], res.slots['collectedMatches']['tokens.right']). Because Pandas data frames cannot store lists, tokens are stored as strings, tab separated.

  • Python 3.11 and 3.12 are now supported

  • Python 3.7 support has been dropped (by rpy2 dependency)

0.7.5

  • Updates recommended RKorAPClient version to 0.7.5
    • fixes collocation scores for lemmatized node or collocate queries
  • Automatically converts again between rpy and py objects, most importantly between Pandas and R data frames, also with newer versions of rpy2
    • Fixes "Hello world" example in Readme.md
  • Updates references / citation information
  • Changes corpusStats to return a pandas.DataFrame by default
  • Advertises Python 3.10 as supported
  • Adds interactive plot examples using Vega-Altair

0.7.1