textreuse 1.0.0

textreuse 1.0.0

This release combines several years of maintenance and feature work into a single 1.0.0 release for CRAN preparation. It improves text input, corpus construction, local alignment behavior, candidate generation, LSH workflows, matrix export, performance, and documentation reliability.

Text input and corpus construction

• TextReuseTextDocument() and TextReuseCorpus() now accept an encoding argument, making it easier to read source files whose text encoding is known or differs from the platform default.
• TextReuseCorpus() now keeps skipped-document bookkeeping deterministic. Skipped documents are reported consistently, and skip metadata is available even when skip_short = FALSE.
• Very short documents are handled more predictably when skip n-grams are used, avoiding assertion failures and making corpus construction easier to diagnose.

Alignment and match inspection

• align_local() now returns an empty local alignment instead of throwing an error when two texts have no matching words. This makes batch alignment workflows easier to run because no-match pairs can be represented directly.
• align_local() gains preserve_punctuation, allowing displayed alignments to keep punctuation from the original texts when that context is useful.
• New count_matches() and matching_tokens() helpers expose absolute match counts and the matched tokens themselves, so users can inspect what drove a similarity score rather than relying only on a ratio.

Candidate generation and comparison

• New token-index helpers find candidate document pairs from shared n-grams, giving users another way to identify likely reuse pairs before running more expensive comparisons.
• pairwise_candidates() and matrix conversion now preserve all document IDs, including documents without returned candidate pairs.
• as_sparse_matrix() provides a sparse matrix representation of candidate results, which is more convenient for downstream modeling, graph analysis, and workflows with many documents.

Locality-sensitive hashing and performance

• lsh_add() can add new documents to an existing LSH bucket cache, so users can extend an index without rebuilding it from scratch.
• lsh_compare() can run comparisons in parallel on non-Windows platforms when options(mc.cores) is set.
• Long-running C++ hashing and n-gram loops now check for user interrupts, so expensive jobs can be stopped more cleanly from R.

Compatibility, documentation, and CI

• Compatibility with current dplyr and tidyr releases has been refreshed.
• README, vignette, reference, and pkgdown examples were regenerated against current package output.
• Stale external links and documentation badges were updated so package checks and the public documentation site are cleaner.
• pkgdown site font sizing is unified across pages.
• AppVeyor is configured to run the R build path instead of MSBuild project detection.

Verification

• pkgdown site build with run_dont_run = TRUE completed successfully.
• R CMD check --no-manual textreuse_1.0.0.tar.gz: OK.
• R CMD check --as-cran --no-manual textreuse_1.0.0.tar.gz: 3 NOTEs, documented in cran-comments.md.