bibnets 0.4.4

CRAN pre-test fix

The four test-equiv-*.R equivalence suites (vs bibliometrix and biblionetwork) have been moved out of the package into a local-only local_testing_and_equivalence/ directory. These developer checks pulled in data.table/biblionetwork, whose OpenMP parallelism caused the “CPU time 4 times elapsed time” NOTE on the Debian r-devel pre-test. They remain runnable locally but are no longer part of R CMD check.
bibliometrix, biblionetwork, and data.table removed from Suggests — they were used only by the relocated equivalence tests.
tests/testthat.R keeps a 2-thread BLAS/OpenMP cap as defence for crossprod() / tcrossprod() in multiply_bipartite().
No user-facing code changes.

All no-run wrappers in reader examples replaced with runnable examples, per CRAN reviewer guidance. File-based readers (read_scopus, read_wos, read_dimensions, read_lens) now use small bundled fixtures under inst/extdata/, reached via system.file(). API-wrapper readers (read_openalex, read_crossref) use an inline data frame matching the upstream column shape so the conversion path runs without a network call. read_biblio examples now demonstrate multi-file, directory, and generic-CSV modes against the bundled fixtures.
New fixtures: inst/extdata/scopus_sample.csv, wos_sample.txt, dimensions_sample.csv, lens_sample.csv (2 records each).

Title renamed to “Importing, Constructing, and Exporting Bibliometric Networks” to reflect the full lifecycle scope.
Description rewritten to lead with attention-weighted networks (lead, last, proximity, circular) and other differentiators (position-aware counting, similarity/dissimilarity normalisations, temporal windows, disparity-filter backbone, historiograph, local citation scoring), dropping the enumeration of standard co-network types.
README intro re-leads with capabilities; numerical method counts removed.
Internal-comment author attributions stripped throughout.

read_lens() no longer inflates output to n^2 rows when neither Lens ID nor ID columns are present.
read_openalex() no longer inflates output to n^2 rows when the id column is absent.
read_scopus() now normalises empty-string DOIs to NA, so is.na(doi) deduplication checks behave as expected.
read_wos() empty-file return now includes the keywords_plus list-column to match the non-empty schema.
read_crossref() no longer crashes with “row names contain missing values” when the issued column has NA entries.

Converter examples (to_igraph(), to_tbl_graph(), to_cograph()) now use @examplesIf requireNamespace(...) so they execute when the suggested package is installed instead of being silently skipped.
read_biblio(), read_bibtex(), and read_ris() now ship runnable examples backed by either the bundled extdata/openalex_works.csv fixture or a tempfile()-based minimal record.
Reference list streamlined across DESCRIPTION, README, vignette, and Rd files.

Added eight new test files covering read_scopus(), read_wos(), read_ris(), read_lens(), read_dimensions(), read_crossref(), read_biblio(), read_openalex(), plus dedicated coverage for R/edgelist.R and build_bipartite_long().
Suite size: 1268 tests (was 499). Package line coverage: 92.5% (was 61.8%).

temporal_network() — builds time-windowed networks with fixed, sliding, or cumulative strategies. Results include a window column for easy stacking.
historiograph() — Garfield-style chronological citation network among the most locally cited documents.
local_citations() — counts within-dataset citations (Local Citation Score).
backbone() — disparity filter for extracting statistically significant edges from dense weighted networks.
prune() — threshold and top-n edge pruning.
read_biblio() — universal reader with auto-format detection (Scopus, WoS, BibTeX, RIS, Dimensions, Lens.org).
read_dimensions() — Dimensions CSV export reader.
read_crossref() — converter for rcrossref::cr_works() output.
to_gephi() — exports node and edge tables in Gephi CSV format; writes nodes.csv + edges.csv when a directory path is supplied.
to_graphml() — pure base-R GraphML writer; no XML package required.
to_cograph() — converts edge list to a cograph_network object with optional node metadata for direct use with cograph::splot().

All edge list functions now sort output by weight descending and reset row names.
local_citations() canonical column order: id, lcs, gcs, year, title, journal, doi.
historiograph() empty-result schema matches non-empty schema.
All readers share a standard column order: id, title, year, journal, doi, cited_by_count, abstract, type, authors, references, keywords, then source-specific extras.
backbone() and prune() use single-pass O(m) node statistics via tapply() / split() — faster on large networks.
temporal_network() converted from for loop to lapply.
read_dimensions() / read_crossref() now apply standardize_authors() and standardize_refs() for consistency with other readers.

Argument count renamed to counting; measure renamed to similarity across all network functions.
co_network() renamed to conetwork().

Initial release.

8 network builders: author_network(), document_network(), reference_network(), keyword_network(), institution_network(), country_network(), source_network(), conetwork().
13 counting methods including harmonic, geometric, golden ratio, adaptive geometric.
6 similarity normalisations: association strength, cosine, Jaccard, inclusion, equivalence.
6 readers: Scopus, Web of Science, OpenAlex, BibTeX, RIS, Lens.org.
Converters: to_igraph(), to_tbl_graph(), to_matrix().
Numerically validated against bibliometrix and biblionetwork.