Skip to contents

Changelog

Source: NEWS.md

codecheck (development version)

New Features

  • New publish_wikibase_pages() rewrites every generated page on the CODECHECK Wikibase from the register and the instance as they are now, skipping pages that have not changed (make wikibase in the register, register#50).
  • The Wikibase Main Page is generated too, and lists every generated page, including Project:Example queries (register#50).
  • Project:Certificates links each certificate’s Wikidata item, as recorded in register.csv (register#50).
  • Project:Wikidata export lists the QuickStatements batches the edit log records as run, and whether register.csv holds each certificate’s item (register#50).

Bug Fixes

  • Project:Wikidata export says whether the export has run, instead of claiming nothing has been sent, and lists the certificates whose checked work has no DOI (register#50).

codecheck 0.31.0

New Features

  • The codechecker lists and the register’s persons.csv take a fediverse account (@user@instance), shown as a rel="me" link on person pages, under sameAs in their Schema.org metadata and in their index.json (register#217).
  • venues.csv takes a fediverse account and hashtags, shown on venue pages and in their JSON (register#217).
  • codecheck_spec_version() uses the version a codecheck.yml names, including the historical https://codecheck.org.uk/spec/1.0, and otherwise the version current at its check_time or the new modified argument, instead of always the newest.
  • Person pages, their stats.json and their Schema.org metadata (knowsAbout) list a codechecker’s fields and languages from codecheckers.csv (register#168).
  • The README and DESCRIPTION list the system libraries needed to install the package, e.g. libpoppler-cpp-dev and libmagick++-dev on Debian/Ubuntu (closes codecheckers/codecheck#91).
  • Rules CC-MET-005 to CC-MET-008 compare the paper metadata with OpenAlex instead of Crossref, which covers arXiv DOIs and holds far more author ORCIDs (closes codecheckers/codecheck#92).
  • The bundled rule files are refreshed from the register: CC-MET-005 to CC-MET-008 are renamed paper-*, keeping their identifiers (register#220).
  • New validate_codecheck_yml_metadata() runs those rules; validate_codecheck_yml_crossref() is deprecated and calls it (codecheck#92).
  • validate_contents_references() takes validate_metadata in place of the deprecated validate_crossref and returns metadata_result (codecheck#92).
  • register_clear_cache() takes certificate identifiers to refresh only what is cached about those certificates, keeping the rest of the cache (make clean_cert CERT_ID=... in the register).

Bug Fixes

  • The manifest table in the certificate escapes &, % and # in file comments and shows missing as the size of a missing file, instead of failing to compile (closes codecheckers/codecheck#93).
  • DESCRIPTION requires R 4.4.0, the version that added the %||% operator the package uses throughout, instead of claiming to work on R 4.0.0.
  • The link text of a fediverse account on person and venue pages is no longer turned into citation spans by pandoc (register#217).
  • A persons.csv whose fediverse column has no account in it no longer stops register_render() with “attempt to set an attribute on NULL” (register#217).
  • Writing persons.csv after resolving Wikidata items keeps the file’s other columns and the rows without an item, instead of rewriting it with orcid and wikidata only (register#217).
  • Every render reads the codechecker lists fresh from GitHub, using the cached copy only when that fails, so a copy cached before the fediverse column no longer hides all accounts until the cache is cleared (register#217).
  • Rule CC-MET-001 orcid-format now also checks an ORCID’s check digit, so a mistyped ORCID is caught before it is looked up (register#209).
  • Rule CC-MET-003 orcid-name-match no longer reports a name as different from its ORCID record only because of diacritics, e.g. Grišiūtė and Grisiute (register#209).

codecheck 0.30.0

New Features

  • The CODECHECK validation rules maintained in the register are bundled with the package and readable with codecheck_rules(), codecheck_rule() and rule_severity(), one set per version of the configuration file specification (register#209).
  • New update_codecheck_rules() refreshes the bundled rules from the register, and codecheck_rules_provenance() reports the register commit each bundled file came from and when it was fetched (register#209).
  • New validate_codecheck_yml_rules() validates a codecheck.yml against the rules of the specification version it declares, with one check function per rule and the severity taken from the rule file, so hardened requirements in 2.0 need no separate code path (register#209).
  • New codecheck_spec_version() reports which specification version a codecheck.yml declares, defaulting to the newest when the version node is absent (register#209).
  • New test asserting that every rule in the register’s rule files is either implemented by a check function or explicitly recorded as not implemented, so a new rule cannot be ignored silently (register#209).
  • validate_codecheck_yml() now runs the rule checks instead of implementing the same requirements a second time, so its error messages name the rule that rejected a file, e.g. CC-CFG-026 certificate-id-format (register#209).
  • A rule reported on its own now carries the rule’s description alongside its identifier, in both the per-rule report and the error messages; where several rules are listed at once, only identifiers and findings are shown (register#209).
  • The ORCID and Crossref comparisons, CC-MET-002 to CC-MET-008, are now rules of validate_codecheck_yml_rules(), requesting each record once per file (register#209).
  • validate_codecheck_yml_rules() checks that the manifest files exist in the bundle (CC-BUN-001), and runs a selection of rules with its new rules argument (register#209).
  • New validate_register_rules() checks the certificate identifier sequence, the venue types and the venues of register.csv (CC-REG-002, CC-REG-004, CC-REG-005), and register_check() runs it (register#209).
  • The tagline of the persons page now splits the number of people into those who conducted checks and those who authored checked works, noting that some people are in both groups.

Bug Fixes

  • An unreachable or rate-limited ORCID or Crossref API no longer fails validate_codecheck_yml_orcid() or validate_codecheck_yml_crossref(); the comparisons are skipped instead (register#209).
  • validate_codecheck_yml_orcid() and validate_codecheck_yml_crossref() report at the rule’s severity, so a missing codechecker, a codechecker without a name or a malformed ORCID now stops without strict (register#209).
  • validate_codecheck_yml_orcid() reads ORCID records from the public API only, and its skip_on_auth_error argument no longer has an effect (register#209).
  • validate_codecheck_yml() runs only the rules it enforces, so validating a certificate in register_check() no longer makes requests for the others (register#209).
  • A codecheck.yml that is not valid YAML or not UTF-8 is now reported as a failure of rule CC-CFG-001 rather than as a parser error, and the remaining rules still report what they can (register#209).
  • Fix R CMD check --as-cran failing with an error about man/wikidata_model.Rd and man/WIKIDATA_MODEL.Rd differing only by case, which is not portable across platforms.
  • Fix a batch of R CMD check warnings and notes: non-ASCII characters in R sources, lost braces and an unknown \n macro in the generated documentation, undocumented and mismatched arguments, undeclared global variables, and rprojroot moving from Imports to Suggests (it is used by the shipped certificate templates, not by package code).
  • The certificate template’s Makefile now ships as Makefile.template and is renamed when copied into a workspace, so its GNU make conditionals no longer trip the portability check.
  • License: MIT now names the accompanying LICENSE file, as CRAN requires.
  • Fix pandoc warning Div at temp.md line 8 column 1 unclosed for every certificate page, and align the JSON link with the cards.

codecheck 0.29.0

New Features

  • The register is now published as linked data, on Wikidata and on the CODECHECK Wikibase (register#50). All 132 certificates are items on Wikidata, e.g. CODECHECK Certificate 2025-007, each stating “review of” (P6977) on the work it checked, e.g. the kite wind tunnel measurements - one query lists all 124 pairs. The staging instance https://codecheck.wikibase.cloud holds the same certificates plus the codecheckers and venues that are only ever linked on Wikidata, never created there: certificate 2020-001, the work it reviews, Stephen J. Eglen, GigaScience. It also carries the overview tables and the procedure, readable without R or a login: Project:Data model (every property and item next to its Wikidata counterpart), Project:Certificates (every certificate item beside its register page), and Project:Wikidata export (what an export would do, before it is done). The step-by-step procedure is the vignette “Export the register to Wikidata”.

  • The CODECHECK Wikibase now has a page of ready-to-run SPARQL examples, Project:Example queries, written by bootstrap_wikibase() (register#50). Ten of them query Wikidata itself, where the certificates are, and say which half of its split graph to send them to; three query the instance for the one thing Wikidata cannot answer, who did the checking, and are generated with the instance’s own property numbers and prefixes.

  • The sort state of the register tables is now reflected in the page URL, so a particular view can be linked, bookmarked or cited: ?sort=-check-date is the register sorted by check date, newest first (a leading - means descending, the column name is the header text slugified). A page with more than one sortable table numbers them in document order, sort, sort2, …, e.g. a person page as ?sort=venue&sort2=-check-date. The sort is applied on load and the URL updated on every header click via history.replaceState(), so the back button is unaffected; an unknown or unsortable column name is ignored and leaves the default order. Entirely in inst/extdata/js/table-sort-init.js, which is all a static GitHub Pages site needs.

  • The statistics dashboard now shows the spread of time between a work’s publication and its CODECHECK certificate: a bucketed histogram and a per-certificate beeswarm plot (each dot linking to its certificate page), plus a summary line, computed by compute_annual_stats() from the existing Work publication date/Check date columns. Deliberately framed as a distribution rather than a speed metric - checking a decades-old work is as notable as a fast turnaround.

  • add_openalex_work_fields() now falls back to reading a work’s publication date straight off its own reference URL when OpenAlex has no record for it at all: citation_online_date/citation_publication_date/citation_date HTML meta tags, schema.org datePublished JSON-LD, or - for a reference that is itself a PDF - the PDF’s own Created metadata via pdftools::pdf_info(). New get_page_publication_date_result() and helpers in R/utils_paper_publication_date.R. A partial date (a bare year, or year+month) is resolved to its calendar midpoint rather than the 1st, so it does not skew interval calculations toward “early in the period”. xml2 moves from Suggests to Imports.

  • A render now warns when a certificate’s paper.reference is a plain (non-archived) PDF link: such links carry no machine-readable publication metadata and are prone to rotting (two of the CMMID COVID-19 reports already 404). New warn_if_pdf_reference(), called from add_openalex_ids(); suggests a web.archive.org snapshot instead, matching the guidance added to the community workflow config spec.

  • A certificate’s “Paper details” box now shows its Work publication date (when known) below the abstract, and a work’s own landing page shows it in the metadata panel directly below the DOI. Work publication date joins the columns carried through to a work page for generate_work_metadata_html() (CONFIG$REGISTER_COLUMNS$works$html), same as OpenAlex already was.

  • preview_wikidata_export() splits a batch that would create more items than Wikidata’s rate limit allows per minute into numbered files to paste in turn (register#50). A background QuickStatements run pushes as fast as the API accepts, so a larger batch stopped at the limit and reported the rest as “No success flag set in API result”.

  • quickstatements_submitted() retires the .qs file a batch was pasted from, renaming it with a .submitted suffix, and preview_wikidata_export() removes a batch file once nothing is left to create (register#50). QuickStatements’ CREATE is not idempotent, so a batch file left lying around after its run is one paste away from duplicating every item in it.

  • preview_wikidata_export() no longer tells you to run the works batch first when there is no works batch to run: certificates left without a review of statement once every resolvable work exists name a checked work that has no DOI (register#50).

  • preview_wikidata_export() writes the checked works to wikidata-works.qs rather than wikidata-papers.qs, matching the noun the register uses everywhere else (register#50).

  • register_render() keeps a register repo’s .zenodo.json contributors current with every codechecker named in the register (register#58), crediting people whose work otherwise only appears on their own person page. New build_zenodo_contributors() and update_zenodo_json(). Zenodo’s contributor vocabulary has no “reviewer”/“checker” term, so every entry is typed "Other"; a .zenodo.json that does not exist is left alone. Only the contributors array is touched - the rest of the file (title, creators, licence, …) stays hand-maintained.

  • A certificate’s “Time of check” is shown as a day, not a timestamp: codecheck.yml files record the check with whatever precision the codechecker had at hand, and the hour of a check is noise to a reader (the register tables already showed the day only). The machine-readable variants keep the recorded precision - the certificate’s index.json and its schema.org JSON-LD datePublished carry the time of day in ISO 8601 where the codecheck.yml gave one, and a bare date where it did not, rather than an invented 00:00:00. register-full.json gains a Check time field alongside the day-precision Check date. New internal helpers format_check_date(), format_check_time_iso() and check_time_has_time_of_day().

Bug fixes

  • A venue whose venues.csv entry labels its identifiers “ISSN (print)” and “ISSN (online)” is now resolved on Wikidata; only the bare “ISSN” label was recognised, so those venues exported with no ISSN at all (register#50).
  • A full render is about three times faster (roughly 200 s to 65 s on a warm cache), from three changes: the certificate PDF links, one HTTP request per certificate, are now resolved up front and in parallel by the new prefetch_cert_links() and cached on disk rather than only for the session - resolving them one by one while writing register.json took 75 s of every render, and now takes 2.5 s; the pages of each filter (venues, works, persons, organisations) render in parallel when register_render(parallel = TRUE), taking that phase from 96 s to 38 s; and detect_report_platform() caches where a report DOI resolves to, which also stops a failed request from silently reporting the platform as “doi.org” and moving the statistics between renders.
  • The test suite no longer empties the user’s R.cache root: test_register_check.R called register_clear_cache() against the real cache, so every build_install_test() run left the next full render re-fetching every API lookup (about six minutes instead of one). It now uses a temporary cache root, as test_organisations.R already did.
  • register_render(verbose = TRUE) prefixes every log line with the seconds elapsed since the render started, so a phase that takes minutes between two summary lines is visible in a redirected log. New with_elapsed_log(); forked workers inherit the handler, so pages rendered in parallel are stamped too.
  • On a phone the register tables now hide their least important columns - the Report and DOI links, which each row’s certificate or work page carries anyway, plus Type, ORCID, ROR, Country and the check-type bars - and say so in a note below the table. New add_column_priority_classes() marks the cells in the rendered HTML after add_sortable_th_attributes(); the markup keeps every column and the JSON, CSV and Markdown exports are untouched, so a wider window brings them back.
  • The register pages render properly on a phone: the navigation menu wraps instead of running off the screen, which left “Statistics” and “About” unreachable, and a table wider than the viewport scrolls in its own container instead of widening the whole document, which cut off the right-hand columns with no way to reach them. Also zeroes the body margin, which only the plain 404 and redirect templates ever saw, since Bootstrap sets it on every other page.

codecheck 0.28.0

New Features

  • New bootstrap_wikibase() creates the CODECHECK Wikibase’s own properties and class items from the data model in R/wikidata.R, each carrying a “Wikidata entity” statement naming its counterpart (register#50). A Wikibase mints its own P-numbers, so that mapping is what lets the two sides line up; it is a dry run by default and idempotent, so the instance can be rebuilt from empty.
  • The bootstrap also creates the properties the model uses as qualifiers and references and one item per platform a certificate can be published on, without which those statements were dropped for want of a target.
  • Each statement in the model now names its Wikibase datatype, validated against WIKIBASE_DATATYPES and reported by wikidata_properties() - a property created with the wrong datatype cannot be changed afterwards.
  • The bootstrap writes a generated Project:Data model page listing every local entity next to its Wikidata counterpart.
  • bootstrap_wikibase() also writes Project:About and Project:Copyrights, which the wikibase.cloud hosting policy requires of every hosted instance (register#50).
  • bootstrap_wikibase() brings an entity whose label has drifted from the model back in line, so a rerun converges on the model instead of only filling gaps.
  • The bootstrap retries a write the server asks it to repeat (maxlag, ratelimited, readonly, 429/503) with a capped doubling backoff, and names a contact in its User-Agent (codecheck.contact option).
  • New load_wikibase_register() writes the whole register to the CODECHECK Wikibase as items - people, venues, papers, then the certificates that refer to them - and generates a Project:Certificates index (register#50). Dry by default, and a rerun updates what it wrote instead of duplicating it.
  • New preview_wikidata_export() shows what an export to Wikidata would do without sending anything, writing the QuickStatements batches a person pastes in under their own account and publishing Project:Wikidata export on the Wikibase (register#50).
  • The preview page’s table of checked works gained a “Wikidata certificate” column, so each row names both the item Wikidata holds for the work and the item for the certificate that reviews it (register#50).
  • preview_wikidata_export() resolves against Wikidata through the Action API’s haswbstatement search rather than the query service, which indexes a new item within minutes and sees every graph it might have landed in; method = "sparql" keeps the old path as a cross-check.
  • preview_wikidata_export() refuses to write a batch of CREATEs the edit log says was already submitted while its entities still do not resolve, since a second paste would duplicate every item; force = TRUE overrides it.
  • Checked works are now created on Wikidata as well as mirrored in the Wikibase, carrying their DOI, title, OpenAlex ID (P10283) and authors as P2093 author name strings (register#50).
  • Enriching a work with its OpenAlex ID now also fetches the publication it appeared in and its publication date, as new Paper ISSN and Work publication date columns in register.json, so an exported work states P1433 published in and P577 publication date (register#50). New add_openalex_work_fields().
  • The CODECHECK Wikibase gets an item for every publication the checked works appeared in, not only for the venues in venues.csv - those name the venues that commission checks, and 22 of the 25 publications the works name are not among them, so their works’ “published in” statements had no target (register#50). Wikidata is unaffected: publications are resolved there, never created.
  • register.json and register-full.json carry each checked work’s Paper ISSN, Paper venue and Work publication date, and the Wikidata export backfills them from cache when the register was rendered before those columns existed (register#50).
  • New verify_wikidata_export() checks what actually arrived: whether each certificate item exists, whether it is visible in the scholarly graph, and whether it states review of on the work it checked.
  • verify_wikidata_export(update_register = TRUE) records each certificate’s Wikidata item in a Wikidata column of register.csv, editing the file line by line so its commented-out rows survive (register#50). Nothing else can re-derive the item offline, and the render needs it to link the exported record.
  • Certificate, work and person pages link their Wikidata record, as a signposting describedby link to Special:EntityData in the page head, as Schema.org sameAs in the JSON-LD, and as a wikidata field in the page’s JSON (register#50). A certificate’s item comes from register.csv, a work’s is resolved from its DOI and a person’s from their ORCID.
  • A certificate page’s CODECHECK details box gained a “Certificate on Wikidata” entry linking the exported record, below “Full certificate”; the entry is left out entirely for a certificate that has not been exported (register#50).
  • Person and work landing pages show the Wikidata item in their metadata panel, next to ORCID and GitHub on a person page and next to the DOI on a work page, rather than only in the page’s head and JSON (register#50).
  • The people with Wikidata items are recorded in the register’s persons.csv, read before anything is resolved and written back after, so a clone renders the same links without network access. register_render() and register_render_cert() gained a persons_file argument.
  • Wikidata lookups are cached per identifier, a confirmed “no item” included, so only an identifier nobody has looked up yet costs a request. A failed lookup is not cached, so an outage does not become a lasting “no item”.
  • New quickstatements_write() and quickstatements_submitted() keep an edit log of both halves of the export, the API writes to the Wikibase and the QuickStatements batches pasted into Wikidata by hand (register#50).
  • New vignette “Export the register to Wikidata”, a step-by-step procedure for the whole export, written to be followed literally by a person or an assistant (register#50).
  • New /organisations/ pages list the works authored and the checks conducted by each research organisation’s people, identified by ROR (register#53). An organisation is only credited for a certificate where the person’s ORCID profile placed them there at the time, and every page says so.
  • Organisation pages carry name, type, location and identifiers from ror.org and a logo from Wikidata, and cross-link with an institution venue that shares their ROR.
  • New orcid_rors() and register_ror_coverage() report which authors and codecheckers have a ROR-identified affiliation in their ORCID profile (register#53).
  • Every rendered page carries FAIR Signposting typed links, expressed as HTML <link> elements because GitHub Pages cannot set HTTP Link headers; certificate pages are Level 1 conformant (register#55). New generate_page_signposting() and one generator per page type.
  • Pages with Schema.org metadata now also write it to an index.jsonld next to the page, so the signposting describedby links resolve to a machine-readable document.
  • Register rendering announces the main register, the full export and each filter before rendering them, with a progress bar per page group (per-page timings under verbose = TRUE).
  • The filtered register CSVs are now written under their own log heading instead of silently.
  • The “all persons” table shows the per-venue-type stacked bar for checks conducted again (register#92); index.json carries the plain counts under "Check types".

Bug Fixes

  • curate_zenodo_record() follows a record that Zenodo has superseded with a new version instead of failing with “Not found”: the register stores the report DOI as it was published, and the id in it stops being the editable one once a new version exists (certificate 2023-011).
  • curate_register_zenodo_records() reports records deposited by another Zenodo account as their own category rather than as errors, and names the certificates concerned - the corrections for those are known and correct, they just have to be made by whoever owns the record.
  • A work landing page shows its OpenAlex ID again: the column was missing from the work pages’ HTML column list, so the metadata panel’s OpenAlex row never rendered even though index.json carried the ID.
  • register_check() warns when two certificates share a report DOI. The register renders both without trouble, so it is invisible there, but the report DOI identifies a certificate in the Wikidata export, where two certificates naming one archived record become a single item and the second silently overwrites the first. New check_duplicate_reports().
  • load_wikibase_register() and preview_wikidata_export() refuse to run when two entities share the identifier the model resolves them on, naming the rows that collide, instead of writing one item for both. New check_export_keys().
  • The Wikibase export no longer ends halfway through when a connection times out or drops: a request that never completed is now retried like a 503, rather than propagating as an error.
  • A certificate item’s label includes its number again: mustache cannot address a key containing a space, so {{`Certificate ID`}} rendered as “CODECHECK Certificate” alone. Register columns are now offered to the model’s templates under an underscored alias.
  • A rerun of bootstrap_wikibase() no longer relabels the properties it created, having counted each as colliding with its own label on the instance and planned title as title (P1476).

codecheck 0.27.1

Bug Fixes

  • register_render()’s codechecker counts (the main statistics.json, every venue’s index.json, and the statistics dashboard) no longer come back empty: preprocess_register() only built the underlying Codechecker column when "codecheckers" %in% filter_by, which stopped happening once 0.27.0 dropped "codecheckers" from the default filter_by. The column is now always built.
  • The statistics dashboard’s “Codecheckers” summary card no longer links to the retired /codecheckers/index.html; it links to /persons/index.html.

codecheck 0.27.0

New Features

  • New data model for representing CODECHECK certificates as linked data, in the CODECHECK Wikibase and in Wikidata (register#50). R/wikidata.R holds only the model and its accessors, no HTTP or QuickStatements. A certificate carries P13046 publication type of scholarly work so it resolves in the query service’s cross-graph join, links to the checked paper via P6977 review of rather than P2860, and gets a venue-derived P31 (AGILE reproducibility reviews are typed separately from CODECHECKs). A checked work whose venue is preprint is typed Q580922 preprint rather than scholarly article; both values keep it in the scholarly graph beside its certificate. The certificate identifier is carried as P528 catalog code qualified with P972 catalog → Q141254857 (the register), until a dedicated external identifier property exists. Only certificates are created on Wikidata; papers, people and venues are resolved against existing items and otherwise created only in the mirroring CODECHECK Wikibase. New wikidata_model(), wikidata_entity_kinds(), wikidata_statements(), wikidata_properties(), wikidata_endpoint(), wikidata_creates(), validate_wikidata_model(), wikidata_pending().

  • Register, venue, person, and work tables gain click-to-sort column headers (vendored stupidtable.js); columns holding a link/title rather than a plain value (Report, Work) are excluded. New add_sortable_th_attributes().

  • New /works/<DOI>/ landing page for a checked paper (register#150): lists every certificate that checked it, a metadata panel (DOI, OpenAlex, venues, authors linked to their own person page), Schema.org (ScholarlyArticle + Review graph), and an index.json/register.json API queryable by DOI - a paper with no DOI has no page. New add_work_key(), normalize_work_key(), get_work_metadata_fields(), generate_work_metadata_html()/_yaml(), generate_work_schema_org(), create_all_works_table().

  • New /persons/<ORCID>/ landing page (register#123), replacing /codecheckers/: covers both roles a person can have, showing “Works authored” and “Checks conducted” as two separate tables, with role counts in stats.json and a Person + Review/ScholarlyArticle Schema.org graph. Only ORCID-identified people get a page. New add_person_records(), explode_person_records(), generate_person_metadata_html(), generate_person_schema_org(), create_all_persons_table(). A new docs/404.html (generate_404_page()) redirects a stray /codecheckers/<X>/ link to /persons/<X>/ and, for an unmatched /works/<DOI>, explains that no CODECHECK has been registered for that DOI.

  • register_render()’s default filter_by is now c("venues", "works", "persons") (was c("venues", "codecheckers")).

  • Certificate pages link each ORCID-bearing paper author to their own /persons/<ORCID>/ page (previously only linked if that author was also a codechecker) and link the paper title to its /works/<DOI>/ page when it has one.

  • register_check() warns about two certificates that share a paper title but not a DOI-normalized work key (a likely same-paper duplicate that would otherwise render as two separate /works/ pages, e.g. register#133/#149), and about an ORCID recorded under materially different names or a name recorded under different ORCIDs across the register. New check_near_duplicate_works(), check_orcid_conflicts().

  • register.md is no longer generated for a person page, since its two tables (works authored, checks conducted) can’t be represented as the one markdown table every other filter’s register.md assumes. New CONFIG$FILTERS_WITHOUT_MD.

  • The “CSV source”/“CSV on GitHub” footer links (formerly “searchable CSV”) now show only on the main, unfiltered register page.

  • The top navigation menu is now shown on every page, not just overview pages; each menu item and the logo gained hover text.

  • .navbar-menu .nav-link font size reduced from 2rem to 0.95rem and its breadcrumb counterpart matched to it.

  • Added a ResearchEquals collection link to the page footer, next to the Zenodo community link.

  • Venue landing pages show a metadata panel with website, contact, description, identifiers and logo, from new venues.csv columns (register#84).

  • A venue’s statistics.json is renamed to index.json, since it now carries more than statistics (register#183).

  • venues.csv gained optional logo_url, website_url, policy_url and publisher columns.

  • New statistics dashboard page at docs/statistics/index.html with charts, a venue grid and a publisher table (addresses register#33, register#48). New render_statistics_page().

  • Venue type colours are now pinned by name in CONFIG$VENUE_TYPE_COLORS instead of assigned by the order types appear in the data.

  • JSON/Markdown export links below register tables are now relative rather than absolute codecheck.org.uk URLs.

  • Venue and codechecker register.md exports carry their metadata as YAML frontmatter instead of embedded HTML.

  • Certificates published on ResearchEquals are now audited against the CODECHECK curation policy, including membership in the CODECHECK and Reproducible AGILE collections. New check_researchequals_record() and researchequals_policy_check().

  • register_render() and register_check() gained check_researchequals_policy (default TRUE), the ResearchEquals counterpart of check_zenodo_policy.

  • A “Go to random certificate” button on the main register page jumps to a random certificate instead of the oldest one (register#188).

  • Certificate pages now carry Highwire Press citation metadata, so Google Scholar can index a certificate and Zotero saves it as a report (register#52).

  • Certificate pages describe themselves in their OpenGraph metadata instead of describing the register as a whole, and gained an og:image of the certificate’s first page.

  • A certificate’s title is now read from the platform it is published on rather than constructed as “CODECHECK Certificate <ID>”.

  • A certificate’s Schema.org JSON-LD gained publisher, inLanguage, the record DOI as identifier and the certificate PDF as encoding.

  • register_render() and register_render_cert() gained prune_unavailable_metadata (default FALSE), controlling whether an OpenAlex ID or abstract confirmed unavailable is removed from the rendered output.

  • Codechecker profiles are now looked up in the institutional and Reproducible AGILE codechecker lists as well as the volunteer one, so those codecheckers’ pages show an avatar and GitHub link (register#215). New get_institutional_codecheckers_data() and get_agile_codecheckers_data().

  • A codechecker’s own page shows a metadata panel with their avatar, ORCID, GitHub profile and the venues they contributed checks to (register#74, register#189, register#83, register#75).

  • A codechecker’s stats.json gained a codechecker field with their name, identifiers and contributed venues (register#78).

  • The all-codecheckers table gained a “Check types” column with a stacked bar per codechecker, and their own page the same breakdown as a donut (register#92, register#207).

Bug Fixes

  • A codecheck.yml that cannot be retrieved no longer aborts the whole register render: the affected entry is rendered without the metadata and a warning names the certificate and the error, instead of the render stopping at the first rate limited or unreachable repository.
  • register_render() rejects an empty register, or a from/to selection outside it, up front with a message saying so, instead of carrying rows of NA into the enrichment steps and failing later with an unrelated error.
  • The download of a codecheck.yml from OSF is now retried like the two OSF API calls preceding it, so a rejection from OSF’s file server (e.g. Forbidden (HTTP 403) when the anonymous quota is exhausted) no longer ends the render.
  • The Zenodo DOI badge in the page footer no longer renders its raw URL as visible link text next to the badge image.
  • Codechecker and venue pages’ Schema.org Review entities no longer embed markdown link syntax in @id, url and name.
  • Pages without Schema.org metadata of their own no longer emit an empty <script type="application/ld+json">; they fall back to the generic CODECHECK website metadata as intended.
  • A certificate’s OpenAlex ID and abstract are now looked up once per certificate instead of three times, and a failed lookup no longer drops the field.
  • researchequals_policy_check() reports a missing reference to the checked paper as a warning, not a failure: ResearchEquals references cannot be added after publication.
  • ResearchEquals certificates whose main file is a document written in that platform’s editor are downloaded correctly again (affects certificate 2026-014).
  • Codechecker pages’ Schema.org JSON-LD includes each certificate’s paper title and URL again.
  • render_cert_htmls() no longer emits a spurious file("") only supports open = "w+"... warning on every render.
  • Poppler’s PDF parsing diagnostics are now captured and classified instead of printing raw to the console, and are reported under parallel rendering too.
  • Removed a duplicate definition of convert_cert_pdf_to_png() that existed identically in two files.
  • The venue table’s Report and Paper Title column widths were fixed; Paper Title was the narrowest column despite carrying the most content.
  • The <meta name="generator"> tag now shows just codecheck <version>, without the git commit hashes.
  • A codechecker’s metadata panel no longer disappears entirely for an author-only person with no codecheckers.csv entry; it now falls back to showing their ORCID.
  • Fixed the navbar’s border colour being silently overridden by a later shorthand border-bottom declaration at equal specificity, so it never actually rendered green.
  • Breadcrumbs on a venue type overview page (e.g. /venues/journals/) no longer collapse to a bare, unlinked “Venues”; they now show “Venues > Journals” like every other venue page.
  • render_register_json() and register_update_stats() no longer duplicate their venue/codechecker/work/person stats-building logic, which could otherwise drift between a full render and a stats-only update.

codecheck 0.26.0

New Features

  • register_check() now checks the register newest-first by default (closes codecheckers/codecheck#79). Pass from = 1, to = nrow(register) for the previous oldest-first order.
  • register_check() now also checks the checked repository itself: organisation/group membership, archived status, CODECHECK badge, license, and codecheck topic tag (closes codecheckers/codecheck#25, #75, #14).
  • The allowed GitHub organisations/GitLab groups for check_repository_org() are now configurable via CONFIG$ALLOWED_REPO_ORGS; github.com/reproducible-agile was added to the default list.
  • validate_codecheck_yml() now checks that codecheck.yml is valid UTF-8, starts with the YAML document marker, and has at least one codechecker entry; register_check() now rejects a register containing duplicate certificate IDs (closes codecheckers/codecheck#9).
  • zenodo_policy_check() gained three checks: the certificate ID must appear in the deposit title, a non-standard certificate PDF filename is now flagged, and deposit membership in the Zenodo codecheck community is checked (closes codecheckers/codecheck#20).
  • register_render() no longer accumulates stale header-attrs-* directories in docs/libs, and gained prune_libs() to remove other unreferenced library directories (closes codecheckers/codecheck#89).
  • Zenodo concept DOI checking is extended to also catch an outdated version-specific DOI, via new is_zenodo_latest_version() (closes codecheckers/codecheck#36).
  • Zenodo “concept DOIs” are now rejected in the report field of codecheck.yml (closes codecheckers/codecheck#36).
  • A Quarto certificate template (codecheck.qmd) is now shipped alongside the R Markdown one (closes codecheckers/codecheck#29).
  • Licence correction now keeps the licences already on a Zenodo record when adding the required CC-BY 4.0.
  • Curation findings that need human judgement (e.g. a non-standard title or repository) are now surfaced instead of guessed.
  • New curate_register_zenodo_records() applies mechanical Zenodo metadata corrections across a whole register in one batch.
  • curate_zenodo_record() gained creator_overrides to control how individual creator names are split or kept as an organisation.
  • Every Zenodo-hosted certificate is now audited against the CODECHECK community curation policy during register_render() and register_check(), with findings reported as a summary.
  • New check_zenodo_record() and curate_zenodo_record() audit and correct a single published Zenodo record’s metadata against the curation policy.

Bug Fixes

  • register_check() no longer aborts the whole run when Zenodo rate-limits it. is_zenodo_concept_doi() and is_zenodo_latest_version() now query Zenodo’s plain record endpoint directly instead of through zen4R’s search-based lookups, which carried a much stricter, easily-tripped limit; requests are paced adaptively from Zenodo’s own X-RateLimit-* response headers rather than a fixed guess, with a cli message logged whenever a wait is actually applied, and a 429 is retried automatically.
  • is_zenodo_concept_doi() no longer misreports a rate-limited lookup as a concept DOI.
  • is_zenodo_latest_version() no longer misreports a record’s own latest version as outdated: it compared record IDs against a second, search-based lookup that is not reliably ordered by version, and could return a stale record even when the checked one was in fact current. It now reads the authoritative is_latest flag already present on the record itself.
  • rprojroot is now declared as a package dependency, fixing a missing-function error on a clean install.
  • The ORCID icon on certificate pages is now clickable and links to the ORCID profile.
  • The ORCID icon in the PDF certificate no longer silently goes missing (closes codecheckers/codecheck#37).
  • A creator recorded as an organisation is now reported as information, not an error, in the Zenodo curation policy check (raised in codecheckers/register#205).
  • ORCID name checks no longer fail a fresh, unauthenticated workspace.
  • A fresh workspace now passes strict CrossRef and ORCID name validation out of the box.
  • The register CSS is now written to the output directory rather than the working directory.
  • create_codecheck_files() now reports the correct target folder (closes #87).
  • OSF certificate retrieval now survives an OSF outage instead of failing with an unrelated parsing error.
  • Rendering no longer rewrites docs/libs/PROVENANCE.csv on every run when nothing changed.
  • A failed library download is no longer stored as if it were the library file.
  • ResearchEquals certificates can be downloaded again after that platform’s API change.
  • A register in which no codechecker has an ORCID or GitHub identifier now renders correctly.
  • Zenodo curation no longer truncates titles that need a human’s review.
  • A clear error message is now shown when a Zenodo token may not edit a record.
  • Codecheckers are now recorded as persons, not organisations, in Zenodo metadata.
  • Alternate identifiers are no longer silently dropped when uploading Zenodo metadata.
  • Record titles now match the curation policy (“CODECHECK Certificate <ID>”).
  • A missing or non-DOI paper reference now raises a clear warning.

codecheck 0.25.0

New Features

  • New register-full.json and register-full.csv exports contain the complete certificate metadata, including authors, codecheckers, and abstracts (closes codecheckers/register#57).
  • New register_render_cert() renders a single certificate by ID without re-rendering the whole register (closes codecheckers/codecheck#84).
  • Visiting /register/certs/ without a certificate ID now redirects to the main register page instead of showing a 404 (closes codecheckers/register#166).
  • Rendering output now uses structured, colored logging with progress bars and section headers.
  • Warnings raised during rendering are now collected and shown as a deduplicated summary at the end of the run.

Bug Fixes

  • Fixed stray libs/ folders left behind by parallel rendering.
  • Fixed broken navigation links on venue type pages.
  • Fixed broken navigation and logo paths on venue type pages.
  • Removed the redundant venue label column from venue type pages.

Performance & Scalability

  • Log messages during certificate rendering are now prefixed with the certificate ID, making parallel-render logs easier to follow.

codecheck 0.24.0

Register Enhancements

  • Internal navigation links now use relative paths, enabling local development and testing.
  • The Paper Title column is wider and the Report column narrower, with shortened displayed URLs.
  • Certificate abstract and summary text now display in full, without a scrollable container.
  • Certificate preview images now auto-paginate every 5 seconds, pausing on user interaction.
  • Certificate pages now use the full viewport width.
  • All register outputs (HTML, Markdown, JSON) are now consistently sorted by certificate ID, except featured.json, which stays sorted by check date (closes codecheckers/register#160).
  • Certificate pages now include an interactive citation generator (APA, Vancouver, Harvard, BibTeX, BibLaTeX, RIS) (closes codecheckers/register#82).
  • JavaScript is now centralized into dedicated files, and the citation library is bundled locally instead of loaded from a CDN.
  • Navigation and breadcrumb text sizes were increased for readability.

Performance & Scalability

  • Certificate HTML can now be rendered in parallel across CPU cores via register_render(parallel = TRUE), roughly 5-6x faster on 8 cores.

Bug Fixes

  • Fixed duplicate certificates appearing on the “no ORCID” codechecker page (closes codecheckers/register#153).
  • Codecheckers without an ORCID now appear on the “All codecheckers” page.
  • Fixed a crash from missing check dates in Schema.org metadata.
  • Fixed the certificate JSON page generator not being available for external use.
  • Fixed the citation generator, which previously loaded indefinitely.

codecheck 0.23.0 (2025-11-12)

Register Enhancements

  • Added a navigation header with the CODECHECK logo, and an overview menu on the main register pages.
  • Added breadcrumb navigation to all register pages (closes codecheckers/register#108).
  • Certificate pages now include a machine-readable index.json (closes codecheckers/register#143).
  • Certificate pages now include Schema.org JSON-LD metadata for search engines (closes codecheckers/register#182).
  • Codechecker profile pages now include Schema.org JSON-LD metadata.
  • Column ordering in register tables is now configurable per view (addresses #101).
  • CSV exports now include all fields available in the JSON output.
  • Added sitemap.xml and robots.txt for search engine discoverability (closes codecheckers/register#126).
  • Codecheckers without an ORCID now get their own page, keyed by GitHub username (closes codecheckers/register#130).
  • Overview pages now show build metadata (timestamp, version, commit) in the footer (closes #105).
  • Replaced inline SVG icons with icon fonts for ORCID, GitHub, and Zenodo.
  • Codechecker pages now display ORCID and GitHub profile links (closes #73).
  • Added a Zenodo community link to the footer.

Bug Fixes

  • Fixed a grouping error in the venue table.
  • Fixed a crash from a missing paper title during CSV generation.
  • Fixed icon font, JavaScript, CSS, and logo paths on nested pages.

codecheck 0.22.0.9000

Certificate Page Improvements

  • Codechecker names on certificate pages now link to their register page rather than their ORCID profile (closes #141).
  • Certificate pages now link the venue type and venue name (closes #142).
  • Certificate pages now include venue-based breadcrumb navigation.

Visual Improvements

  • ORCID icons now use the official ORCID brand color.
  • Updated the Zenodo icon to the official logo.
  • Improved icon alignment across register pages.

Infrastructure Improvements

  • Temporary HTML section files are now cleaned up after rendering.
  • Register CSS moved out of inline styles into a dedicated stylesheet.
  • External libraries (Bootstrap, Font Awesome, Academicons) are now managed locally instead of via CDN, with provenance tracking.

Bug Fixes

  • Fixed a rendering error from a missing table details name.
  • Fixed broken venue type hyperlinks.

Venue Configuration and Label Integration

  • Venue information now loads from venues.csv instead of being hardcoded.
  • Venue pages now link to their open GitHub issues.
  • Breaking change: register_render() now requires a venues_file parameter.

GitHub Issue Validation

  • Certificates are now validated against their GitHub issue: a missing issue fails validation, a closed or unassigned issue raises a warning.

ORCID Validation Improvements

  • ORCID validation now handles authentication failures gracefully, with a new skip_on_auth_error option for CI and offline use.

Manifest Rendering Enhancements

  • Certificates can now render more file formats in the manifest section, including TIF, GIF, EPS, SVG, JSON, TSV, and multi-page PDFs.
  • A missing or unsupported manifest file no longer fails the whole certificate; it now shows an error message in the PDF instead.

codecheck 0.22.0

Certificate Automation and Validation

  • Certificate IDs can now be retrieved automatically from the matching GitHub issue.
  • codecheck.yml can now be validated for syntax errors and auto-completed with missing fields.
  • Paper and author/codechecker metadata can now be validated against CrossRef and ORCID records.
  • Placeholder certificate IDs and report DOIs are now flagged with a visible warning in rendered PDFs.

Lifecycle Journal Integration

  • Article metadata can now be retrieved automatically from Lifecycle Journal to populate codecheck.yml (closes #82).

Bug Fixes

  • Fixed Zenodo certificate upload failing due to an API signature change.
  • Fixed ORCID icon links in PDF certificates.
  • Fixed handling of NULL/empty fields during certificate rendering.
  • Fixed Zenodo alternate identifiers not being set per the curation policy.

Enhancements

  • Zenodo certificate upload can now include additional files, including the certificate source (.Rmd/.qmd).
  • A new Zenodo record’s DOI is now written back to codecheck.yml automatically, and the record is submitted to the CODECHECK community.
  • Zenodo metadata uploads now comply with the CODECHECK curation policy (publisher, resource type, identifiers).
  • DOI validation is now platform-agnostic (Zenodo, OSF, ResearchEquals, …).

codecheck 0.21.0

  • Added a pkgdown documentation site.
  • Fixed a typo in abstract text retrieval.
  • Fixed repository URL validation to properly check HTTP responses.

codecheck 0.20.0

  • Renamed fields in register output for consistency.
  • Updated footer contact information and removed individual maintainer references.

codecheck 0.19.0

  • Fixed namespace issues.
  • Added support for specifying a from/to range when rendering the register.

codecheck 0.18.0

codecheck 0.17.0

  • Added support for downloading certificates from the ResearchEquals platform.

codecheck 0.16.0

  • Added support for accessing codecheck.yml in a subdirectory of a GitHub repository via github::org/repo|subpath.
  • Added a TL;DR section to the README for template usage.

codecheck 0.15.0

  • Improved handling of PDF import in the certificate template.

codecheck 0.14.0

  • Added TU Delft Data Champion Centre (DCC) and eLife as venues; removed unused venue entries.
  • Fixed Zenodo hyperlinks in register output.
  • Hardened handling of missing ORCIDs in the codechecker field.
  • Added venue hyperlinks to register pages.
  • Fixed a missing codechecks column.

codecheck 0.12.0

  • Added support for retrieving codecheck.yml from Zenodo records via zenodo::1234567 and zenodo-sandbox::1234567.

codecheck 0.11.6

  • Fixed JSON output including markdown hyperlinks by mistake.

codecheck 0.11.5

  • Fixed imports to reflect API changes in the zen4R package.

codecheck 0.11.4

codecheck 0.11.3

  • Added logging of cache usage for check and render operations.

codecheck 0.11.2

  • Fixed codechecker hyperlink generation when the ORCID ID is missing.

codecheck 0.11.1

  • Fixed certificate pages that have no certificate preview available.

codecheck 0.11.0

  • Added individual certificate HTML pages.
  • Added abstract retrieval from CrossRef and OpenAlex.

codecheck 0.10.1

  • Fixed venue type page hyperlinks.

codecheck 0.10.0

  • Fixed URL links in venue and venue type tables.

codecheck 0.9.0 / 0.8.0

  • Replaced the repository column in the register with platform-prefixed specifications (GitHub, OSF, GitLab, Zenodo).

codecheck 0.7.0

  • Added Amsterdam UMC as a venue.

codecheck 0.6.0

  • Split the register by type (journal, conference, community, institution).
  • Added filtered CSV files by venue and codechecker.

codecheck 0.5.0

  • Added support for multiple output formats (HTML, JSON, Markdown).

codecheck 0.4.0

  • Added codechecker-specific register views.

codecheck 0.3.0

  • Added remote configuration retrieval.

codecheck 0.2.0

  • Added register rendering functionality, supporting multiple venues and codecheckers.

codecheck 0.1.0

  • Initial release: basic CODECHECK workspace creation, certificate generation, and Zenodo integration.

codecheck 0.0.0.90xx

  • Added a NEWS.md file to track changes to the package.