NEWS

SpaMTPdb 0.99.4

Added spaMTPdbGeneReference() for a versioned HGNC quarterly archive with approved/previous/alias symbols and stable gene identifiers. The reference is retrieved from the official fixed archive and verified locally or in the cache; no changing current-file or live symbol lookup is used.
HGNC versions and provenance are separate from the existing RaMP 3.0.7 registry. Experiment resources remain owned by SpaMTPData.

Added a maintainer-only shared release workflow for annotation database and native experimental files in the existing Zenodo version family. Logical resource ownership, Hub registries and resource versions remain separate.
Shared preparation verifies all files and preserves published registries. Uploads use an explicit new-version or existing-draft target, validate the version family, resume matching files, and reject conflicts without deleting or overwriting remote files. Default execution is an offline dry run.
Removed automatic publication and independent-record creation from the legacy upload entry point. Publishing remains a separate manual action.

Added explicit roxygen2 importFrom declarations for every Imports package and regression coverage for namespace dependency declarations.
Replaced PascalCase resource exports with spaMTPdbResource(), spaMTPdbResources(), spaMTPdbBundle() and spaMTPdbVersion(), synchronized with SpaMTP >= 0.99.3. The RaMP 3.0.7 resource contents and checksums are unchanged.
Verify local files by default, allow explicit unverified development fixtures, and reuse a verified download cache before Hub access, including offline. Bundle defaults now follow the requested resource version.
Updated the SMILES recipe to the native deconvolveSMILES() API and refuse accidental overwrites of an existing structure resource. data-raw entry points delegate to shipped recipes instead of keeping duplicate code.
Made the vignette offline-reproducible and documented the responsibilities and coordinated versions of all three packages.

Version bump only. The Zenodo download fallback shipped inside 0.99.0, so an installation cached by version alone could still hold the pre-Zenodo build that failed with "AnnotationHub does not yet contain ...". Bumping the version lets those caches invalidate.

Initial Bioconductor submission scaffold.
Added versioned AnnotationHub metadata for RaMP 3.0.7 and legacy metabolite annotation resources.
Resource files are published on Zenodo rather than in the Bioconductor Hub bucket, which no longer accepts data from external contributors. Hub records point at the Zenodo copy through Location_Prefix and RDataPath.
Added a verified download and cache path so a resource can be retrieved from its immutable Zenodo URL before the AnnotationHub records are published. Downloads are checked against the recorded size and MD5 checksum.
Added inst/scripts/zenodo_upload.R to deposit staged resources on Zenodo, and taught inst/scripts/stage_resources.R and inst/scripts/make-metadata.R to regenerate the manifest and Hub metadata from a Zenodo record ID.
Moved resource_manifest.csv out of inst/extdata, because AnnotationHubData::makeAnnotationHubMetadata() reads every CSV in that directory as Hub metadata.
Recorded RaMP-derived resources with SourceType: MySQL; the Hub controlled vocabulary has no SQLite term.
Added local-resource support for development before Hub ingestion.
Added an independent, precomputed smiles_features resource for functional groups and positive-, negative-, neutral-, and alkali-adduct priors.