Skip to contents

plasticdb 0.99.6

Style fixes from the 0.99.5 build report: one example line over 80 characters and one indent that was not a multiple of 4. No behaviour change.

plasticdb 0.99.5

Changes from the first Bioconductor build report (0.99.4). No behaviour change.

  • Tests compare doubles parsed from JSON with expect_equal(), which fixed a last-ulp failure on macOS arm64.
  • Added the biocView SequenceMatching.
  • abort_from_response(), pdb_download() and to_upload() are split into small internal helpers (none over 50 lines).
  • Examples use \donttest{} instead of \dontrun{}; job examples run only when PLASTICDB_API_KEY is set, and write to tempdir().

plasticdb 0.99.4

The repository is renamed VictorGambarini/plasticdb (Bioconductor needs it to match the package name); the site moves to https://victorgambarini.github.io/plasticdb/. No code changes.

plasticdb 0.99.3

Adapted to server 2.90.0 (the contract is re-vendored).

  • job_results() on an hmm_screen job gains the columns profile_accession (character), domain_score, domain_c_evalue, domain_i_evalue (double) and hmm_from, hmm_to, env_from, env_to (integer; 1-based, inclusive). They are NA on DIAMOND hits and on HMM jobs run before server 2.90.0.

  • On HMM hits pident, length, qcov and scov are always NA, never 0, so identity and coverage filters do not apply to them. evalue and bitscore are the full-sequence values; subject_seq_id is the profile name.

  • The HMM TSV download has 8 trailing columns after calibration_status; read it by header name.

  • New HMM test fixtures: a hit with the new fields and an old-style row with all of them null.

  • The README and package documentation state the code’s provenance (AI-assisted), per Bioconductor’s AI and third-party code policy.

  • Security hardening: GitHub Actions are pinned to commit SHAs, Dependabot keeps them current, and SECURITY.md describes private vulnerability reporting.

plasticdb 0.99.2

Adapted to server 2.87.1, 2.88.0 and 2.89.0 (the contract is re-vendored).

  • Behaviour change: proteins_get() now asks for view = "summary" by default (the server’s own default since 2.89.0, which the package used to override with "full"). A summary carries the record’s fields and counts instead of the embedded interactions, indirect_interactions, variants, kinetics and properties lists; pass view = "full" for those, or read them as tibbles with proteins_interactions(), proteins_variants(), proteins_kinetics() and proteins_properties(). as_aastringset() works on either shape.
  • proteins_interactions() and evidence_interactions() take evidence_level ("all", "direct", "indirect") in place of kind, as the server and the new evidence lists name it. kind still works but is deprecated (a plasticdb_deprecated warning, once per session); only evidence_level is ever sent, and different values in both signal plasticdb_argument_error. Code that passed the level by position is unaffected.
  • New top-level evidence lists, flat tibbles of the spine as the release tables have it: interactions_list(), experiments_list() and observations_list(), with interactions_get(), experiments_get() and observations_get() by id. Filters: plastic, include_subtypes, plastic_id, microorganism, protein, evidence_level, paper_id, doi, polarity, evidence_method, substrate_form, measurement_type, has_quantitative; experiments add interaction_id, observations interaction_id and experiment_id. All are pinnable.
  • proteins_list(include_sequence = TRUE) (also variants_list()) adds a sequence column. proteins_sequences() and variants_sequences() use it, live and pinned, so they no longer download the FASTA file as well (they still fall back to it when the server does not return sequences). Each record’s metadata no longer carries a sequence column in mcols().
  • proteins_fasta() answers a pinned release (server 2.87.1; it used to be live only).
  • build_tree() needs an API key on the server since 2.88.0, as the package already required; the contract test now checks the contract has no anonymous option for it.

plasticdb 0.99.1

  • Jobs (R3), mirroring the Python client: submit_job() and the typed helpers annotate_gene(), annotate_genome(), compare_genomes(), annotate_list(), pathway_analysis() (target_plastic), hmm_screen() and build_tree(); the synchronous search_sequence(); wait_job(), job_get(), job_status(), job_results() (hits paged from /jobs/{id}/results, pathway steps, the annotated taxa table, or a tree as Newick text, an organism x plastic matrix and, with ape, a phylo), job_comparison(), job_download(), job_share(), job_shares(), job_unshare(), job_shared(), jobs_list(), job_provenance() and job_replayed().
    • Inputs are file paths (plain or gzip), FASTA text, raw bytes, Biostrings::AAStringSet/DNAStringSets or (taxa tables) data frames; each upload is named after its contents (.gz exactly when it is gzip).
    • Every submission carries an Idempotency-Key (idempotency_key =, or a new UUID), reused on each retry. With an API key, submissions are retried like reads (timeouts, connection errors, 5xx); without one a POST is resent only when it never reached the server, or on a 429. Idempotent-Replayed is reported by job_replayed().
    • New conditions: plasticdb_api_key_required (before anything is sent), plasticdb_job_error, plasticdb_job_failed, plasticdb_job_timeout, plasticdb_results_expired (410 expired) and plasticdb_idempotency_key_reused (422).
  • Adapted to server 2.86 and 2.87:
    • Record redirects are told apart by reason: engineered_variant or merged (with merged_into; kind may be "microorganism", so microorganisms_get(follow_merges = TRUE) follows organism merges). A variant’s accession on a per-protein table is a 404 engineered_variant, now also plasticdb_not_found. Withdrawn records (and variants of a withdrawn protein) are plasticdb_gone with reason "withdrawn" and accession. Every condition has reason.
    • proteins_fasta() and proteins_structure() given a variant’s accession return the variant’s; proteins_get() takes a numeric id.
    • variants_get() and variants_fasta() answer a pinned release.
    • evidence_scope = "enzyme_only" on microorganisms_get() and stats_summary().
    • Every search route_key is a string.
    • The dataset ZIP’s 503 is JSON (reason "archive_unavailable"); a sequence search’s 504 timeout is not retried.
    • A cached pinned download is revalidated with its ETag (If-None-Match; a 304, or an unreachable server, uses the cached copy). Release artefacts are also checked against X-PlasticDB-SHA256. last_response() reports etag, sha256, source_sha256 and replayed.
    • The SDK contract is re-vendored and the fixtures re-captured from 2.87.0.
  • A second vignette, Reproducible analyses with a pinned release.

plasticdb 0.99.0

  • First version, for submission to Bioconductor. Mirrors the Python client (plasticdb-py) phases P1 and P2:
    • plasticdb_client() with the plasticdb.* options and the PLASTICDB_API_KEY, PLASTICDB_RELEASE, PLASTICDB_BASE_URL and PLASTICDB_CACHE_DIR environment variables; retries that honour Retry-After; classed conditions (?"plasticdb-conditions").
    • Release pinning with ?release=, provenance on every result (provenance()), cite() as a bibentry, and a BiocFileCache of answers from past releases (live answers are never cached).
    • Every public data endpoint: proteins, variants, microorganisms, plastics, papers, the evidence tables, search, pathways, statistics, the change feed, downloads and releases. Tables are tibbles; sequences are Biostrings::AAStringSets.
  • Jobs (annotation, genome comparison, pathway analysis, HMM screens, taxonomy trees) are not included yet.