Skip to content

Indet markers, alternatives, sensu notes and triage batches B1–B3 - #253

Merged
mdoering merged 17 commits into
mainfrom
fix/issues-2026-11-18
Nov 18, 2026
Merged

mdoering merged 17 commits into
mainfrom
fix/issues-2026-11-18

Conversation

@mdoering

@mdoering mdoering commented Nov 18, 2026 •

Copy link
Copy Markdown
Member

Fixes #251 and #252, plus three related changes to sp.-like markers and alternatives, then batches B1–B3 of the issue triage: #232, #234, #237, #238, #239, #240, #242, #245, #247. One commit per change.

Commit Change ChecklistBank COL XR
e2d9d70 A slash between two lower-case words no longer flags "authorship marked as uncertain" (Acalitus sp. brevitarsus/phyllereus) 303 rows lose the flag 0
a5ab510 agg. with no epithet before it reads like sp.: Acalitus agg. brevitarsus/phyllereus, Taraxacum agg. → informal, SPECIES_AGGREGATE 28 0
0a157e0 Alternatives after an infraspecific rank marker: Sylvia cantillans subsp. inornata/iberiae → informal on the species, rank from the marker; no doubled marker in canonical names (var. var. [unnamed]) 52 names + 26 rows 0
ba0c970 #252: sp. after an epithet is a misspelt ssp. only before a final epithet of 4+ letters; otherwise it opens the phrase of an unnamed species. sp. inc./sp. inq. become a taxonomic note; sp. nov., Authors after an epithet is a nova note 443 + 33 2
e1d55bd #251: dotless sl. before sp. is sensu lato; s.l. sp. / s.s. sp. read like Genus sp. with the note 25 2
ded6e47 #238: the virus check's word boundaries know non-ASCII letters, so the author Vágvölgyi (GV) or Šrna (RNA) no longer makes a name a virus 86 26
cdb1281 #239: a compound epithet in -phages is an eater, no virus (Cuthona poritophages) 20 0
4ff3083 #240: Genus_01 sp_01, Genus-01 are placeholders 0 0
0301a8a #245: Sp. 12, Spec. 1, Sp., Species 1 with no taxon are placeholders (decided: no informal anchor, as #108) 17 0
6c7d512 #234: Phoridae_sp. 873 reads like Phoridae sp. 873 1 0
d210075 #242: Dynestinae sp2, Mus spec2 → informal, phrase sp2 as written 568 2
3c57fdc #232: cf. VERBENACEAE parses as cf. Verbenaceae 0 0
8f5bdae #237: Jonesia luteola et al. 2015 → authors al., 2015, as with a separate authorship 24 0
67b455c #247: a 1–2 digit number after the authors (POW.sp.01, truncated years like Kohmann, 19) is left unparsed instead of dropped; pages and year ranges stay 4,664 11

Decisions worth a look:

  • agg. gets rank SPECIES_AGGREGATE where sp. gets SPECIES; the word aggregate is left alone because it is also an epithet.
  • sp. before a final lower-case epithet of four letters or more stays a misspelt ssp. (about 40 real typos in ChecklistBank, e.g. Oryzaephilus surinamensis sp. mercator).
  • Several words or a qualified epithet before sp. anchor the name on the genus (Prionospio aff. iberian sp. c), the same as for alternatives. This re-baselines one golden row, Euxoa nr. idahoensis sp. 1clay.
  • "Agathidium POW.sp.01” does not contain new warning “author with attached digit” #247 makes 4,664 ChecklistBank rows (3,275 names) partial: mostly truncated years and strain codes split at a dot (Escherichia coli P0301867.5). The number was dropped silently before; the data is no worse, the loss is now visible.
  • catch “et al.” in authorship when primary author is missing #237: a lower-case author before et al. (Arcanobacterium haemolyticum collins et al. 1983, 3 rows) still reads as an epithet; it can't be told from a real one.
  • Pipe-separated name lists with a rank marker (Sparganium erectum ssp. neglectum|Sparganium erectum subsp. neglectum) now follow the alternatives rules; 3 of 36 become OTHER as two genera.

Two build commits:

  • 55a1a92 links the Windows synchronization library into the R binding. The Windows R job has failed on main since 7a55576 (2026-11-15) with undefined reference to WakeByAddressSingle.
  • cc3747a and the next commit move the Java binding to name-parser-api 6.0.1, first the snapshot, then the release. Its NameFormatter (gbif/name-parser 1188287) now writes names as the Rust formatter does: no doubled markers, subg. with the bacterial subgen., and the author of Navicula var. fasciata Grunow.

All workspace tests, clippy and fmt pass locally.

"authorship marked as uncertain" and doubtful were set for any letter, slash, letter, so
alternative epithets (`Acalitus sp. brevitarsus/phyllereus`) and designations (`Barbus sp.
(big mouth/big eye)`) were flagged although they have no authorship. A capitalised word must
now stand on one side of the slash, as an author does (`Smith/Jones`, `Smith/de Vries`). The
rule only narrows Java's: 303 of 332,983 ChecklistBank rows with a slash between letters lose
the flag, nothing else changes.
`Acalitus agg. brevitarsus/phyllereus` became the species `agg` with the infraspecific epithet
`brevitarsus` and the author `phyllereus`, and `Acalitus agg.` lost its marker to become the
uninomial Acalitus. Like `sp.`, an `agg.` with no epithet before it now opens the phrase of an
unnamed taxon anchored on the genus or subgenus, ranked SPECIES_AGGREGATE: `agg.
brevitarsus/phyllereus`, `agg.`. The formatter takes `agg` as the species marker of a phrase,
so the canonical name round-trips. The word `aggregate` is left alone, it is an epithet too
(`Centaurea aggregate Fisch. & C.A.Mey.`, `Rumphella aggregate`). After its species the marker
still makes a species aggregate.

ChecklistBank: 28 rows change, 26 of them a bare `Genus agg.`; COL XR has none.
…s rank

`Sylvia cantillans subsp. inornata/iberiae` became the subspecies `inornata` with the author
`iberiae`, and `Acalitus brevitarsus subsp. brevitarsus or phyllereus` left `or phyllereus`
unparsed. The alternatives check now accepts an infraspecific rank marker before the last
epithet: the name is informal, anchored on the species, ranked by the marker, with the marker
and the alternatives as its phrase (`subsp. inornata/iberiae`), left partial as all
alternatives are. A marker restricted to one code sets it (`serovar` is bacterial).

The formatter no longer writes the rank marker a second time when the phrase spells it out in
any spelling (`ssp.`, `subsp`), which also fixes `Asplenium hookerianum var. var. [unnamed]`
and `Nechamandra subsp. ssp. GP-2014a` (26 ChecklistBank rows).

ChecklistBank: 52 distinct names with alternatives change, 36 of them pipe-separated name
lists, which now follow the alternatives rules (3 become OTHER as two genera); COL XR: none.
`sp.` between epithets was read as a misspelt `ssp.`, and a word before it dropped:
`Prionospio iberian canyons sp. c` became the scientific `Prionospio iberian c`. It is now a
misspelt `ssp.` only before a final lower-case epithet of four letters or more that is no note
word (`Oryzaephilus surinamensis sp. mercator (Fauvel, 1889)`). Otherwise it marks an unnamed
species, its phrase running to the end: on the species after one epithet (`Abies alba sp. c`,
as `sp. 1`), on the genus after several or a qualified one (`Prionospio iberian canyons sp.
c`, `Ramazzottius aff. szeptycki sp. n. 1 ALL-2022a`). The formatter adds no `sp.` when the
phrase holds one.

`sp. inc.` and `sp. inq.` after an epithet, read as the subspecies `inc` before, are a
taxonomic note on the species, which keeps its authorship (`Munidopsis aries sp. inc. (A.
Milne Edwards, 1880)`). A nova note right after the epithet and followed by a comma and authors
now ends there (`Heptapterus carmelitanorum sp. nov., Azevedo-Santos, ...`); after an
authorship the text after the comma stays unparsed as before.

ChecklistBank: 443 rows with `sp.` after a word change, plus 33 `sp. inc.`/`sp. inq.`; COL XR:
2. One golden row re-baselined: `Euxoa nr. idahoensis sp. 1clay` now hangs off the genus.

Closes #252
`Mangelia sl. sp.` became the species `Mangelia sl`, a scientific name that passed every filter
for informal names into the COL extended release. The dotless `sl.` now counts as sensu lato
in the middle of a name too, noted as `s.l.`, and an indetermined species after a sensu note
(`s.l.`, `s.str.`, `s.s.`) stays in the name instead of being left unparsed: `Mangelia sl.
sp.` and `Mangelia s.l. sp.` read like `Mangelia sp.`, with the note. In a separate authorship
(`s. l. sp.`) the marker stays unparsed, as before, since it would become an author there.

ChecklistBank: 25 of 1,404 rows with a sensu note followed by more text change; COL XR: the 2
names of the issue.

Closes #251
…#238)

The virus check matched `GV` in the author `Vágvölgyi` and `RNA` in `Šrna`: its word boundaries
knew only ASCII letters, so the accented letters around them counted as word ends, and the whole
name became an unparsable virus (`Coprinellus fuscocystidiatus L. Nagy, Házi, Papp &
Vágvölgyi`). The boundaries now know every letter. The URL encoding in the issue only seemed to
matter: unencoded, the `&` cut the name before the author.

ChecklistBank: 86 rows (62 names) parse again; COL XR: 26 names, all with Vágvölgyi among the
authors. No measurable slowdown over the 6.4M COL XR names.

Closes #238
`Cuthona poritophages`, a nudibranch eating Porites, was rejected as a virus for its
`-phages`. A clean binomial whose epithet is a compound ending in `-phages` now parses: the
epithet names an eater (`Diplogaster coprophages`, `Labroides phthirophages`). Phages are named
in the singular (`Synechococcus cyanophage`) or carry a strain, and the plural of a viral word
stays viral (`Lactococcus bacteriophages`, `Bacillus phages`).

ChecklistBank: 20 rows (11 names); COL XR: none (the COL names carry an author and a year,
which already saved them).

Closes #239
The word `Genus` is no genus (#211), but only when a space followed it: `Genus_01 sp_01` and
`Genus_01` became the genus "Genus" with the phrase `01 sp_01`, `Genus-01 sp-01` the genus
"Genus-01". An underscore or a dash after it now makes the name a placeholder too. No
ChecklistBank or COL XR name has the shape; the issue's names come from GBIF occurrences.

Closes #240
…#245)

`Sp. 12`, `sp. 16`, `Spec. 1`, `Sp.` and `Species 1` became the scientific uninomial "Sp" or
"Spec", the number dropped without a trace. Like a numbered group with no taxon before it (#108),
they now are unparsable placeholders: there is nothing to hang an informal name off. An epithet
after the marker still makes it an abbreviated genus (`Sp. albiflora`).

ChecklistBank: 17 rows (10 names), among them `Sp` given with the ranks class, order, family and
phylum; COL XR: none.

Closes #245
`Phoridae_sp. 873` made `sp` the epithet and left `873` unparsed as an unlikely year, and
`Phoridae_sp. BOLDABA6985` took the code for an author. As for `Sus_spec_2` (#209), the
underscore before a dotted `sp.`, `spp.` or `spec.` now stands for a space: the name reads like
`Phoridae sp. 873`.

ChecklistBank: 1 row (`Trichosporon_sp.`); COL XR: none.

Closes #234
`Dynestinae sp2`, `Mus spec2` and `Mus sp12` made `sp2` the epithet of a scientific name. A
species marker glued to its number is now the marker of an unnamed species, as `sp.2` is: the
name is informal, with the token as its phrase, kept as written. After an epithet it hangs off
that species, as `sp. c` does (#252): `Mus alba sp2`. The formatter takes the glued marker as
the phrase's own and adds no `sp.` before it.

ChecklistBank: 568 rows; COL XR: 2 (`Gaprivervirus sp18`, `Leucotheavirus sp4`).

Closes #242
`cf. VERBENACEAE` made "Cf" the uninomial and the family its author, while `cf. Verbenaceae` and
a lone `ACANTHACEAE` parse. The leading cf./aff./nr. now also takes a name in capitals, which a
later step writes in title case, so `cf. VERBENACEAE` parses as `cf. Verbenaceae` does. No
ChecklistBank or COL XR name has the shape.

Closes #232
`Jonesia luteola et al. 2015` made `al` an infraspecific epithet and dropped `et`, with the
warnings "Removed: et" and "name was quadrinomial". An `et al.` right after the epithet now opens
the authorship, as it does after an author, so the name parses as it does with `et al. 2015`
given apart: authors `al.`, year 2015. A lower-case author before it (`Arcanobacterium
haemolyticum collins et al. 1983`) is still read as an epithet.

ChecklistBank: 24 rows (23 names); COL XR: none.

Closes #237
A number of one or two digits ending the authorship was dropped without a trace: the `01` of
`Agathidium POW.sp.01`, of `Agathidium alba Smith, 01`. It is no year, so it is now left
unparsed and the name partial. Most are truncated years (`Foissner, Berger & Kohmann, 19`) or the
tail of a strain code split at a dot (`Escherichia coli P0301867.5`). The end of a year range
(`1829/44`) and a page after `p.` or a colon (`Kimsey, 1991: 61`) are left as they were.

ChecklistBank: 4,808 rows (3,400 names) of 1.06M candidates; COL XR: 11.

Closes #247
@mdoering mdoering changed the title Indet markers, alternatives and sensu notes (#251, #252) Indet markers, alternatives, sensu notes and triage batches B1–B3 Nov 18, 2026
The R binding stopped building on Windows on main since 2026-11-15 (7a55576): the Rust std now
calls WaitOnAddress, WakeByAddressSingle and WakeByAddressAll, which live in
synchronization.lib, and the R package linked without it ("undefined reference to
WakeByAddressSingle"). PKG_LIBS of Makevars.win now adds -lsynchronization.
The api's NameFormatter now writes names as the Rust formatter does (gbif/name-parser 1188287):
no second marker before a phrase that spells it out (`Acalitus agg. brevitarsus/phyllereus`,
`Sylvia cantillans subsp. inornata/iberiae`), the botanical `subg.` for a subgenus with the
bacterial `subgen.`, and the author of `Navicula var. fasciata Grunow`. The binding builds
against the snapshot until the api is released; a binding release needs the api release pinned
first.
The api release with the formatter that writes names as the Rust formatter does; replaces the
6.0.1-SNAPSHOT dependency, so the binding can be released again.
@mdoering
mdoering merged commit 1700bbd into main Nov 18, 2026
9 of 11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Dotless sl. for sensu lato becomes an epithet: Mangelia sl. sp. → Mangelia sl

1 participant