orthonym.validation.stereo_descriptors#

Note

Internal API. Names and behaviour may change between releases.

The descriptor check of a name for its input: is every CIP stereodescriptor right?

A name that OPSIN reads only without its stereodescriptors (a lowercase r/s, a descriptor OPSIN 2.9.0 does not know) is shipped by the default tier’s stereo-layer carve-out after the stereo-stripped parse proved the constitution , and the best-effort tier’s cis/trans offer starts from the same name. Neither may ship a name whose descriptors are wrong, so both ask:func:stereo_block_units for the reading by which the name describes the input. orthonym.namer imports:func:stereo_block_units and stereo_descriptors_verified() as _stereo_block_units and _stereo_descriptors_verified; its callers resolve those names at call time.

class orthonym.validation.stereo_descriptors.StereoBlockUnits(text, span, multiplicity, units)#

Bases: object

One descriptor block of a name and what the accepted reading of it binds (stereo_block_units()).

text: the block as rules.stereochemistry.strip_stereo_blocks returns it ('(1r,4r)'). span: (start, end) in the name, the block with its following hyphen (name[start:end] == '(1r,4r)-'). multiplicity: how many copies of its group the name cites the block for (1 outside a multiplied group). units: the stereo units of the input the reading binds to the tokens of the block, multiplicity per token, as ('a', atom_idx) (a stereocentre) or ('b', bond_idx) (a double bond), in ascending order.

text: str#
span: Tuple[int, int]#
multiplicity: int#
units: Tuple[Tuple[str, int], ...]#
orthonym.validation.stereo_descriptors.stereo_block_units(name, smiles, reduced)#

The reading by which name describes the stereo of smiles, or None when there is none. One:class:StereoBlockUnits per descriptor block of name, in the order rules.stereochemistry.strip_stereo_blocks removes them; the stereo-layer carve-out and the best-effort prefix-order fallback ask only whether the reading exists (stereo_descriptors_verified()), and a caller that rewrites a block reads its span and its atoms here, from the reading that was accepted.

critic-1 / R63 (claims conformance, 2026-09-27): is every stereodescriptor of name right, for a name OPSIN rejects only for its stereo layer?

The stereo-layer carve-out of the validity gate ships such a name after judged the stereo-STRIPPED parse (reduced, rules.stereochemistry.strip_stereo), so the constitution is OPSIN-checked and the descriptors are not. They were shipped unchecked, and some were wrong: ‘(7R,8S)-7-chloro-1-(3,5-dihydroxyphenyl)-2- methylundecyl acetate’ for an input whose stereocentres are the name’s C1 and C2 (C7 is unspecified, C8 is a CH2), and ‘(2S,3R)-3-oxoestr-4-en-17β-yl 3-phenylpropanoate’, whose C2 is a CH2 and C3 the ketone carbon. “NAMING OF STEREOISOMERS” (the Blue Book): “They are preceded by a numerical or letter locant to describe the position of the stereogenic unit when such locants are present” – a descriptor asserts the configuration of the unit at its locant, so a descriptor on a CH2 is a wrong name, not a narrower OPSIN grammar.

Generalises namer._pseudoasymmetric_name_verified (r/s only) to every CIP descriptor. A reading exists iff, for some reading of reduced onto the input:

  1. every descriptor block is a list of CIP code tokens (R, S, r, s, E, Z), each with its locant where the unit has one; anything else (rel, cis, RS, R*) is not verifiable;

  2. every token lands on DISTINCT stereogenic units of the input whose CIP label (perception.stereo.assign_stereochemistry, the centres engine) is that code – a stereocentre for R/S/r/s, a double bond at that atom for E/Z. A located token is read in OPSIN’s own numbering of reduced (the $_AV atom labels of -o extendedsmi), so a locant on a CH2, a heteroatom or out of range finds no such unit; a locant-free token may be any unit with its code. A block cited under a multiplier (bis[...], tris[...]) is cited for every copy of its group: each of its tokens lands on as many distinct units as the multiplicity (_block_multiplicity()), one per copy;

  3. every stereo unit of the input that the CIP labelling labels (the units a token can name, as in 2.) is either one of those cited units or one that the OPSIN parse of reduced already specifies (the retained parent’s implied configuration, a 17β,…): a name that cites a block for fewer copies than the input holds leaves a labelled unit uncited and has no reading;

  4. the input with exactly the cited units made unspecified has the FULL InChIKey of reduced’s OPSIN parse – so each unit the parse specifies is what the rest of the name says. A token that restates a unit the parent already fixes (tropane’s 1R,5S) is kept specified on both sides.

Conditions 2 and 3 fix every labelled unit of the input, and condition 4 the constitution, the parent’s units and any other stereo the InChI sees, so the cited codes fix the units they name and the name denotes the input. Condition 3 is stated on the units, not through the InChI of the reduced input: that InChI drops a lone ring centre whose partner is unspecified, so a leftover centre of a pair of ring centres would not show in the key. A tagged atom that carries no CIP label (one of the two ring centres of cis,trans-1,3,5-trimethylcyclohexane that are not the pseudoasymmetric one) is not a unit a descriptor can name, and is left to condition 4. Fails closed (None) on anything it cannot establish.

orthonym.validation.stereo_descriptors.stereo_descriptors_verified(name, smiles, reduced)#

Is every stereodescriptor of name right for smiles, reduced being the name with its descriptor blocks stripped? True iff:func:stereo_block_units finds a reading; the conditions are stated there.