PDB and mmCIF extraction

results = crabwurcs.extract_pdb_file("model.cif")
for result in results:
    print(result.glycan.to("wurcs"))
    for residue in result.residues:
        print(residue.node_index, residue.chain, residue.sequence_number)
    for warning in result.warnings:
        print(warning)  # residue 2uA A:4: named α-L-IdopA, coordinates say α-L-GulpA2S (C2)

Coordinates decide

Every residue is searched for a monosaccharide ring (five or six atoms, one oxygen) in the bond graph, whatever the residue is called. From the coordinates alone:

• the carbon numbering comes from the ring graph: the ring carbon next to the ring oxygen that carries an exocyclic heteroatom is the anomeric carbon, C1 precedes it in a ketose, and the numbering walks the ring away from the ring oxygen and on along the exocyclic carbon chain;
• every stereocentre's configuration is the sign of the signed volume of its three heavy-atom substituents, ordered by the ring graph alone (previous carbon, next carbon, exocyclic heteroatom; the ring oxygen takes the place of the previous carbon at the anomeric carbon). That gives D/L, the parent sugar, deoxy positions and uronic acids;
• the anomer compares the anomeric centre with the anomeric reference atom (IUPAC 2-Carb-6.2);
• substituents (N-acetyl, N-glycolyl, N-sulfate, amino, O-sulfate, phosphate, phosphocholine, O-methyl, O-acetyl, fluoro) and glycosidic linkages come from what the ring atoms are bonded to, in the same residue or in another (GLYCAM keeps sulfates in SO3 residues).

Bonds are perceived from distances within each atom's valence (bonds inside a residue first, then explicit CONECT/LINK records, then contacts between residues), so a clash between residues does not break a ring. Metal ions are never bonded. wwPDB Chemical Component Dictionary coordinates of 49 components fix the conventions in the test suite.

Names are hints

A residue name fills in only what the coordinates cannot show: atoms missing (a crystal structure without hydrogens: an absent O4 or N-acetyl group), a centre too close to planar to read (normalised volume below 0.15), or a residue whose ring atoms are missing (read from its name, as before). Names come from the pinned wwPDB CCD component table, concrete registry names and GLYCAM residue codes (upper case D, lower case L).

Every disagreement between a name and the coordinates, and every part the coordinates could not decide, is a warning. policy chooses what decides:

policy	on a disagreement
coordinates (default)	the coordinates decide; a warning
names	the name decides where it says something; a warning
strict	PdbError listing the disagreements

Warning kinds: name-contradicts-coordinates, missing-atoms, distorted-geometry, unreadable-geometry, no-ring, unrecognized-group. ExtractedGlycan.readings gives, per residue, the WURCS UniqueRES read from the coordinates and the one its name describes, its ring and backbone atoms, and the Fischer side and normalised volume of every centre.

In Rust, read_glycans_from_file(path, &PdbReadOptions::with_policy(...)) returns a PdbReading with glycans, warnings and residues; the extract_glycans_* functions read with the default policy.

Records read

PDB text is read from the records the extractor uses: the ATOM/HETATM records of the first model and the CONECT and LINK records. Free-text records (such as banner REMARK lines) and later models are skipped. Of an mmCIF file, the first model is read. Of alternative locations, the atoms without one and those of the first are read.

The attachment site is the residue an explicit CONECT or LINK record bonds the glycan to (the bond at the root's anomeric atom first; never a substituent residue such as SO3), otherwise the root residue. Insertion codes are retained. PDB/mmCIF writing, conformer generation, and coordinate optimization are intentionally outside the 0.3 scope.