genome-chromosomes

One place where chromosome codes are interpreted. pg.chr.resolve() is the primitive: it maps codes to a requested scheme and reports what it could not map, without ever aborting or dropping. The two policies on top of it decide what to do with the unmappable remainder:

• pg.chr.canonical() for identity/join paths — normalizes recognized codes and passes unrecognized contig labels through unchanged, so a non-standard contig still matches itself;
• pg.chr.recode() for consumers that need a closed set (PLINK numeric, bigsnpr, PRS-CS) — drops what it cannot map and records it as provenance.

The same file holds the one "chr:start-end" region grammar and the membership predicate built on it, pg.in.region(), because a region names a chromosome and must canonicalize it the same way everything else here does.

Supported output schemes mirror the export formats of PLINK 2.0 (https://www.cog-genomics.org/plink/2.0/data#export): "26" — Always numeric (X/Y/XY/PAR1/PAR2/MT -> numeric codes)

"M" — Autosomal numeric; X/Y/M single character

"MT" — Autosomal numeric; X/Y/MT bare labels

"0M" — Autosomal numeric; 0X/0Y two-character; MT

"chr26" — "chr"+numeric code; PAR1/PAR2 unchanged

"chrM" — "chr"-prefixed; MT -> "chrM"; PAR1/PAR2 unchanged

"chrMT" — "chr"-prefixed canonical label; PAR1/PAR2 unchanged

A code is classified as one of "ok" — recognized: an autosome 1..n.autosomes, X, Y, XY, PAR1, PAR2, M/MT, or the PLINK numeric equivalent, in any of PLINK's spellings and with or without a chr prefix

"missing" — no usable coordinate at all: NA, "", "NA", ".", "-". Never joinable, and always safe to drop

"unrecognized" — a label that is not a standard chromosome but may still be meaningful to the caller: scaffolds and decoys ("GL000209.1"), composite labels ("1_q21"), "0" (PLINK's unplaced marker), junk