Module refs 

Bibliography input adapters per ADR-0030.

Parses three input shapes into an iterator of Refs with optional entry_key provenance back to the source bibliography:

• Plain refs: one doi:… / arxiv:… / bare-DOI / bare-arXiv id per line, with #-prefixed comments and blank lines tolerated. The existing doiget batch <refs.txt> shape.
• CSL-JSON: a JSON array of entries with id (citation key), DOI, and optionally archivePrefix = "arXiv" + eprint fields. Parsed via the workspace’s existing serde_json — no new dependency.
• BibTeX / BibLaTeX (.bib): deferred to a follow-up slice (the biblatex crate adds cargo-vet exemption churn that is independent of the slice 1 wire shape; users with a .bib library can re-export it as CSL-JSON from Zotero today).

Identifier-pick priority per ADR-0030 D3: doi > arxiv > pmid (PMID adapter parking until the Ref::Pmid variant lands in a later slice; current code carries the rule through without producing a Pmid ref).

Parse-error policy per ADR-0030 D5: a single entry’s failure is captured per-entry and does NOT abort the whole batch. The caller decides whether to skip-and-warn (default) or fail-closed (--strict).

Structs

ParsedEntry

One successfully-parsed bibliography entry.

Enums

Format

Input-shape discriminator per ADR-0030 D4.

ParseError

Why a single bibliography entry failed to produce a Ref.

Functions

detect_format

Detect the input format per ADR-0030 D4.

parse_csl_json

Parse a CSL-JSON document — a JSON array of objects, each with at least an id (citation key) and one of DOI, or archivePrefix

parse_input

Parse text per format, dispatching to the matching shape parser. path is consulted only when format == Format::Auto to drive detect_format.

parse_plain_refs

Parse plain refs — the existing batch input format. One ref per non-blank, non-comment line. entry_key is always None for this shape; plain refs have no citation-key concept.