Escape XML reserved characters when writing JATS-formatted text to database

[rhigman]

Raw characters <, >, & etc were not previously being escaped when writing JATS fields such as abstract, leading to bugs downstream such as https://help.thoth.pub/#ticket/zoom/363 (where a stray < character used as a "from" symbol caused truncation/malformation of the Crossref XML output).

[ja573]

Issues Fixed

1. normalise_crossref_abstract_jats had no input validation

normalise_crossref_abstract_jats passed content directly to jats_to_ast, which uses html5ever — a lenient HTML parser. Legacy content with raw < (stored before the escape fix) would be silently misinterpreted. For example, "<p>1 < 2</p>" was handled differently depending on whether < was followed by a letter (parsed as a tag) or a space (treated as text), producing silently wrong output instead of an error.
Fix: Added validate_jats_subset(content, ConversionLimit::Abstract)?; before jats_to_ast at mod.rs:395. Malformed XML is now rejected early with a clear error.

2. write_jats_content double-escaped attribute entities

The old code used String::from_utf8_lossy(&attr.value) which reads raw bytes from quick-xml without resolving XML character references. When xml-rs then wrote the attribute via event_builder.attr(), it re-escaped the & in &, producing &.
Fix: Replaced with attr.decode_and_unescape_value(reader.decoder()) in both Event::Start and Event::Empty branches (crossref.rs:336, crossref.rs:383). Entities are resolved before xml-rs encodes them, giving exactly-once encoding.

Test adjustments

Tests for <break/> and nested <p> in the Crossref export were updated from expecting silent normalization to expecting rejection, matching the stricter validation policy. New tests verify exactly-once entity preservation for text (& → &) and link attributes (&/"/' survive without double-escaping).

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 37b4db6c1e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".