Skip to content

sprint-i-3-xml-block-mixed-content.md

Latest commit

 

History

History
84 lines (69 loc) · 3.46 KB

sprint-i-3-xml-block-mixed-content.md

File metadata and controls

84 lines (69 loc) · 3.46 KB
Raw
id sprint-I.3
title XML Block and Mixed-Content Extraction
phase I
status complete/accepted
branch sprint/i-3-xml-block-mixed-content
worktree ../sc-compose-worktrees/sprint/i-3-xml-block-mixed-content
target develop

Sprint I.3 — XML Block and Mixed-Content Extraction

Purpose

Close GitHub issue #193 Gap 1. Extend the known-template XML adapter so a placeholder occupying an element's content can recover a rendered block that contains text and the contract-approved child markup. The implementation must reuse I.2's raw matcher for candidate-value matching and retain XML provenance.

Dependencies and exact targets

  • I.1 accepted XML block/mixed-content contract;
  • I.2 accepted raw-text public matcher seam;
  • crates/sc-composer/src/extract/xml.rs, XML path/source types, and shared extraction report code;
  • CLI and Python extraction parity fixtures;
  • docs/requirements.md, docs/architecture.md, ADR-0013 or the accepted ADR-0012 amendment, and the error registry.

Deliverables

  • Replace the current equal-child-count-only path for the approved full-content placeholder case with a bounded matcher over the rendered inner content.
  • Preserve element occurrence paths, source kind, capture order, and ambiguity diagnostics for block captures.
  • Use a deterministic serialization or text projection exactly as chosen by I.1; do not depend on incidental quick_xml formatting.
  • Support realistic <description>, <references>, and <workflow> blocks, including multiline text and contract-approved inline child elements.
  • Reject multiple placeholders in one block, dynamic names, unsupported control-flow output, mismatched static child structure, and captures that exceed existing input/occurrence limits.
  • Add equivalent library, CLI, Python, positive corpus, and negative-boundary tests.

Acceptance criteria

  • The exact #193 Gap 1 shape recovers the block value without an unsupported child-structure error.
  • Text-only block values, multiline values, escaped entities, repeated element names, and nested approved inline elements have stable evidence.
  • A block placeholder cannot consume a sibling or silently hide a static mismatch; structural paths remain accurate.
  • The public report and diagnostics are identical across Rust, CLI --json, and Python.
  • Existing scalar XML behavior, malformed XML rejection, dotted-expression rejection, input limits, and ambiguity handling remain unchanged.

Required validation

Use the authoritative Phase I validation checklist. Add a focused corpus artifact containing the template, rendered input, expected report, and the requirement/ADR trace for every case. Include a review note explaining why each block is a realistic customer document. The accepted I.3 corpus is evidence/i-3-xml-block-mixed-content.json.

Removal path

If block/mixed-content matching fails QA, remove the XML structural extension and its fixtures while retaining I.2 raw mode and the pre-existing scalar XML adapter. Do not leave a partial child-node fallback enabled.

Out of scope

  • dirty prefixes before the XML root (I.4);
  • unmatched/truncated XML markup, multiple roots, post-root content, or unknown-template identification;
  • loop/branch reconstruction;
  • a second XML-only text matcher;
  • treating every XML text node as a raw block when the template does not have the approved full-content placeholder form.