# `8.6 修正.XML` workbook-to-XML conversion

## Metadata

- Date: 2026-08-11
- Status: Converted and independently accepted by the active processor
- Scope: user-supplied `8.6 修正.XML` workbook and generated upload artifact
- Confidence: High for the generated artifact; no original workbook bytes were changed

## Input classification

- The supplied file is an XLSX ZIP package despite its `.XML` filename (`application/vnd.openxmlformats-officedocument.spreadsheetml.sheet`).
- It contains three sheets, with 181 data rows on `Sheet1`; every row has the same `20260806` / `06-08-26` business-date pair.
- The first sheet exposes the flattened Opera `RES_DETAIL` fields needed by the ARR parser, plus separate flattened comment and trace columns.

## Conversion and validation

- Generated a new UTF-8 `RES_DETAIL` XML with the fixed path
  `LIST_G_GROUP_BY1/G_GROUP_BY1/LIST_G_RESERVATION/G_RESERVATION`.
- Mapped the workbook's required reservation fields without changing business values; flattened comment and trace values
  were placed in the parser's existing nested `RES_COMMENT` and `TRACE_TEXT` paths. The workbook remained read-only.
- Output artifact: `8.6 修正-可上传.XML`, SHA-256
  `27b41f4338967ba2b5ae03161fa7eb7b63929a9708c0106e01a740e6c2cc1c91`.
- `xmllint`, strict UTF-8 decoding and ARR processor 4.0.0 all pass. Processor result is `success` for business date
  `2026-08-06`: 181 source rows, 37 rate-code exclusions, 29 duplicates, 115 retained rows, zero price-unmatched
  rows and zero review items. No daily/Finance/database write was performed.

## Boundary

This artifact is a deterministic reconstruction from the supplied workbook, not a byte-for-byte restoration of the
malformed original Opera export. Upload the generated XML as a new task; keep the source workbook and the original
malformed XML as historical inputs.