[Maya-Kid]

Word writes a non-breaking hyphen as an element of its own, <w:noBreakHyphen/>, not as a character in w:t. People type it with Ctrl+Shift+-, and converters write it to keep "2024-01-15" or "010-62345678" on one line. docx_xml_to_text reads the text of w:t and the few elements that stand for a break or a tab. So the hyphen is dropped, and the numbers on either side of it run together.

Problem

A Word file written by python-docx, with non-breaking hyphens in a date and in a phone number, reads like this on dev 2e7ba78:

合同于 20240115 签订，联系电话 01062345678。

The page shows 2024‑01‑15 and 010‑62345678.

• Time extraction no longer sees a date.
• Search for either number finds nothing.

It is the same loss #813 fixed for breaks in table cells and #1019 fixed for breaks on slides.

Change

docx_xml_to_text writes <w:noBreakHyphen/> as a hyphen, in a paragraph and in a table cell, next to the w:tab case. It writes the ASCII hyphen-minus rather than U+2011, so "2024-01-15" reads as the date it is.

Verification

Run on macOS (arm64): Rust 1.98.1; the workspace suite against PostgreSQL 16 with pgvector 0.8.2, with Poppler 26.06.

• a_non_breaking_hyphen_keeps_the_numbers_on_either_side_apart (crates/utopia-ingest/tests/docx_breaks.rs) covers a hyphen inside a run and one in a run of its own.
  • In a paragraph it gives 合同于 2024-01-15 签订，电话 010-62345678。.
  • In a table cell it gives | Sample | 2024-01-15 |.
  • On dev with only the test added, it fails: 合同于 20240115 签订，电话 01062345678。.
• The python-docx file above reads 合同于 2024-01-15 签订，联系电话 010-62345678。.
• cargo test --locked -p utopia-ingest: 124 passed.
• cargo fmt --all --check and cargo clippy --locked --workspace --all-targets -- -D warnings are clean.
• UTOPIA_TEST_REQUIRE_DB=1 UTOPIA_TEST_REQUIRE_PDFTOTEXT=1 cargo test --locked --workspace: 1,345 passed, 0 failed, 6 ignored. That run had this change merged with the spreadsheet-number fix opened alongside it.

[WaylandYang]

Approving. Thanks @Maya-Kid. Reviewed at 1eba5d1: w:noBreakHyphen is written as a plain hyphen in a paragraph and in a cell, so a date and a phone number keep their parts. The ASCII hyphen is the right choice over U+2011.