| """Turn a WIPO ``ipc_scheme`` XML file into a flat, ordered list of :class:`IpcNode`. |
| |
| Entries nest physically in the XML (subgroups inside their main group, groups inside |
| the subclass...). Alongside the classifiable entries there are: |
| |
| - ``t`` subsection titles (plain headings between classes, no symbol of their own) |
| - ``g`` guidance headings that span a range of main groups |
| - ``n`` notes |
| - ``i`` catchword indexes |
| |
| None of those are classification targets. Guidance headings are kept as context for |
| the entries that sit under them; the rest are skipped. |
| """ |
|
|
| from __future__ import annotations |
|
|
| import xml.etree.ElementTree as ET |
| from collections.abc import Iterator |
| from pathlib import Path |
|
|
| from .model import IpcNode |
| from .symbols import level_of_kind |
|
|
| NS = "{http://www.wipo.int/classifications/ipc/masterfiles}" |
| TITLE_PART_SEPARATOR = "; " |
|
|
|
|
| def parse_scheme(path: Path | str) -> list[IpcNode]: |
| """Entries in depth-first order, so every parent precedes its children.""" |
| root = ET.parse(path).getroot() |
| return list(_walk(root, parent=None)) |
|
|
|
|
| def _walk(element: ET.Element, *, parent: str | None) -> Iterator[IpcNode]: |
| heading: tuple[str, str] | None = None |
| for child in element: |
| if child.tag != f"{NS}ipcEntry": |
| continue |
| kind = child.get("kind", "") |
| level = level_of_kind(kind) |
| if level is None: |
| if kind == "g": |
| |
| |
| heading = (_title_of(child), child.get("endSymbol") or child.get("symbol", "")) |
| continue |
| symbol = child.get("symbol", "") |
| under_heading = heading is not None and level == "group" and symbol <= heading[1] |
| yield IpcNode( |
| symbol=symbol, |
| level=level, |
| depth=int(kind) if kind.isdigit() else 0, |
| parent=parent, |
| title=_title_of(child), |
| heading=heading[0] if under_heading and heading else "", |
| ) |
| yield from _walk(child, parent=symbol) |
|
|
|
|
| def _title_of(entry: ET.Element) -> str: |
| """Join the title parts, dropping cross references such as "(edge trimmers A01G 3/06)". |
| |
| Cross references point *away* from the entry; embedding them would attract exactly |
| the texts that belong elsewhere. |
| """ |
| body = entry.find(f"{NS}textBody/{NS}title") |
| if body is None: |
| return "" |
| parts = [] |
| for part in body.findall(f"{NS}titlePart"): |
| text = part.find(f"{NS}text") |
| if text is not None: |
| value = "".join(text.itertext()).strip() |
| if value: |
| parts.append(" ".join(value.split())) |
| return TITLE_PART_SEPARATOR.join(parts) |
|
|
