feat(layout): write line caches for paragraphs holding fields or highlights

CHANGELOG.md

@@ -131,6 +131,11 @@
   지원 밖(`a document without a header part (Contents/header.xml)`)으로 알린다. 줄 배치 정보 쓰기도 그런 문서에는 쓰지 않는다.
 - (실험) 줄 배치 정보 쓰기가 문단 모양이 한글·영문/숫자 자동 간격(`hh:autoSpacing`)을 켜고 한글 음절 바로 옆에 영문 글자나
   숫자가 있는 문단을 건너뛴다. 한/글은 그 사이를 넓히는데 그 폭을 아직 재지 않아, 쓴 줄 나눔이 한/글과 다를 수 있었다.
+- (실험) 줄 배치 정보 쓰기가 글 사이에 필드(하이퍼링크·누름틀·날짜·메일 머지)나 형광펜이 든 문단에도 쓴다. 필드의 시작·끝은
+- (실험) 줄 배치 정보 쓰기가 글 사이에 필드(하이퍼링크·누름틀·날짜·메일 머지)나 형광펜이 든 문단에도 쓴다(구역 첫 문단의
+  필드와 메모 필드는 한/글이 배치 정보 없이 저장해 건너뛴다). 필드의 시작·끝은
+  줄에 자리를 차지하지 않고 줄 위치(textpos)에서 하나에 8글자씩이며, 줄이 나뉘는 자리에 있으면 다음 줄이 그것으로
+  시작한다. 형광펜 표시는 자리도 줄 위치도 차지하지 않는다. 한/글이 그렇게 놓는다. 전에는 그런 문단을 건너뛰었다.
 - FormFit과 쪽 수 추정이 글꼴 폭 표에 없는 가나(히라가나·가타카나)를 그 글꼴의 한글 음절 폭으로 잡는다(함초롬바탕
   972, 한컴 고딕 932). 한/글이 그렇게 놓는다. 전에는 전각(1 em)으로 셌다.
 - 쪽 수 추정(실험, `estimate_pages`)이 탭 정의를 한 번만 읽는다. 전에는 글자 모양을 볼 때마다 머리 정보의 탭 정의 전체를

src/hwpx/layout/line_cache.py

@@ -16,7 +16,9 @@
 Only a paragraph Hancom would lay out from its text alone gets a cache: a body paragraph of text runs in a
 section of one column where no object lets text run beside it. Tabs to the default tab stops, line breaks,
 controls that take no room on a line (a header, a footer, a page number, the section's own definitions in its
-first paragraph) and a hanging bullet or number label in the paragraph's size are fine. Its runs may differ in
+first paragraph), the start and end of a field (a hyperlink, a click-here field, a date, a mail merge field) anywhere in the text but
+in a section's first paragraph, a
+highlight and a hanging bullet or number label in the paragraph's size are fine. Its runs may differ in
 face, each character taking its own run's advance, and in size: each line is then as tall as its largest
 character, unless a larger one past the line's end may be the one that did not fit on it (Hancom writes the line as
 tall as that one), and has no tab; a larger empty run ending the paragraph makes its last line's text as tall. A
@@ -115,6 +117,12 @@
 #: The text children a cached paragraph may hold, as python-hwpx reads them into its text. With a cache Hancom
 #: draws a tab as wide as its ``hp:tab`` says, so a tab gets the width Hancom lays it out at too.
 _TEXT_MARKS = frozenset({"tab", "lineBreak", "nbSpace", "fwSpace"})
+#: Marks in the text that take neither room on a line nor a text position: a highlight's start and end.
+_EMPTY_MARKS = frozenset({"markpenBegin", "markpenEnd"})
+#: The kinds of field whose start and end take no room on a line and eight characters each in a line's text position
+#: wherever they stand (one standing where a line breaks starts the next line). A memo's field, or a field in a
+#: section's first paragraph, Hancom was seen to save without a cache, so those are left for it.
+_FIELD_KINDS = frozenset({"HYPERLINK", "CLICK_HERE", "DATE", "MAILMERGE"})
 #: What Hancom writes on a tab that reaches a default tab stop: no leader, a left stop.
 _DEFAULT_TAB = {"leader": "0", "type": "1"}
 
@@ -320,8 +328,8 @@ def lay_out(room: list[float]) -> list[int]:
         return None
     if mixed and _taller_past_a_break(text, sizes, starts, metrics):
         return None
-    if not _sure_of(text, starts, style, looks) or any(place in starts[1:] for place in controls):
-        return None  # a line starting at a control: on which side Hancom starts it is not known
+    if not _sure_of(text, starts, style, looks):
+        return None
     if any(lay_out([room + shift for room in widths]) != starts for shift in (-_MARGIN, _MARGIN)):
         return None  # a line break this close to a line's end may fall elsewhere in Hancom
     tabs = _tab_widths(text, starts, widths, size, style, looks) if "\t" in text else []
@@ -533,26 +541,38 @@ def _wraps_text(section: Any) -> bool:
 def _control_places(runs: list[Any], first: bool) -> list[int] | None:
     """Where in the paragraph's text each of its controls stands, or None when the paragraph holds anything but
     text and controls that take no room on a line: a header, a footer, a page number and the other page controls
-    (and in a section's first paragraph, the section's own definitions), all of them before the text."""
+    (and in a section's first paragraph, the section's own definitions), all of them before the text, and the start
+    and end of a field anywhere."""
 
     places: list[int] = []
     length = 0
     for run in runs:
         for child in run:
             name = _pages._local(child)
             if name == "t":
-                if any(_pages._local(mark) not in _TEXT_MARKS for mark in child):
+                if any(_pages._local(mark) not in _TEXT_MARKS | _EMPTY_MARKS for mark in child):
                     return None
                 if "\n" in (child.text or "") or any("\n" in (mark.tail or "") for mark in child):
                     return None  # a line break typed into the text, which Hancom writes as hp:lineBreak
                 length += len(_pages._t_text(child))
             elif (name == "secPr" and first or name == "ctrl" and _takes_no_room(child, first)) and not length:
                 places.append(length)  # before the text only: one among it was seen to break the line elsewhere
+            elif name == "ctrl" and not first and _field_mark(child):
+                places.append(length)
             else:
                 return None
     return places
 
 
+def _field_mark(ctrl: Any) -> bool:
+    """Whether *ctrl* holds only the start of a field of :data:`_FIELD_KINDS` or a field's end."""
+
+    marks = list(ctrl)
+    return bool(marks) and all(
+        _pages._local(mark) == "fieldEnd" or _pages._local(mark) == "fieldBegin" and mark.get("type") in _FIELD_KINDS
+        for mark in marks)
+
+
 def _takes_no_room(ctrl: Any, first: bool) -> bool:
     kinds = {_pages._local(child) for child in ctrl}
     return bool(kinds) and kinds <= (_PAGE_CONTROLS | ({"colPr"} if first else set()))

tests/test_layout_line_cache.py

@@ -18,7 +18,7 @@
 HANCOM_LAID_OUT = ["line_cache_tabs.hwpx", "line_cache_line_breaks.hwpx", "line_cache_hanging_indents.hwpx",
                    "line_cache_indents.hwpx", "line_cache_labels.hwpx",
                    "line_cache_mixed_sizes.hwpx", "line_cache_taller_past_a_break.hwpx", "line_cache_table_cells.hwpx",
-                   "line_cache_mixed_faces_tabs.hwpx", "line_cache_space_runs.hwpx", "line_cache_tab_punctuation.hwpx", "line_cache_script_faces.hwpx"]
+                   "line_cache_mixed_faces_tabs.hwpx", "line_cache_space_runs.hwpx", "line_cache_tab_punctuation.hwpx", "line_cache_script_faces.hwpx", "line_cache_fields.hwpx"]
 
 
 def _body(document: HwpxDocument) -> list:
@@ -444,3 +444,23 @@ def test_glyphs_in_a_shape_of_several_script_faces_get_a_cache() -> None:
     assert any(ch.isascii() and ch.isalnum() for ch in texts) and "※" in texts
     for mine, theirs in written:
         assert _segments(mine) == _segments(theirs)
+
+
+def test_a_paragraph_holding_fields_or_a_highlight_gets_a_cache() -> None:
+    # A field's start and end take no room on a line and eight characters each in its text position, wherever
+    # they stand (one where a line breaks starts the next line); a highlight's marks take neither.
+    document = HwpxDocument.open(FIXTURES / "line_cache_fields.hwpx")
+    hancom = HwpxDocument.open(FIXTURES / "line_cache_fields.hwpx")
+    for paragraph in _body(document):
+        for cache in paragraph.findall(f"{HP}linesegarray"):
+            paragraph.remove(cache)
+
+    write_line_caches(document)
+
+    written = [(mine, theirs) for mine, theirs in zip(_body(document), _body(hancom)) if _segments(mine)]
+    with_fields = [mine for mine, _ in written if mine.find(f".//{HP}fieldBegin") is not None]
+    with_highlights = [mine for mine, _ in written if mine.find(f".//{HP}markpenBegin") is not None]
+    assert len(with_fields) >= 10 and with_highlights
+    assert any(len(_segments(mine)) > 1 for mine in with_fields)
+    for mine, theirs in written:
+        assert _segments(mine) == _segments(theirs)