[airmang]

Addresses #550 as an opt-in.

Background

A .hwp paragraph's PARA_LINE_SEG is the same line layout cache as hp:linesegarray in HWPX. python-hwpx writes new paragraphs without one and leaves the layout to Hancom Office, which lays out a paragraph without a cache when it opens the file. Hancom does take a cache that is present as written, so a cache can only be written when it is exactly the one Hancom would write — a wrong one moves or squeezes text.

Some other HWP viewers place a picture from its paragraph's PARA_LINE_SEG and skip it when there is none, which is what #550 reports.

Change

hwpx.layout.line_cache.write_line_caches(doc) (experimental, opt-in) now also writes the one line Hancom writes for a paragraph holding nothing but a plain picture set as a character:

• vertsize = textheight = picture height + outer margins, or the paragraph's character size when larger
• baseline = 85 % of that
• spacing = the paragraph's percentage spacing over its characters, or over the line when the picture affects the line spacing

A picture with a caption, a shadow or effects, or in a paragraph not spaced by a percentage, is left for Hancom.

Calling it before saving to .hwp carries the caches into PARA_LINE_SEG:

from hwpx.layout.line_cache import write_line_caches
write_line_caches(doc)
doc.save_to_path("out.hwp")

Saving without calling it is unchanged.

Tests

• line_cache_pictures_alone.hwpx: a Hancom-saved document of 120 picture paragraphs (character sizes, picture heights, outer margins, affectLSpacing) between text paragraphs; every cache written matches Hancom's.
• A picture paragraph's cache reaches PARA_LINE_SEG in an .hwp save; a picture with a caption, a shadow or a fixed spacing gets none.

🤖 Generated with Claude Code