summaryrefslogtreecommitdiff
path: root/comphelper
diff options
context:
space:
mode:
authorJustin Luth <justin.luth@collabora.com>2019-10-22 15:50:19 +0300
committerMiklos Vajna <vmiklos@collabora.com>2019-11-22 10:47:44 +0100
commit334409fbde555a957cd34e295cc27f2c2bf6e194 (patch)
tree24367e1330f55490077fa89ec2cd7550ddab2080 /comphelper
parenta2fbc48afdaf693f1a24e1de5cea21250bfab5ed (diff)
tdf#128153 docx/VML: apply style properties to shape text
This replaces LO 4.3 commit 8766011bccfd0f12f8dd77d2f94eb16e2e8c3471 DOCX import: set document-level font size default based on default para style ...and is needed for tdf#118947, since bogus DEFAULT_VALUEs really hinder determining what a table style should override. Shape text should inherit the font size from the style that is specified. In many cases, it will not be specified, and therefore the default style was appropriate, but in cases where a style IS specified, then of course use that fontsize ... and every other character and paragraph property. HOWEVER, I only added the properties used in vml import for now, and I skipped asian/complex fontnames since VML only handles CharHeight, and not CharHeightAsian/Complex. -note: this does not affect direct formatting - it just sets default value at the shape level, not at the paragraph level. So far I have only looked at DOCX:VML - which satisfies the unit tests. There are other codepaths that lead to PushShapeContext though, and this should be easy to expand to other import situations. I've tried lots of asserts to find unit tests that should be modified, and so far they all seemed to point to VML - although round-tripping doesn't use VML and still requires at minimum the CharHeight property to be overridden, so limiting non-VML to that to maintain backward compatibility, and reduce regression footprint. Since we have to emulate here (since paragraph styles are not supported on Draw text), a perfect solution cannot possibly be found - specifically in cases where multiple paragraphs exist in one shape with different styles applied, or where some pararaphs apply a paragraph property and others do not. Compromise 1: For ambiguous paragraph styles, fallback to the default paragraph style. Rationale: Normally, most styles inherit from default and only change a couple of properties. So MOST of the properties will match the normal style. The chances that the default style will be correct are more likely than that some other random default would be. -note: no existing unit tests were ambiguous Compromise 2: Ideally, each paragraph could report whether it had DIRECT formatting, and the paragraphs could be walked through instead of the shapes. But I don't think that is reasonably possible in this SyncProperties situation. At first I had a grabbag framework setup that monitored when a paragraph property was set, and then skipped applying that property. But I later noticed that the PropertyState for paragraph properties actually did seem to reflect that - which is a better solution if it works properly. Regression potential: -for VML: should be limited to non-charheight properties where the default style sets some weird default, and an ambiguous style sets it back to the program default. Prior to this patch, the default style value wouldn't apply. On the flip side (and more likely scenario), non-ambiguous cases will use the correct value and look more like MSWord, as seen in many existing unit tests that now use corrected fonts. -for non-VML: should be none since I limit it to only CharHeight which was previously emulated by changing the program default. Change-Id: I8f1fb7ed01f990dbf998ebe04064c2645a68e1aa Reviewed-on: https://gerrit.libreoffice.org/81365 Tested-by: Jenkins Reviewed-by: Justin Luth <justin_luth@sil.org> Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
Diffstat (limited to 'comphelper')
0 files changed, 0 insertions, 0 deletions