At this point, there are no simple TextClass objects in the main LyX code, and it is impossible to create them, since the TextClass constructor is protected. Only LayoutFile and DocumentClass objects can be constructed, and for the most part these are constructed only by their respective containers: BaseClassList and DocumentClassBundle. There is an exception: LayoutFile does have a public default constructor, but if anyone knows how to make it go away, please do.
There will be one or two more commits along these lines, but these will be simple renamings. For example, BaseClassList should be LayoutFileList.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23343 a592a061-630c-0410-9148-cb99ea01b6c8
dependent, we need the buffer pointer. In fact we already have a
pointer to the text inset. So as soon as inset know their buffer we
can easily switch to a buffer local word list.
* there might be missing some more places to update the words from a
buffer, like for example after backspace or more important when the
cursor leaves a paragraph.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23244 a592a061-630c-0410-9148-cb99ea01b6c8
Sorry, Andre, but this broke not only the modules stuff but the general handling of TextClasses. I'm not opposed to doing this sort of thing, but it's going to be a little more complicated. I'll do it when I get a bit of time, or I can explain what the issue is here if you want to do it.
I'll separately re-commit some of the cleanup here.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23189 a592a061-630c-0410-9148-cb99ea01b6c8
Also a few cosmetical changes where I xould not resist.
Would be nice if the 'modules people' could verify that their baby
is strill alive...
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23154 a592a061-630c-0410-9148-cb99ea01b6c8
* completion support for mathed
* experimental completion support for text
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23104 a592a061-630c-0410-9148-cb99ea01b6c8
So, when we have an older LyX file, it will look like this:
\begin_inset ERT
status open
\begin_layout Standard
this that
\end_layout
\end_inset
which is now invalid, because ERT uses only PlainLayout. So I had put some code into Text::readParToken, where the layout for a paragraph gets set as it is read:
if (par.forceEmptyLayout()) {
// in this case only the empty layout is allowed
layoutname = tclass.emptyLayoutName();
} else if (par.useEmptyLayout()) {
// in this case, default layout maps to empty layout
if (layoutname == tclass.defaultLayoutName())
layoutname = tclass.emptyLayoutName();
} else {
// otherwise, the empty layout maps to the default
if (layoutname == tclass.emptyLayoutName())
layoutname = tclass.defaultLayoutName();
}
This turns out not to work, because par.forceEmptyLayout() and par.useEmptyLayout() always return false here, because par.inInset() always returns a null pointer, because the paragraph's inset hasn't yet been set when Text::readParagraph() gets called from Text::read() gets called from InsetText::read(). The solution is to set the paragraph's inset when it is created, which means passing a pointer to the various read() routines along the way.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@23057 a592a061-630c-0410-9148-cb99ea01b6c8
- file format change
- new example files and an example address database
- remove ugly hack from scrlttr2.layout and fix bug 4529
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22764 a592a061-630c-0410-9148-cb99ea01b6c8
* src/output_latex.{cpp,h}:
- (latexParagraphs, TeXEnvironment, TeXOnePar, TeXDeeper): pass Text element
instead of ParagraphList. This is necessary to detect whether we are in the main text.
- check if we are in the main text for language switches, if needed.
* src/Buffer.cpp:
* src/insets/InsetEnvironment.cpp:
* src/insets/InsetText.cpp:
* src/mathed/InsetMathMBox.cpp:
- adapt to new latexParagraphs definition.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22666 a592a061-630c-0410-9148-cb99ea01b6c8
Problem: Before macro redefinitions in a child buffer did not see the
parent original definition because an Buffer::updateMacros() call had
updated the redefinition_ flag to "false" after the parent buffer had
temporarely been set to =0 for latex output of the child.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22656 a592a061-630c-0410-9148-cb99ea01b6c8
The main change is to support/ForkedCalls.{h,cpp}. We introduce a static variable IAmAChild and a corresponding accessor. This is set to true in a new fork() method, in the branch taken by the child. (Note: fork() is safe cross-platform, as it just returns -1 if we don't have fork().) This ForkedProcess::iAmAChild() method is then used to protect GuiView::message().
As Abdel has pointed out, there may be other such calls to be protected, e.g., the emission of the Buffer::changed() signal. Those are not addressed here.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22587 a592a061-630c-0410-9148-cb99ea01b6c8
though in this case the warning was "fixed" before it did its job.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22561 a592a061-630c-0410-9148-cb99ea01b6c8
avoidable ~5% overhead when loading the UserGuide. This is an attempt on
rectifying the situation.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22532 a592a061-630c-0410-9148-cb99ea01b6c8
The changes that accomplish this part are in ModuleList.{h,cpp}, configure.py, and the *.module files themselves. This is a format change, and the lyx2lyx is in those files.
By itself, that change would not be major, except for the fact that we do not want the module to be represented in the UI by its filename---e.g., theorems-std---but rather by a descriptive name, such as "Theorems". But that change turns out to be wholly non-trivial. The mechanism for choosing modules was the same as---indeed, was borrowed from---that in GuiCitation: You get a list of modules, and choosing them involves moving strings from one QListView to another. The models underlying these views are just QStringListModels, which means that, when you want to know what modules have been selected, you see what strings are in the "selected" QListView. But these are just the descriptive names, and we can't look up a module by its descriptive name if it's been translated. That, indeed, was the whole point of the change to the new representation.
So, we need a more complicated model underlying the QListView, one that will pair an identifying string---the filename minus the extension, in this case---with each item. This turns out not to be terribly difficult, though it took rather a while for me to understand why it's not difficult. There are two parts:
(i) GuiSelectionManger gets re-written to use any QAbstractListModel, not just a QStringListModel. This actually seems to improve the code, independently.
(ii) We then subclass QAbstractListModel to get the associated ID string, using the Qt::UserRole slot associated with each item to store its ID. This would be almost completely trivial if QAbstractListItem::itemData() included the QVariant associated with this role, but it doesn't, so there are some additional hoops through which to jump.
The new model, a GuiIdListModel, is defined in the files by that name. The changes in GuiSelectionManger.{h,cpp} make it more abstract; the changes in GuiDocument.{h,cpp} adapt it to the new framework.
I've also updated the module documenation to accord with this change.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22501 a592a061-630c-0410-9148-cb99ea01b6c8
- odocfstream: properly fix plaintext output by getting rid of ctor with std::string argument as this can be mixed up with std::ofstream(std::string) ctor where the argument is a file. UTF8 is now the default encoding.
- PreviewLoader::Impl::startLoading(): catch another potential iconv exception.
git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@22332 a592a061-630c-0410-9148-cb99ea01b6c8
