https://www.mail-archive.com/lyx-devel@lists.lyx.org/msg198191.html
The root of the issue is in Buffer::reload() called during "Save As". After
loadLyXFile() there, all the insets have been deleted, and therefore the Inset
pointer GuiView::documentBufferView()->cursor().inset() is dangling. Immediately
after loadLyXFile(), reload() calls updateBuffer() which causes a segfault.
While debugging I got other segfaults caused by the same dangling Inset pointer
in Cursor, notably: 1) a trace identical to the second one from
<http://www.lyx.org/trac/ticket/10520>, and 2) a similar segfault in the
critical path after emergency saving (call to inMathed()).
Having to "refresh" by hand the inset pointer cache in CursorSlice is very
unsatisfactory, but there does not appears to be a consistent strategy for
managing these Inset pointers in CursorSlice.
We use the more comprehensive cochineal package if available and fall
back to the crimson package, if not.
Math support is provided by means of newtxmath
Fixes: #9974, #9976
From: "Joel A. Kulesza" <jkulesza@gmail.com>
Date: Mon, 24 Oct 2016 17:37:58 -0600
Subject: [PATCH] Add "Swap & Reverse" to math delimiter dialog
When "Keep matched" is unchecked, a button becomes enabled to "Swap &
Reverse" the left and right delimiters. This is expected to be of use
with line-wrapped equations featuring one or more set of delimiters that
break across the lines. When "Keep matched" is checked, the button is
visible but disabled.
The most common use case is expected to be the user entering a pair of
unmatched delimiters on the first line of an equation (e.g., "(" and
"(None)"), entering the inner text, going to the next line, and
inserting the opposite set of delimiters (e.g., "(None)" and ")").
This button will negate the need to find the correct corresponding
combination. However, it relies on the dialog's memory of the previous
unmatched set.
This change addresses Ticket #10457
-----------
Modifications by spitz to the original patch:
* Only enable the button if an unmatched pair is selected
* Consider l7n when locating the string "(None)"
* Add an accelerator and a tooltip to the dialog
* Simplify the code a bit
Fixes: #8831
This introduces
* a new inset member isPartOfTextSequence() that returns
whether the inset produces something visible at the current position
in the text flow
* an isOpenPunctuation() test that returns whether a character is in the
'Punctuation, Open' unicode class. This is used instead of just checking
for two (Western, ASCII) opening brackets
It also fixes the isChar() and isLetter() value of InsetSpecialChar,
since some types have not been assigned correctly.
* b `british' (``inner quotation'')
* w >>swedishg>> ('inner quotation') ["g" = Guillemets]
* f <<french>> (``inner quotation'')
* i <<frenchin>> (<<inner quotation>>) ["in" = Imprimerie Nationale]
* r <<russian>> (,,inner quotation``)
Rename "french" to "swiss"
Also rename "single/double" to "secondary/primary" ("inner/outer" in
the UI) and "left/right" to "opening/closing". Note that the inset
identifier string is left as is ("s/d" and "l/r")
The current char-based implementation gets increasingly unreadable,
especially if styles are added that do not follow the strict
single-double paradigm.
The QTextLayout handling is terribly slow on Qt 4.8.7, but some
caching has been added in Qt5 that makes it much faster. For some
reason, it is not that slow with Qt 4.8.1.
Caches are introduced for the three following methods
* width(doctring), controlled by CACHE_METRICS_WIDTH. This cache already
existed, but the code has been cleaned up
* getTextLayout, controlled by CACHE_METRICS_QTEXTLAYOUT (disabled by
default on Qt5, which does its own caching). This is used for pos2x
and x2pos and now for drawing of text too. The previous code used a
trivial caching scheme of the last used QTextLayout, but now they
are properly kept in a QCache. Moreover, the cacheEnabled() property
is enabled for these QTextLayout object (not sure what this does).
* breakAt, controlled by CACHE_METRICS_BREAKAT. This is the only user
of QTextLayout which did not have some kind of caching already.
For some weird reasons related to Argument-dependent look-up, the
qHash(docstring) function has to be defined in std namespace, since
lyx::docstring is actually std::basic_string<wchar_t>.
[NOTE: this version has profiling hooks, enabled by commenting out the line
#define DISABLE_PMPROF
that should eventually be removed.]
The xfrac package is the "state of the art" for "split-level" (nice) fractions.
Character replacements look consistent, scale properly and fit in the line.
Fixes#5220.
A dot in the <lang><side><times> string means: keep current value.
This expert feature allows to change arbitrary quote features in one step.
E.g.,
inset-forall Quotes inset-modify changetype f..
=> change all quote insets to French style, maintaining current side
and times setting
inset-forall Quotes inset-modify changetype ..s
=> change all quote insets to single quotes, keeping style and times
inset-forall Quotes inset-modify changetype g.s
=> change all quote insets to German single quotes, keeping left/right
setting
Any idea where to document this?
All typographic quotation marks are supported, if the proper
LaTeX macros (as opposed to ligatures) are used.
So no need for forcing straight quotes.
This is in line with what babel-hebrew suggests.
* Adjoining closing Single + double quote becomes double + single quote
(for English, Swedish and German, LaTeX export as ''').
* French double quotes are converted to << >> in the LaTeX file and to
double inverted question/interrogation marks in the output, if the font
encoding is set to [None] or OT1 but the global default is T1. (test
for lyxrc.fontenc instead of the document-specific fontenc setting in
InsetQuotes.cpp).
* Quote type ignored for LyXHTML: always "English" quotes used.
See #10451
Also output the right one depending on whether the unit is "mu" or
not. Indeed, the two macros are identical, except that \mkern only
accepts unit mu, and kern does not accept it.
With this patch, both macros accept all units, but on writing beck,
the right one is used.
It may happen (for example with on the fly spell checking) that
GuiFontMetrics::pos2x is called with a negative pos value. Add a
special case for this situation.
Fixes bug #10506
This helper method is used to set the inset dimension cache at metrics
time. However this is already done by the metrics code itself
(MathRow::metrics, TextMetrics::redoParagraph), so that there is no
need to do this same work again in a different place.
Also, InsetPhantom::metrics is removed, since it does not do anything
interesting.
This commit comes with a mild risk attached (it might be that some of
these calls were useful after all!), but all of this is fixable.
This helper method is used to set the inset position cache at drawing
time. However this is already done by the drawing code itself
(MathRow::draw, RowPainter::paintInset), so that there is no need to
do this same work again in a different place.
Note that the inset positions are still set in InsetTabular::draw for
each cell.
This commit comes with a mild risk attached (it might be that some of
these calls were useful after all!), but all of this is fixable.
Up to now Inset::dimension was either a helper function to access
CoordCache, or... something else. This created problems to properly
use it.
In particular, the definition of InsetText::dimension created problems
for child classes. Removing this definition (actually renaming it to
dimensionHelper) allows to streamline the code.
* endInsertRow was incorrectly placed before endResetModel, sometimes leading to
bad access by the view before it was informed that the model was empty. In
fact begin/endInsertRows were redundant with begin/endResetModel.
* Empty blocks are not supposed to be shown. This happened when moving the
cursor. A valid block is now selected intead. Due to the design/optimisation
of the model update, this requires a first update with an empty
model. Simplify the old code.
Tweak the algorithm so that a BOX math row element can have some
spacing. To this end, MathRow::before/after do not look at the type of
an element for deciding when to skip it, but rather to its math class.
In the new setting, the spacing algorithm works on all elements, but
skips the MC_UNKNOWN elements as if they were not present. As a
consequence, the two element types BEGIN and END have been replaced by
a single DUMMY (they can be recognized from their class).
To simply the code, add a new `mclass' argument to the
MathRow::Element constructor (default is MC_UNKNOWN).
Macro nesting is now recorded into the macro inset itself. This allows
the ArgumentProxy inset to determine whether it is editable or not by
looking at its macro.
Remove code in the metrics and draw methods of ArgumentProxy: this
code is AFAICS not active anymore, since arguments are linearized into
math rows.
Use Changer idiom to change locally the values of MecticsInfo::base.macro_nesting.
This is a long standing issue, present since the new math macros
inception in version 1.6. It manifests as a display issue when a
macro with optional arguments appears in the optional argument of
another macro. In this case the display is messed up and it is
difficult, if not impossible, changing the arguments as they do not
appear on screen as related to a specific macro instance. It also
manifests as latex errors when compiling, even if the latex output
is formally correct, due to limitations of the xargs package used
to output the macros. Most probably, both aspects have the same
root cause, as simply enclosing in braces the macro and its
parameters solves both issues. However, when reloading a document,
lyx strips the outer braces enclosing a macro argument, thus
frustrating this possible workaround.
This commit solves the display issue by correctly accounting for
macros with optional arguments nested in the argument of another
macro, and circumvents the xargs package limitations causing errors
by enclosing in braces the macros with optional arguments appearing
in the argument of an outer macro when they are output. This means
that when loading an old document with such macros and saving it
again, the macro representation is updated and will have these
additional braces. However, as such braces are stripped by lyx on
loading, there is no risk that they accumulate.
See also this thread:
http://www.mail-archive.com/lyx-devel@lists.lyx.org/msg197828.html
This fixes the "bad keming" of math ERT (in fact wrong metrics) which recently
was worsened by InsetMathChar substitutions and their MathClass spacing.
Also fix a small inefficiency: always prefer:
Changer dummy = (currentMode() == TEXT_MODE)
? pi.base.font.changeShape(UP_SHAPE) : Changer();
over:
Changer dummy = pi.base.font.changeShape((currentMode() == TEXT_MODE)
? UP_SHAPE : pi.base.font.shape());
The former only records and restores a value when the condition is satisfied,
and does not cost anything otherwise.
- Allow using logical values for icon sizes as the standard sizes may
differ among different icon sets
- Do not allow setting sizes smaller than smallIconSize
When the logical sizes differ and the icon set is changed, the correct
sizes are established only after a restart.
Addressing #10481.
This patch adds the new 'needauth' option for converters launching
external programs that are capable of running arbitrary code on behalf
of the user. These converters won't be run unless the user gives explicit
authorization, which is asked on-demand when the converter is about to
be run (question is not asked if the file is cached and calling the
converter is not needed).
The user prompt has a 3rd button so that he/she's not prompted again
for (any converter over) the same document (identified through
buffer->absFileName()).
Two preference options are added:
lyxrc.use_converter_needauth_forbidden disables any converter with
the 'needauth' option, which is meant to force user to an explicit
action via the preferences pane, before being able to use advanced
converters that can potentially bring security threats;
lyxrc.use_converter_needauth enables prompting the user for 'needauth'
converters, or bypasses the check if not enabled, falling back to the
previous behavior.
So, the first option is for maximum security, the second is for
maximum usability.
The current math style should have no impact on \kern1mu. This is another
benefit of a9eb87a8.
Testcase:
\begin{align*}
& \kern1em a\\
& {\scriptscriptstyle \kern1em a}\\
& \kern18mu a\\
& {\scriptscriptstyle \kern18mu a}
\end{align*}
- Adds LFUNs for setting the icon size
- Adds icons sizes to the Toolbars menu
- Uses the Toolbars menu as application context menu
- The context menu can now be user defined in stdcontext.inc
Fixes ticket #10428.
In fact having an extra parameter "bool cond" is no longer useful because it can
now always be emulated with a ternary operator:
Changers dummy = cond ? do_change() : Changer();
* Factor code for easier maintainance.
* Avoid computing metrics several times. This duplication explained the
exponential blowup during the metrics phase for nested fractions (see
b2b87330). This happened in particular when using lyxproofs which heavily uses
nested \dfracs for on-screen drawing.
* Call MetricsBase::changeScript instead of MetricsBase::changeFrac for
\nicefrac and \unitfrac.
Move math style to FontInfo and compute the font sizes for scriptstyle and
scriptscriptstyle according to standard proportions: 0.73 and 0.55.
This is simpler and more accurate. It also fixes the font size of
${\scriptscriptstyle {\textstyle A}}A$ which exposed the limitations of the
previous approach.
There is no reason to reserve pixel space in macros replacement text,
which is not editable. This makes macros more compact and eases the
writing of lib/symbols.
* introduce new InsetMath::drawMarkers and friends that do nothing
when nested inside a macro. This required to move macro_nesting
inside MetricsBase, and to pass MetricsInfo & to metricsMarkers.
* keep track of nesting when drawing rows or macros.
Empty insets should use a minimal amount of space, especially when
they are part of a built-in macro in lib/symbols.
With this change, blue rectangles signal actually editable places.
Empty macros in editable data are shown as grey boxes, but they do not
appear when further nested.
This is done by adding a new type BOX of MathRow::Element object and a
MetricsInfo::macro_nesting that keeps track of macros (and is reset to
0 in editable macro arguments).
* new MathRow class which contains the description of a MathData
object in terms of math class and spacing
+ macros and their arguments used in the MathData object are
linearized (replaced with their contents) so that all math insets
are typeset as a string together. To this end, we introduce a
method addToMathRow to InsetMath and MathData. This method allows
to linearize recursively a MathData object.
+ It is then necessary to set manually the dimension and position of
the macros and arguments.
+ the class class and spacing are computed using the MathClass helpers.
The MathRow data is cached in the MathData object in a bufferview-dependent
way (different dpi for different screens).
* delegate most of the work MathData::metrics/draw to MathRow metrics/draw.
The case of draw is trickier, since many draw() methods rely on their
metrics without any spacing added.
This implements the relevant math typography rules described in the
Appendix G of the TeXbook. More precisely, for each atom
+ the class is computed by implementing rules 5 and 6 of Appendix G
+ the spacing is computed according to the table p. 170
This code is not used at this point.
This done according to the TeXbook. This class replaces the individual
isMathXXX() methods. The mathClass() method (currently unused) is
provided for the following insets:
* InsetMathChar (with a revised list of affected characters)
* InsetMathSymbol: the class is given by the `extra' field
Operators defined in lib/symbols (e.g. \log) are MC_OP
* InsetMathFrac is MC_INNER (except nicefrac and units)
* InsetDelimiters is MC_INNER
* InsetStackrel is MC_REL
* The class of InsetScript is the class of the last element of its
nucleus (yes, it is a hack, but doing it right is more work).
Remove the explicit spacing that was done in the different insets. The spacing
will be reintroduced properly in a forthcoming commit.
* set up a replacement of *, -, and : by the adequate symbols (#9893)
* fix the wrong character selection and operator spacing in \text mode
* hide some internal symbols from the auto-completion.
* Fix spacing in lib/symbols after recent commits about math spacing, as well as
older spacing issues (e.g. \Join).
* InsetMathKern now uses the same em value as other math length commands.
What is nice is that the kerning amount now matches the ones found in the
packages definition (modulo 10mu that lyx currently adds between relations).
Testcase: $\CheckedBox\LEFTcircle\RIGHTcircle\photon\gluon\vcentcolon\dblcolon\Coloneqq\eqcolon\models\hookrightarrow\bowtie\hookleftarrow\Join\APLinv\neq$
This is a long wanted feature, although it does not go all the way to
fix#6604 (private-cut/private-paste).
Additionally, it fixes a crash that can happen when using undefined
branches. This is done by making the action when pasting unknown
branches configurable.
Fixes bug #6570.
An overlong word containing a hyphen could be broken anywhere, instead of after
the hyphen.
Example: compare the line breaking of
aaaaa-aaaaa-aaaaa-aaaaa
with
aaa aaaaa-aaaaa-aaaaa-aaaaa
(with a very narrow window)
This also improves Chinese text in some situations (#10299)
1) Distinguish expanding characters from separators, to fit with Qt's notion of
expanding character which comes from the Unicode std. CountExpanders() is moved
to FontMetrics to fix a discrepancy with the duplicate implementation from
598f7e4a.
2) Make these expanders stretch on-screen proportionally to the em of the font.
If a row mixes large and small text, LyX let us see which spaces are set in the
bigger font.
3) Now that the stretch is defined in ems, add a limit such that an expander
never stretches more than 1.5em to avoid weird and hard to read justified lines.
4) Add a return boolean to setSeparatorExtraWidth for future use.
When \multicolumn{ncol}{align}{content} is parsed and the ncol
parameter is not a numeric value, this parameter is swallowed
and replaced with '1'. Hence, if the file is subsequently saved
a dataloss would occur. With this commit, \multicolumn is not
interpreted when ncol is not a numeric value and is left as is.
See also #10466
It turns out that Cursor::getFont() does not return a fully realized
font in some cases. This patch uses real_current_font instead, which
seems better anyway.
I suspect that all uses of getFont() should be removed, but this is
not code that I know well.
Fixes bug #10478.
This means that, when a string contains high-plane Unicode characters,
the length of a docstring and the corresponding QString will be
different: Qt will encode these characters using several 16bit
characters.
We have additionally to take into account QTBUG-25536, which implies
that sometimes qstring_to_ucs4(toqstr(s)) !=s. It is not clear whether
this bug can be a problem in other places.
Fixes bug #10443.
1. We must always output all (diverging) options, including
default options; if not, default options might get overwritten.
2. Do not output options in \setotherlanguage, since we might have
multiple "other languages" varieties from the same language (such
as naustrian, nswissgerman). And the options are output for the
language switches anyway.
Hence, LaTeXFeatures::getPolyglossiaLanguages() does not have to record
varieties. This was not done correctly anyway, since the map allowed
for one entry per language only.
Math macros can be displayed on screen by providing a different
representation than the one used for latex output. This representation
is actually used by lyx even while it is being updated. This leads to
printing useless error messages on the terminal. For example, a macro
parameter has to be entered as \#1 and, if the macro is already used in
a math inset, lyx prints on terminal the error message "Math parse error:
missing token after \\" as soon as one hits the \ key, followed by
"MathMacroArgument::MathMacroArgument: wrong Argument id: -48" as soon as
one hits the # key. So, this is not a useful information and simply
clutters the terminal output. On the other hand, the input is sanitized
even if one stops input after hitting either \ or #, so that no further
messages are issued. Hence, those error messages are simply pointless.
1/ simplify logic that triggers margin selection painting
2/ fix logic for end of paragraph markers. They are now selected when
there is an end row margin selection.
3/ avoid rounding problems due to int/double conversions.
Also use the exact amount of vertical space TeX adds after a
math display (instead of 1ex) for vertically shifting the box.
We have to use \belowdisplayshortskip here, instead of
\belowdisplayskip, because the math formula is typeset by alone
in a box, and thus there is no following line.
This fixes two particular problems
* with Qt5, it seems that QFontMetrics::width does not return the
correct value for some Arabic text; this patch uses QTextLayout
instead to compute a string width
* Likewise, the undocumented layout flags TextForceRightToLeft and
TextForceLeftToRight do not work with Arabic text; this patch uses
unicode override characters instead.
It might be that the two issues are related. In any case, they do not
happen with latin text where right-to-left direction is enforced. And
they do not happen with Qt4.
Additionally, remove some dead code in GuiFontMetrics::pos2x().
Fixes bug #10436.
LyX assumes that everything in \lyxdeleted is struck out by ulem
and increases the corresponding counter. However, deleted display
math material is struck out using tikz. As we also take into
account the deletion of underlined display math (in order to
properly position such material vertically), we have to take
care that the count is correct.
No newline is written after \begin{document}, such that
the afterParbreak method would return false. This misleads
the code that outputs a display math in an ulem command
to emit a newline command instead of \noindent, causing
latex errors. This occurs only if the math is at the very
start of a document, without anything before it.
- fileformat change
- I could not update the tex2lyx test files because of bug #10440
- maybe the lyx2lyx routine could be improved but since a document can also begin with another layout than "Standard" I could not use the lyx2lyx function put_cmd_in_ert
It should be now possible underlining or striking out any kind
of math inset containing any math construct indigestible to ulem.
While this was already possible for inline math insets, they could
have break if an aligned environment was used, for example.
This is now possible also for diplay math. Even if this can be
nonsensical and not visually perfect, at least no latex errors
should be generated if one tries to.
This should avoid performance problems related to the window update machinery.
Moreover this fixes a crash introduced by 82808fea when closing a file.
Note that GuiWorkArea::Private already had a read_only_ member, but it
was unused.
Also rename LyXVC::vcname() to LyXVC::vcstatus() since it now contains
directly the UI string to be shown.
Font changes are brought inside the \lyxdeleted macro, just before
outputting the latex code for the math inset. The inset writes a
signature before itself and this is checked by \lyxsout for recognizing
a display math. So, the font changes confuse \lyxsout, which also
swallows the first macro at the very start of \lyxdeleted. The result
is that the font changing command is not seen by latex and \sout is also
used to further strike out the formula already striked out by tikz.
This commit makes sure that the expected signature actually appears
just after the opening brace of \lyxdeleted. It also accounts for a
paragraph break occurring just before the math inset, in order to not
introduce too much vertical space, which is noticeable when using
larger font sizes.
LaTeX refuses to break a line when it is empty. But we have to start
a new line here, otherwise the whole displayed equation would be
typeset as it were inline with previous content. The solution is to
put a zero-length space just before the line break. Moreover, this
is the right thing to do, as it simulates the extra space that is
normally added in this circumstance.
Showing deleted display math by enabling "Show Changes in Output" was
only possible with dvi (through dvipost). Although LyX strikes out
such formulas on screen, it was impossible obtaining an output
directly using pdflatex (or other engines producing pdf) because
ulem cannot cope with display math material and gives errors.
The solution is to strike out by ourselves such deleted formulas.
I took into account several options. One of them would produce
an output similar to dvipost (which strikes out each element), but
would have required much more changes in the output routines.
Eventually, I opted for using tikz, which gives a more clean
output (as it requires to simply adding a preamble and a postamble
to the latex code of any displayed math, instead of a mark up
tailored to each particular math construct). The look of the pdf
output is similar to the way LyX strikes out the equations on screen.
Now that there is a signal to inform the controller when the buffer view has
changed, it is no longer necessary to store a perishable pointer to a buffer
view in the dialog view. (Indeed, it was only really useful for checking whether
the new buffer view pointer was different from the old one.)
TODO: other dialogs should be audited for similar hazards.
This enables error reporting for the preamble, provided the preamble is written
using the new InPreamble layouts.
In the future, I find it preferable to deprecate the usual preamble in favour of
InPreamble layouts rather than implementing error reporting for the usual
preamble. This requires some improvements to code editing in the buffer view
first (line breaking behaviour, syntax highlighting).
texstring is a pair of a docstring and a corresponding TexRow. The row count in
the TexRow has to match the number of lines in the docstring.
otexstringstream is an output string stream that can be used to create
texstrings (i.e. it's an odocstringstream that records the TexRow information
and let us extract a texstring from it).
texstrings can be passed around and output to otexstream and otexrowstream,
which produces an accurate TexRow information by concatenating TexRows.
