There is function `NK_INTERN int nk_to_upper(int c)` in `UTIL` section. Implementation is almost the same to `NK_INTERN int nk_to_lower(int c)`. But only `nk_to_lower` used in Nuklear's code. To remove this warning we can:
1) Comment i.e. remove `nk_to_upper` from code. But it can break someone's code, if he used it in module with `NK_IMPLEMENTATION` defined.
2) Hide it under define. Same.
3) Just use it somewhere in code.
I replaced two calls of `nk_to_lower` to `nk_to_upper`. It removes the warning, code works the same. The bad side - complication of the code. "Why `nk_to_upper` is used here, but not `nk_to_lower` as one line up?"
Maybe there exists better solution? Or just add the comment about this line to code?
When NDEBUG defined:
nuklear.h: In function 'nk_popup_begin':
nuklear.h:20753:22: warning: variable 'panel' set but not used [-Wunused-but-set-variable]
nuklear.h: In function 'nk_nonblock_begin':
nuklear.h:20850:22: warning: variable 'panel' set but not used [-Wunused-but-set-variable]
Added another version of `nk_tree` that does not store internal state
to hold the tree state (minimized/maximized). Main advantages is
possibily less memory consumptions and more user control.
In addition a new set of button overloads were added to directly pass
in a button style struct which allows an easier way to have custom
button visuals for specific buttons.
Added additional `nk_group` version which uses an external scrollbar
offset instead of internally managing the memory. In addition
`nk_list_view` is introduced which allows drawing huge lists inside a
`nk_group`(#269).
I simplified a lot of API calls by pulling panel memory management
inside the library. All API calls which previously required a panel
as parameter are now handling their panel themself.
Previously every widget border was drawn by two overlapping rectangles
to limit the number of primitives required to be supported by any
potential backend. The biggest problem from using this approach
is drawing overhead and no real way to get widget transparency. So
I changed border drawing to use stroked rectangles instead of a
second filled rectangle drawing call.
using memcmp and decide whether to draw a new frame or not:
* added NK_ZERO_COMMAND_MEMORY macro
* when NK_ZERO_COMMAND_MEMORY is defined, inside
nk_command_buffer_push function there is an additional call
to NK_MEMSET after allocating memory for new drawing command
For more information please look inside the changelog. This is more of
an internal fix and hopefully does not break any existing code. I also
removed `nk_strfmt` from the API and made it internal only.
If you had any kind of popups like for example comboboxes inside a
deeper panel structure using nk_group, then `nk_xxx_close` would only
unblock input for the first level and not the whole panel stack. This is
now fixed and works correctly.
This is a breaking change which fixes bad closing behavior in
non-blocking popups. If the maximum size of a non-blocking popup was
not reached and you clicked to close the popup under it could happen
that it wouldn't be closed. This is now fixed in combination with
more popup control by allowing to set both the maximum height as
well as popup width.
Since it is not as clear on how to use the menubar correctly as I
thought added an assert into `nk_menubar_begin` with explainations on
how to use it correctly and stop wrong ussage.
Since it was not as apparent as I thought it is I added some
documentation warning to set the same flags each time you
include "nuklear.h" except `NK_IMPLEMENTATION` which is only
allowed to be defined once in *one* source file.
Functions `nk_combo`, `nk_combo_seperator`, `nk_combo_string`,
`nk_combo_callback`, `nk_combobox`, `nk_combobox_seperator`, `nk_combobox_string`,
`nk_combobox_callback` all calculated the the height of the combo box
panel wrong and forces a scrollbar even if it was not required. The
calculation is now fixed and should produce the right result.
This is basically a hack to fill the gaps. I have absoultly no idea why
I get some of these gaps but hopefully it is fixed and does not come
back to hunt me in the future.
I removed the default vertex type and instead create an API to allow the
user to specify the vertex struct layout. Of course you are still
bound on what the library actual has on vertex data but the type and
sequence of the data is now configurable. This commit is quite
experiental and could introduce some bugs, but so far it look fine.
This commit fixed some bugs from the last commit as well as some general
changes and fixes of some old bugs. For overview of all changes please
look inside `CHANGELOG.md`.
This is a rather experiemental commit to remove some internal complexity
that build up over time handling panel padding and border. There still
could be some bugs but I tried my best to keep everything working.
Previously it was only possible to scroll windows and groups by mouse.
No it is possible to scroll popups, as well as nonblocking popups like
menus, contextuals and comboboxes. In addition I fixed a drawing bug
in dynamically growing panels.
To account for hash collisions which can be quite nasty I added
the string representation of the window name for additional
assurance. I also added an assert to check and alert if two windows
have the same name.
It is now possible to temporarily push some style values like color,
style items, font, ... into a stack modify the value and as soon as
you are done just pop the changes. Main benefit is less boilerplate
code to manually save and restore configuration properties.
SDL iOs implementation provides key press and release in
one frame and while nuklear previously was keeping track
of transistions it did not check it. It hopefully should now be
possible to correctly interpret up and down key in the same frame.
Edit widget width `NK_EDIT_SIG_ENTER` flag now keep their active
state after enter was pressed. Main reason is to be able to keep
typing after commiting.
`nk_window_is_closed` previously checked if the window is hidden
instead of checking if it is closed. This behavior is now fixed and I
added an additional function `nk_window_is_hidden` to check if a
window is currently hidden.
Nuklear now has three different property versions for int, float and
added by this commit a double version. Internally an additional change
happend. Now the type of the property is actually taken into account
and floating pointer errors due to casting are less of an issue.
Since floating point to and from string conversion is a total pain
and really complex it is really hard to provide a realtive simple
non-standard library solution. But to solve both problems I now added
a way to overwrite the library intern float to string conversion
implementation if you really need more precision.
Since nuklear internal string to number conversion is quite imprecise,
and a correct solution is way to complex (I dare you to look at the
implementation of strtod, dtoa, ...) I now provide a new define to
overwrite the library function with either your own or the standard
libray implementation.
Under certain circumstances it is possible that because of floating
point error `nk_widget` will mark a widget as read only if used in
combination with `nk_layout_row`. The fix is rather simple in this
case since the additional floating point precision is not really needed
it is possible to just cast to int and remove the problem all togther.
There is now a difference between hiding a window with
`nk_window_show` and closing a window by clicking the close button in
the window header or calling `nk_window_close`. Only in the case of
activly closing a window will be removed at the end of the frame
besides getting removed if not updated inside the frame.
To support both string formats with string combined with length as
well as zero terminated strings I added a zero terminated string
version to `nk_edit_xxxx` with `nk_edit_string_zero_terminated`.
Fixed scrollbar auto hiding behavior to also work correctly if
the window is not hovered or manipulated in any way.
In addition I fixed a small panel padding bug in `nk_layout_widget`
for all panels outside windows which require a test on
`layout->flags` instead of `window->flags`.
The dynamic immediate mode layout API with `nk_layout_row_begin`,
`nk_layout_row_push` and `nk_layout_row_end` caused an infinite
scrollbar bug by wrongfully calculating the required widget sizes.
This bug is now fixed and everything is correctly calculated.
Previously I packed standard library variable arguments as well as
file operation into one define `NK_INCLUDE_STANDARD_IO`. This caused
some problems and I split the define into two. So if you want the old
behavior you need to define `NK_INCLUDE_STANDARD_IO` as well as
`NK_INCLUDE_STANDARD_VARARGS`. If you only want file operations but
not variable arguments only define `NK_INCLUDE_STANDARD_IO` and
if you only want varargs but not file operations please only define
`NK_INCLUDE_STANDARD_VARARGS`.
If you move the cursor to fast while scaling a window there is a
chance the mouse will miss the scaler and the scaling operation will
halt. Dragging in itself is quite hard but I implemented the hack I
use for property, window moving and slider for the window scaler and
scaling hopefully now works correctly.
This is quite an old bug and a result of being nearly stateless.
If you move/scale a window and another window is behind it the other
window will be acitvated and the move/scale operation will be stopped.
To fix this I now require no mouse down for activating by hovering.
I hope this does not cause any problems and fixes the problem for good.
I had a long standing bug inside nuklear which prevents you from
swapping out a font while the code is being build. This bug is now
fixed but the fix requires _ALL_ `nk_user_font` data structures
provided by `nk_style_set_font` to be persistent.
A previous push request added some macros into the header file and
I missed that it was outside the header include guard. This should
now be fixed and not cause problem if included multiple times in
different location.
Window scrollbar will if enabled by window flag hide if no user
interaction happens for a set time duration.
In addition the internal window/table data structs are now located
inside the header part of nuklear to remove the need to allocate a
pool and therefore simplify memory management.
Hopefully I finally fixed the hole inside combobox popup which only
occur if you set the window border size greater one. In addition
I fixed some subpixel drawing artifacts that come up if you use
antialising in combination with vertex buffer output.
I completely forgot to add the edit scrollbar style and
had a incorrect scrollbar bounds calculation. Both bugs
should now be fixed. Furthermore the edit ROM flag know
prevents any manipulating input instead of all input.
So it should be possible to move or copy inside the edit
field but not to modify the content.
For gui elements like tree tabs, tree nodes and windows themselves,
you can repeatedly click the same "minimize" region and it keeps
toggling that element. This was not the case with "popup" type of
elements. This commit slightly changes how the popups close once
they are already open.
Like before, the user can close the popup by just clicking outside
the popups region. Now the user can also close the popup by clicking
the "header" region of the popup.
Previously if a dynamic back buffer is full and reallocated then the
current allocation alignment will be alignet at the direct memory
address instead of memory address minus size. This is now fixed and
alignment should work correctly.
It is now possible to use mouse scrolling as well as shortcuts for
windows as well as groups. But there is still some unwanted behavior
if you scroll the window and while scrolling hover over a group then
the group will be scrolled and not the window. I will look into fixing
it.
This commit extends nuklears scrollbar with some basic shortcuts.
Included are page up/down as well as jump to beginning and
end of window. The last scrollbar change is page down/up by mouse
click in empty scrollbar space.
Text was previously wrongfully software clipped if text was only
partially visible. The wrong behavior is now fixed and text should
be clipped only if the text is completely outside the clipping rect.
Surprising turn out, the C++ standard comitee consists of a bunch of
retards invalidating completely fine code and requesting super
idiotic work arounds for their new versions. I have no fucking idea
why these spaces are required and honestly I don't even want to know.
Whatever the reason it should now compile fine with C++11 until the^
next C++ version will probably invalidate it again. To a better
future I can't wait.
Previously if you used movable or scalable groups you had to add
the group border width to the panel to access the updated position
or size. Nuklear now takes control of the process and the new
position/size can be read after `nk_group_end` from `nk_panel`
directly.
Extended toggle style and changed toggle (checkbox + radio button) to
depend on the font height instead of row height. The main reason is
that it is easier to configure the cursor inside.
This commit adds a previously missing `group` flag to each group panel
and therefore allows to check for groups inside `nk_panel_end` and
draw the correct border color.
Fixed a small bug introduced by the last commit which prevent
groups from being moved. In addition I fixed an old scaling bug for
groups. It should now be possible to have movable and scalable
groups.
Fixed window y-padding and added scrollbar cursor padding and extended
scrollbar styling properties to separate between cursor and scrollbar.
In addition I added a property to control tab indentation. Finally fixed
clipped horizontal scrollbar for group.
While the last commit fixed some issues it still had a bug which
caused wrong scrollbar target size calculation. So this commit
hopefully fixes the problem.
Previously if a group had a scrollbar and is not completly visible
inside the window then the its scrollbar would use the clipped
height instead of its actual height as scroll reference. This should
now be fixed and work correctly.
In addition I split the tab/node button style into two one for the
minimized and one for the maximized state change.
Previously window borders were wrongfully drawn with already updated
size onto a window drawn with old size. This is now fixed and scaling
is a delayed by one frame.
In addition I hopefully finally fixed a window movement bug which
occurs if you drag one window header over another one really fast.
This would cause the previously not selected window to be wrongfully
selected and moved.
This commit adds another init function to font atlas which allows to
specify one allocator for temporary and one for permanent allocation.
In general this makes it managing memory a little bit more fine
grained. This is not a breaking change and the old API still works
without problems.
For some reason `xlib` still draws text even if the scissor rect has
a width or height of `0`. I don't know if this is a bug on my end or
on xlib but whatever the source it is now fixed.
It's now possible to define a specific color for each line or
column entry inside a chart. This purely extends the API and all
previous implementation still work.
This is a breaking change to hopefully noone. I removed all drawing
callbacks inside `nk_style` structs. The reason is that they are
quite ugly and a lot of work to maintain in the future. In addition
I added `nk_selectable_image_text`, `nk_selectable_image_label`,
`nk_select_image_text` and `nk_select_image_label` for combined
text and image selectables.
While mouse grabbing already works in x-direction for vertical and
in y-direction for horizontal scrollbars moving over scrollbar
boundries in the other direction would fail. The problem is now fixed
and scrollbar should work as expected. Note: interestingly slider
already work correctly even though I did not do anything to make it
work correctly.
Hopefully fixed selection bug for empty lines and made edit widget
usable while being only partially visible. I also added two addition
modes for a vim like expierence with view, insert and replace mode
instead of default replace mode, but only added the new behavior into
the x11 demo so far.
This commit fixes a stupid reallocation bug I had inside the library.
Basically I get a pointer inside a growing buffer which can be
invalidated by a second allocation and regrowth of said buffer.
Under normal circumstances I do not do this but in one instance I
actually did. That is for allocation temporary normals. This issue
should now be fixed and work correctly.
This fixes some small harmless unitialized variable access violations
found by either valgrind or ASAN. Nothing really ground breaking but
still cleaner.
Previously you had to first activate a window by clicking if it was
visually blocked by another window to move it around which was a
little bit blocky. It should now be possible to directly move and
activate a window.
Previously buttons would show up as hovered if another widget was
being hovered in another window. This behavior is now fixed and
everything should work correctly now.
Fixed some non-portable implementation details inside window and
table memory pool which could cause problems for platforms other
than x86/x64 and fixed some strict aliasing problems.
Tables inside windows get removed if a windows is not longer in use
inside `nk_free_window`. But I noticed that they are only removed
if they were not used the last frame, because of a copy & paste bug.
I removed the additional check and every table should now rightfully
be marked as free. I don't know if this bug actually caused any
problems, since I checked memory consumption and there were no
symptoms of wrong behavior before.
This patch hacks the chart API to support display of multiple data sets in a
single chart instances. Tested briefly with the line chart.
Signed-off-by: Jan Viktorin <iviktorin@fit.vutbr.cz>
Previously I only used the `__FILE__` and a user `ID` to make
sure to get an unqiue identifier out of it. Like @ands pointed out
this is not enough. Since it is possible to have multipe instances
with the same `title` in the same `__FILE__`. So to fix that I had
the idea to string concat `__FILE__` and `__LINE__` into a string.
This allows to have the same sequence of `IDs` in different places
in one `__FILE__`.
Previously it was not possible to check if currently any window or
widget is active. To fix that I added function `nk_item_any_active`.
To get correct result you should only call it after all UI code
ran but before calling `nk_clear`. There is currently no other way to
achive the same result inside UI code.
It is now possible to control button click behavior to trigger either
on button pressed or button released. By default nuklear now used
on button pressed but `#define NK_BUTTON_TRIGGER_ON_RELEASE` lets
you change button behavior to trigger on release. I furthermore
enhanced slider dragging behavior to not force you to hover the
slider at all times.
Editfield width inside properties were previously calculate by text
length + padding and always the size of the edit cursor, even in case
were no one is editing the property. Now the cursor size is only
added while editing.
Fixed a small copy & paste error which pushes the wrong texture
into an empty draw list. In normal cases this never occurs but is
still a big if used independent of libraries use cases.
`nk_tree__push` was not supposed to be seen as private but more like
the more extended way of handling duplicate tree elements. But I used
double underscore which is the general indicator for private. I renamed
it know to nk_tree_push_id, to hopefully make it clearer.
This commit fixes a bug that occurs if you move the cursor inside
text and then back to the end of the text and insert a character.
Caused was the bug by an incorrect check if the cursor is at the
end of the string.
Previously if you use `nk_edit_string` and use flag
`NK_EDIT_CLIPBOARD`, callbacks were not passed. I only tested it with
`nk_edit_buffer` and set the callbacks directly. This wrong behavior
is now fixed and should work copy & paste should behave correctly.
I previously used a conditional style so I don't have to compare
floats directly with eachother because of warning. This caused some
bugs and I now reverted back all conditions back to normal.
In addition I oversampled the default font to make it look better.
Thanks to @codemeow I fixed some bugs and warnings found with
cppcheck. NOTE: it seems like all static analyzer fail to
detect the assert macro or hate combining assert ans if checks.
Some minor issues: an unused variable, fixed a comparison of int against
float, and a pragma warning (pop) that didn't have a matching pragma
warning (push). Potentially more serious was a possible buffer overrun,
where a pointer was being tested against a null terminator, instead of
the char it points to.
There were a few casts from void * missing causing problems when
compiling as C++, also the assert function from <assert.h> was being
used instead of NK_ASSERT(...) in one place.
Fixed window movement for windows without header and some small
wrong calculation inside the panel layout code. In addition I
added another example implementing a simple calculator.
This is the first release version of nuklear (previously: zahnrad).
As for those who no the old version will notice: a lot has changed.
Most obvious should be the two biggest changes. First the name change
because I got critique that the name is hard to comprehend and
remember (understandable for non-germans) and the second is the
transistion from four files (zahnrad.h, zahnrad.c, stb_truetype
and stb_rect_pack) to one single header library file nuklear.h.
I am not 100% convinced that using a single header library is the
right choice here but so far I haven't encountered any problems.
Noticable should be as well that nuklear now directly embeds three
stb libraries: stb_truetype, stb_rect_pack and stb_textedit. Like
in previous versions the first two are optional and the library
can be compiled without. stb_textedit on the other hand powers
the text edit implementation for single as well as multiline
text manipulation. The text edit implementation is still relative
new and untested so you can expect some bugs I have not found yet.
In the demo department a lot changed as well. All platform demos
now don't compile one big demo but instead contain a simple
demo and small abstraction layer over the platform. Main benefit is
better understandablity improved ease of use. The old demo
is now split up and transfered into the example folder while each part
is self contained and compileable. (All examples use glfw I don't now
if this is the best platform but it is at least the simplest.
I also removed the apple demo because I don't have an apple system
and cannot make sure the new version runs with the old version.
Finally a lot of small bugs have been fixed as well as bugs found by
clang analyzer and coverity.
