Alex,
This is very exciting stuff and although I am a poor C++ programmer I
would like to do what I can to help the process along - I will checkout
the branches later and see how I go.
Many thanks for all your work!
Regards,
Phil.
On 2018-03-31 11:03, Alex Valavanis wrote:
Hi All,
Here's a summary of my work at the Boston Hackfest.
Before I continue, I'd like to thank everyone who donated their money
or time to support this Hackfest. This has been a really useful event
- it's so much easier to deal with tricky design challenges and
project management stuff when we can meet face to face. It has also
brought together people with very diverse expertise across the project
(UI, web, build, extensions, project management etc). This week,
we've figured out a path to deliver long desired features (multipage
support etc), roadmapped our final route to deliver Inkscape 1.0, done
some great website design work, and handled some tricky code structure
issues that I'll discuss in this message.
If you're able to donate some money to support Inkscape, and future
Hackfests, we really appreciate this. You can find the donation link
on our website:
https://inkscape.org/en/support-us/hackfests/
It would also be great to get some more of the regular contributors
from our wider community to join us at future events (i.e., designers,
bug-team, experienced users... not just C++ developers!). The
developers can learn so much more about the project by discussing it
with you.
So... what have I been doing this week?
== What are Actions? ==
Basically, an "action" is a thing that Inkscape can do (e.g., flip an
object horizontally), which we can identify using an ID code (e.g.,
"SP_VERB_OBJECT_FLIP_HORIZONTAL"). We can then hook up buttons, menu
items etc to activate these actions. In this example, we would set up
our Flip Horizontal button in the select-tool toolbar so that when the
user clicks it, it activates the required action.
My work this week has focused on improving the way we implement
actions in our code, which will
* Allow Inkscape to run on machines with future versions of the Gtk+
library (Gtk 4+)
* Provide much better separation between the behaviour and user
interface (i.e.,, separate pieces of the code will describe what
Inkscape "does", and what it "looks like")
* Make it much easier to create new "flavours" of the Inkscape user
interface - for example, an "Inkscape for Kids" or "Inkscape CAD"
* Make it easier for "non-programmers" to create and reorganise
custom toolbars and dialogs
* Allow us to remove a lot of older "C" code that is very difficult
for current "C++" developers to maintain
== Where we are now? ==
In the Inkscape code, we have a mechanism called "verbs" that acts as
a kind of action mechanism... it provides a way of using an ID code to
look up the required commands to perform an action.
On top of this, we have another structure called SPAction that
essentially hooks up a couple of signals to the verb.
Finally(ish), we use objects called GtkActions from the Gtk+ library,
which let us create GUI elements (buttons etc) that we can add to our
menus or toolbars that activate the required action.
We have also got a few custom sub-classes of GtkAction that let us
create special kinds of tool-item that can activate actions... for
example, one of these creates a spin-button (a box with a number in
it, and "up/down" arrows) that can also handle units, basic
arithmetic, and provides a label.
== What's the problem? ==
First, the GtkAction widget (and hence all our custom sub-classes) is
deprecated and has been removed from Gtk+ 4, so we need to change.
Since Inkscape 1.0 is intended as a "stable" release, we want to make
it as future-proof as possible.
The main reason for this deprecation is that the GtkAction widget is a
slightly weird mix of paradigms: it mixes information about how to
display a tool/menu item with information about behaviour. This
causes some problems... what if we wanted to run a "headless" Inkscape
on a system that has no display? What if we wanted to provide several
different GUI options for different kinds of user to trigger the same
behaviour?
This is one of the reasons why we have the "verbs" system is in place,
as it separates behaviour from the GUI. However, there is still quite
a lot of code "overhead" needed if we want to hook up Inkscape's Verb
behaviour to a GUI element. Essentially, it's "programmers only"
territory. What if a designer wanted to add a new toolbar button to
their Inkscape to handle a useful feature that is currently buried
deep in a menu? What if a teacher wanted to strip out a lot of
"advanced" behaviour from the interface to make it easier to introduce
Inkscape to their class? At the moment, they have to edit the C++
code, and recompile it... a slow and complicated process!
== Where are we going? ==
We need to move to using GActions (a feature in the Glib/GIO library).
This *only* describes the action's behaviour and provides an
"activated" signal that we can trigger by simply knowing the action's
name. This is completely separated from the GUI. These actions can
be activated very easily from a whole range of places, for example:
* By attaching the action's name to a GUI element in the C++ code
(replacing our SPAction/GtkAction code)
* By activating the action (using its name) from a command-line
interface (replacing some of our verbs code)
* By attaching the action's name to elements in a customised GUI file
* From external applications, using the DBus system
Number 3 in the list provides some very exciting opportunities for
future releases, as non-programmers will be able to make a customised
user interface by drawing one in the Glade application. You then just
attach the name of the required Inkscape action to each GUI element.
Want a new button? Just edit Inkscape's user-interface file in Glade
and see the changes immediately next time you start... no C++
knowledge needed!
== What have I done so far? ==
I've started working through Inkscape's toolbars, and have replaced
the deprecated GtkAction-based widgets with standard widgets (buttons
etc). Some of these now hook up to GAction descriptions of behaviour,
while others are just based on handling GUI signals (e.g.,
Button::clicked).
So far, I have migrated the toolbars for the Select, Dropper and
Connector toolbar, and have killed off a couple of the custom
GtkAction-based widgets that were only used in these.
In the process, I have also migrated the toolbar code to C++, which
should make it a little easier for the current core developers to
maintain. Since all the data is now stored as C++ member variables,
rather than as GObject data, we now benefit from better compile-time
and run-time type-checking that will help to safeguard against errors
being introduced in the code in future.
I have placed the updated toolbars in the Inkscape::UI::Toolbar
namespace and made them into proper sub-classes of the Gtk::Toolbar
class.
== How can you help? ==
This important refactoring project must be done before Gtk+ 4 is
widely used, and represents a major change to the way that our
toolbars are implemented. It represents a couple of thousand lines of
code changes and will inevitably cause some regression bugs. I will
need your help in thoroughly testing that all the toolbars work
correcly!
All of these changes are currently in a git branch
("gtk-actions-migration") and will not be merged into the master
branch until preliminary testing has been completed. In other words,
this will not "mess up" your stable version of Inkscape and will never
be part of the current 0.92.x release series! I intend to merge this
work well in advance of the Inkscape 1.0 release, so that we have
plenty of time for very thorough developer and user testing, and
several preliminary bug-fix "alpha" pre-releases before the final
stable release.
For the very brave: please checkout the "gtk-actions-migration" branch
in git, rebuild and test. Please note that this is work-in progress.
Do not use this experimental branch for your regular Inkscape drawing
work!!! The Node toolbar will not work... the Select, Dropper and
Connector toolbars are highly experimental; all the other toolbars are
"at risk"! If you don't know how to do this, check out our
developer guide.
For the quite brave: I'll merge this experimental branch into our
master development branch in a few weeks once (a) I've completed the
basic work, (b) I've had a good amount of feedback on the branch and
know that it "works" to some extent. I'll message you again at this
stage and ask everyone to help look for bugs. Again, this will only
be visible to developers and users of our experimental "trunk" PPA in
Ubuntu.
For the intrepid: If you're interested in helping with the code
migration - message me and I'll help to divide up the tasks.
== The end ==
I hope this gives an insight into the work I've started doing this
week, and will continue over the coming weeks. Thanks once again to
everyone who has contributed time, money or even just attention and
enthusiasm for the Hackfest.
Best wishes,
Alex Valavanis (valavanisalex)
Inkscape developer
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites,
Slashdot.org!
http://sdm.link/slashdot
_______________________________________________
Inkscape-devel mailing list
Inkscape-devel(a)lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/inkscape-devel
--
Philip Rhoades
PO Box 896
Cowra NSW 2794
Australia
E-mail: phil@...1587...