Declarative User Interfaces for Python

The main change here is the addition of a new example file, which demonstrates how the various date-related widgets look and work, and documentation to match.

While doing that work, I made a number of other more minor changes, including reducing the Sphinx warnings, which included exposing another example that had been hidden.

I discovered all of these issues when testing with the desktop example.

I have started testing against my application (psiexperiment) and may have more to report later.

Examples to check

• [ ] aliases
• [ ] applib
• [ ] dynamic
• [ ] functions
• [ ] layout
• [ ] stdlib
• [ ] styling
• [ ] templates
• [ ] tutorial
• [x] widgets
• [ ] workbench

Broken examples

• [ ] drag_and_drop.enaml (drag and drop broken)
• [ ] file_dialog.enaml (browse button broken)
• [ ] flow_area.enaml (changing flow layout)
• [ ] image_view.enaml (set to None)
• [ ] ipython_console.enaml (possibly just need to install qtconsole),
• [ ] vtk_canvas.enaml (most likely need to install vtk)
• [ ] web_view.enaml

I've been playing around with trying to use Enaml for creating mobile apps (see enaml-native) in python using native widgets. It is working, however loading enaml is awful slow (takes over 8 seconds on a Nexus 5) which is pretty much a killer for the whole concept.

After some investigation this is due to the yacc.yacc call in enaml/core/parser.py#L3984 taking almost 8 seconds to complete. (There's also jnius.autoclass taking forever but I'm working on that as well)

05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:          1124963 function calls (1115859 primitive calls) in 22.236 seconds
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:    Ordered by: cumulative time
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:    ncalls  tottime  percall  cumtime  percall filename:lineno(function)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:        33   10.232    0.310   12.482    0.378 /data/user/0/com.jventura.pyapp/assets/python/site-packages/jnius/reflect.py:149(autoclass)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.000    0.000   11.285   11.285 /data/user/0/com.jventura.pyapp/assets/python/bootstrap.py:22(router)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.000    0.000   11.285   11.285 /data/user/0/com.jventura.pyapp/assets/python/bootstrap.py:33(handle)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.000    0.000   11.285   11.285 /data/user/0/com.jventura.pyapp/assets/python/bootstrap.py:100(start)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.008    0.008   11.285   11.285 /data/user/0/com.jventura.pyapp/assets/python/enamlnative/android/app.py:51(start)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.000    0.000   11.277   11.277 /data/user/0/com.jventura.pyapp/assets/python/enamlnative/android/app.py:60(get_view)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:      25/1    0.001    0.000   10.954   10.954 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/widgets/toolkit_object.py:147(initialize)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:        25    0.001    0.000   10.952    0.438 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/application.py:351(create_proxy)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:        25    0.000    0.000   10.951    0.438 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/application.py:325(resolve_proxy_class)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:        30    0.000    0.000   10.950    0.365 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/application.py:127(resolve)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:      25/1    0.001    0.000    9.258    9.258 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/core/declarative.py:103(initialize)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.022    0.022    8.264    8.264 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/__init__.py:36(imports)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.115    0.115    8.243    8.243 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/core/import_hooks.py:8(<module>)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.090    0.090    8.023    8.023 /data/user/0/com.jventura.pyapp/assets/python/site-packages/enaml/core/parser.py:8(<module>)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.005    0.005    7.719    7.719 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:3214(yacc)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.001    0.001    6.869    6.869 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:2100(__init__)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.690    0.690    6.523    6.523 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:2533(lr_parse_table)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.000    0.000    3.958    3.958 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:2509(add_lalr_lookaheads)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    1.812    1.812    1.816    1.816 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:2283(find_nonterminal_transitions)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:     53114    1.332    0.000    1.804    0.000 /data/user/0/com.jventura.pyapp/assets/python/site-packages/ply/yacc.py:2163(lr0_goto)
05-26 12:30:49.509 19789-19789/com.jventura.pyapp I/pybridge:         1    0.007    0.007    1.696    1.696 /data/user/0/com.jventura.pyapp/assets/python/enamlnative/android/factories.py:103(drawer_layout_factory)
more excluded ...

Does enaml require the use of ply? Could some alternative be used? If not, I've been thinking about cythonizing ply, but before going that far I thought I'd check here and see if that would even work.

Thanks!

This is a meta to track the changes associated with Python 3.8:

Required changes (meaning enaml won't run without them):

• bytecode changes:
  • MAP_ADD: the expected order on the TOS changed
  • BREAK_LOOP, CONTINUE_LOOP, SETUP_LOOP, SETUP_EXCEPT have been removed and as a consequence the code_generator for_loop and try_squash_raise need to be updated.
  • ROT_FOUR, BEGIN_FINALLY, CALL_FINALLY, POP_FINALLY have been added which may impact try_squash_raise.
  • END_FINALLY and WITH_CLEANUP_START changed of behavior
  • CodeType signature changed

Those require to update wbyteplay and in particular the stack trace calculation and since we will be dropping Python 2 it may be the proper time to switch to https://github.com/vstinner/bytecode which a more modern interface to the Python bytecode.

Other changes (required for feature parity with normal Python):

• support for assignment expression (may cause grammar issue, and is tightly related to https://github.com/nucleic/enaml/pull/348)
• positional only argument (change to the parser but should not be a big problem, except it may mean the AST for function may have change which would imply more pervasive changes)
• f-strings now have a new formatting mode marked by '='
• continue is now allowed in finally clause (not sure we were actively preventing it on the parser side)
• the rules for unpacking in yield and return have changed.

This is a proxy for Qt's QButtonGroup class, which allows manual management of button groupings. This alleviates the restriction of requiring all buttons to be within the same container to ensure that only one button in the desired grouping can be selected at a time.

I don't think this is ready to merge in its current state. The requirement to manually hook up the button to the button group in the 'activated' event handler is very clunky, but I'm not sure at present of a better way of handling this.

I'm also not sure that it makes sense to have the button group as an actual widget, but I wasn't sure of a better way to handle this either. The Timer widget (a proxy for the QTimer class) appears to work in the same way that this patch currently does.

I'm pretty sure there is some other documentation that needs updating, such as anything that claims the only way to implement exclusitivity is through a common container parent. I'd appreciate some guidance on this, and on if any additional testing is required.

I also wanted to say thank you for such an amazing piece of work. I found Enaml just three weeks ago, and it's been an absolute joy to work with. I didn't think it was possible for desktop application development to be so easy and fun. I know a large part of the software development industry is moving towards the web, but there are absolutely still plenty of important use cases where a desktop application makes more sense, so it's very nice to see toolkits like this still being developed.

I'm not sure why but git made me commit my other changes so this also includes adding autocomplete to the scintilla widget.

I can break it apart if needed.

Hi, just trying the example files, yet some are acting "weird" :

enaml-main\examples\styling\banner.enaml

The dark gradient is way too dark to see it is actually a gradient. Use this instead :

'stop: 0 #1356A9, stop: 0.3 #8AAFDC, stop: 0.58 #E0E4E0, stop: 0.68 #F8D8B1, stop: 0.848 #D39B8A, stop: 0.8499 #9C7F73, stop: 0.85 #D79F88, stop: 0.851 #E2BF9B, stop: 1 #817F73)'

enaml-main\examples\templates\basic.enaml

ok_clicked and cancel_clicked are obviously not handled like intended, nothing is printed into the console.

enaml-main\examples\widgets\buttons.enaml

While selecting RadionButton from the second container (two lowest radios) you can unselect radios from the first container.

enaml-main\examples\widgets\dual_slider.enaml

Sliders' tab are way too tiny :

enaml-main\examples\widgets\file_dialog.enaml

The file dialog opens relatively to its parent window (or not, even when parent=None). How is it possible to open the dialog always on screen which is the parent window yet centered ? Or best, save its position and last path for next time.

enaml-main\examples\widgets\flow_area.enaml

Item(FlowItem).Html doesn't show up until 'Ortho Stretch' is at least 1 and up or 'Preferred Height' is at least 173 and up.

enaml-main\examples\widgets\image_view.enaml

The motion of the ImageView is pretty weird because related of the spacer's width. Even if the window's height is high enough, if its width is narrowed the image goes up.

enaml-main\examples\widgets\mdi_area.enaml

Tiling : get 3 mdi windows, tile them, move the top left and top right around, tile them again, the title bar of the top right window is not redrawn correctly.

Cascading : get 3 mdi windows, cascade them, move the top left one around, cascade them again, the middle one gets buried. I mean, the position should be set to top alternatively so that we can read every title bar like the first time. Btw, is there a way to define the horizontal and vertical spacing of the tiled windows ?

enaml-main\examples\layout\basic\stack.enaml

Maybe there should be one.

enaml-main\examples\tutorial\plugin\plugin.enaml

Maybe there should be one.

enaml-main\examples\tutorial\scintilla\scintilla.enaml

Maybe there should be one.

Adds an enaml-compileall command that patches python's compileall to also compile enaml files.

Python 3 has legacy and optimize flags that are not used. https://github.com/python/cpython/blob/3.6/Lib/compileall.py#L95

I'm trying to filter a table, however this simple example is not working in python3 (on the current master branch).

from enaml.widgets.api import Window, Container, Label, Field


enamldef Main(Window):
    Container:
        attr colors = ['red', 'blue', 'yellow', 'green']
        Field: search:
            placeholder = "Search..."
        Label:
            text << '{}'.format(search.text)
        Label:
            text << '{}'.format([c for c in colors
                                    if not search.text or search.text in c])

The second Label displays correctly initially but never updates thereafter. The first Label works as expected. Both work fine in Python 2.7.

I tried a few more and it seems to be only with generator expressions on the RHS, so pulling the filtering out into a function is a workaround (and maybe a better to do it anyways).

from enaml.widgets.api import Window, Container, Label, Field

STATIC_COLORS = ['red', 'blue', 'yellow', 'green']

def echo(text):
    return text

def echo_gen(text):
    return [text for i in range(3)]

def echo_global(text):
    return [c for c in STATIC_COLORS if text in c]

def echo_ref(text, colors):
    return [c for c in colors if text in c]

enamldef Main(Window):
    Container: container:
        attr colors = ['red', 'blue', 'yellow', 'green']
        Field: search:
            placeholder = "Search..."

        Label: t1:
            #: Works
            text << '{}'.format(search.text)

        Label: t2:
            #: Works
            text << '{}'.format([search.text])

        Label: t3:
            #: Works
            text << '{}'.format(echo(search.text))
        Label: t4:
            #: Works
            text << '{}'.format(echo_gen(search.text))
        Label: t5:
            #: Works
            text << '{}'.format(echo_global(search.text))
        Label: t6:
            #: Works
            text << '{}'.format(echo_ref(search.text, colors))
        Label: t7:
            #: Doesn't update in 2.7 (master) OR 3.5
            text << '{}'.format({i:search.text for i in range(3)})
        Label: t8:
            #: Doesn't update in 3.5
            text << '{}'.format([search.text for i in range(3)])
        Label: t9:
            #: Doesn't update in 3.5
            text << '{}'.format([c for c in colors
                                    if not search.text or search.text in c])

The dict one is not working in 2.7 or python 3 (master branch). Interestingly, if I use enaml==0.9.8 it won't run unless that's commented out (gives a NameError: global name 'search' is not defined) so I assume dict comps never were supported.

Python 3.8 introduced many changes to the bytecode. This PR will focus on addressing them and allow to run enaml on Python 3.8 using Python 3.7 syntax. New syntax introduced in Python 3.8 will be the focus of a different PR.

The first step (completed) towards supporting Python 3.8 is to switch from byteplay to bytecode (more modern interface and actively maintained) since Python 3.8 introduce once again many changes to the bytecode.

The code requires ~https://github.com/vstinner/bytecode master~ bytecode 0.10.0 and pass all the tests on Python 3.8, 3.7 and 3.6. ~The next step is to adapt for Python 3.8.~

With enaml 0.8.0 (2b2dfd6050b5190d1a251446b1c9dfc69b640525) and atom 0.3.3 (6dcf3e411bfc7ebb2ee8faffa7c1a0d5ab85b99b), and using the version of PySide in the latest anaconda, many of the examples raise a RuntimeError. E.g.

$ enaml-run form.enaml 
Traceback (most recent call last):
  File "/home/warren/local_enaml/bin/enaml-run", line 9, in <module>
    load_entry_point('enaml==0.8.0', 'console_scripts', 'enaml-run')()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/runner.py", line 64, in main
    window.show()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/widgets/window.py", line 331, in show
    self.activate_proxy()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/widgets/toolkit_object.py", line 206, in activate_proxy
    child.activate_proxy()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/widgets/toolkit_object.py", line 206, in activate_proxy
    child.activate_proxy()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/widgets/toolkit_object.py", line 203, in activate_proxy
    self.activate_top_down()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/widgets/toolkit_object.py", line 219, in activate_top_down
    self.proxy.activate_top_down()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/qt/qt_toolkit_object.py", line 68, in activate_top_down
    self.init_widget()
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/qt/qt_field.py", line 78, in init_widget
    self.set_submit_triggers(d.submit_triggers)
  File "/home/warren/local_enaml/lib/python2.7/site-packages/enaml-0.8.0-py2.7-linux-x86_64.egg/enaml/qt/qt_field.py", line 168, in set_submit_triggers
    widget.lostFocus.disconnect(handler)
RuntimeError: Failed to disconnect signal lostFocus().

This updates the code generator/compiler to work with Python 3.11, it builds on the pegen branch because the main branch just fails due to invalid col offsets in the ast.

• It seems like the python docs are wrong with MAKE_FUNCTION as there is now no need to push a qualname (https://github.com/python/cpython/issues/93270).
• Since POP_BLOCK was removed, I'm not sure if try_squash_raise can still be implemented?
• There is still a weird problem with decl funcs with default args. Inspect shows the signature is correct and funchelper shows the correct number of defaults but for some reason it does not use the defaults?

On a positive note I get about a 10-15% speedup with this on enaml-web :).

This there a wait to do this "natively" from within Enaml :

https://stackoverflow.com/questions/8218900/how-can-i-change-the-cursor-shape-with-pyqt

.enaml files appear completely opaques to most tools and while we cannot do much in a generic sense we could improve the user experience inside Python files using definition found in an enaml file by providing .pyi for enaml files.

The generation could be made from the AST generated by the parser:

• for Python blocks we could empty the body of function nodes and simply unparse the AST (auto formatting after could handle messed up line numbers)
• for enamldef we could use a class and generate type annotations for attr, aliases, and event and copy the annotations of declarative func. For attr, enaml syntax already has a type annotation, for aliases we would need to resolve to what the alias point to and if it is an atom member we should recover the matching type useful in an annotation, for event we would simply use the atom member
• for template something similar could work too but it will need more thoughts

I just downloaded Enaml and am trying to follow the Person tutorial, but it's incomplete. It doesn't tell me how to name the files or how to run them, so I just get an error like this:

14:34:02 $ enaml-run lala.enaml
Could not find component 'Main'

In my opinion, tutorials should assume zero knowledge on the part of the reader, so the tutorial should tell me the names of the files and the command I need to see the UI.

For #493 . It works by saving a reference to the compiler node to each declarative object and copies the filename and line number during compilation.

Unfortunately this breaks any existing caches because of the additional parameters, I'm not sure what to do about that, it could maybe provide empty defaults for the filename and lineno?

Also it currently doesn't catch errors when manually updating an expression, eg so I'm not sure if it should be doing a try/except in the expression engine itself.