# Copyright (c) Jupyter Development Team. # Distributed under the terms of the Modified BSD License. """Interact with functions using widgets.""" from collections.abc import Iterable, Mapping from enum import EnumMeta as EnumType from inspect import signature, Parameter from inspect import getcallargs from inspect import getfullargspec as check_argspec import sys from IPython import get_ipython from . import (Widget, ValueWidget, Text, FloatSlider, FloatText, IntSlider, IntText, Checkbox, Dropdown, VBox, Button, DOMWidget, Output) from IPython.display import display, clear_output from traitlets import HasTraits, Any, Unicode, observe from numbers import Real, Integral from warnings import warn empty = Parameter.empty def show_inline_matplotlib_plots(): """Show matplotlib plots immediately if using the inline backend. With ipywidgets 6.0, matplotlib plots don't work well with interact when using the inline backend that comes with ipykernel. Basically, the inline backend only shows the plot after the entire cell executes, which does not play well with drawing plots inside of an interact function. See https://github.com/jupyter-widgets/ipywidgets/issues/1181/ and https://github.com/ipython/ipython/issues/10376 for more details. This function displays any matplotlib plots if the backend is the inline backend. """ if 'matplotlib' not in sys.modules: # matplotlib hasn't been imported, nothing to do. return try: import matplotlib as mpl from matplotlib_inline.backend_inline import flush_figures except ImportError: return if (mpl.get_backend() == 'module://ipykernel.pylab.backend_inline' or mpl.get_backend() == 'module://matplotlib_inline.backend_inline'): flush_figures() def interactive_output(f, controls): """Connect widget controls to a function. This function does not generate a user interface for the widgets (unlike `interact`). This enables customisation of the widget user interface layout. The user interface layout must be defined and displayed manually. """ out = Output() def observer(change): kwargs = {k:v.value for k,v in controls.items()} show_inline_matplotlib_plots() with out: clear_output(wait=True) f(**kwargs) show_inline_matplotlib_plots() for k,w in controls.items(): w.observe(observer, 'value') show_inline_matplotlib_plots() observer(None) return out def _matches(o, pattern): """Match a pattern of types in a sequence.""" if not len(o) == len(pattern): return False comps = zip(o,pattern) return all(isinstance(obj,kind) for obj,kind in comps) def _get_min_max_value(min, max, value=None, step=None): """Return min, max, value given input values with possible None.""" # Either min and max need to be given, or value needs to be given if value is None: if min is None or max is None: raise ValueError('unable to infer range, value from: ({}, {}, {})'.format(min, max, value)) diff = max - min value = min + (diff / 2) # Ensure that value has the same type as diff if not isinstance(value, type(diff)): value = min + (diff // 2) else: # value is not None if not isinstance(value, Real): raise TypeError('expected a real number, got: %r' % value) # Infer min/max from value if value == 0: # This gives (0, 1) of the correct type vrange = (value, value + 1) elif value > 0: vrange = (-value, 3*value) else: vrange = (3*value, -value) if min is None: min = vrange[0] if max is None: max = vrange[1] if step is not None: # ensure value is on a step tick = int((value - min) / step) value = min + tick * step if not min <= value <= max: raise ValueError('value must be between min and max (min={}, value={}, max={})'.format(min, value, max)) return min, max, value def _yield_abbreviations_for_parameter(param, kwargs): """Get an abbreviation for a function parameter.""" name = param.name kind = param.kind default = param.default not_found = (name, empty, empty) if kind in (Parameter.POSITIONAL_OR_KEYWORD, Parameter.KEYWORD_ONLY): if name in kwargs: value = kwargs.pop(name) elif default is not empty: value = default elif param.annotation: value = param.annotation else: yield not_found yield (name, value, default) elif kind == Parameter.VAR_KEYWORD: # In this case name=kwargs and we yield the items in kwargs with their keys. for k, v in kwargs.copy().items(): kwargs.pop(k) yield k, v, empty class interactive(VBox): """ A VBox container containing a group of interactive widgets tied to a function. Parameters ---------- __interact_f : function The function to which the interactive widgets are tied. The `**kwargs` should match the function signature. __options : dict A dict of options. Currently, the only supported keys are ``"manual"`` (defaults to ``False``), ``"manual_name"`` (defaults to ``"Run Interact"``) and ``"auto_display"`` (defaults to ``False``). **kwargs : various, optional An interactive widget is created for each keyword argument that is a valid widget abbreviation. Note that the first two parameters intentionally start with a double underscore to avoid being mixed up with keyword arguments passed by ``**kwargs``. """ def __init__(self, __interact_f, __options={}, **kwargs): VBox.__init__(self, _dom_classes=['widget-interact']) self.result = None self.args = [] self.kwargs = {} self.f = f = __interact_f self.clear_output = kwargs.pop('clear_output', True) self.manual = __options.get("manual", False) self.manual_name = __options.get("manual_name", "Run Interact") self.auto_display = __options.get("auto_display", False) new_kwargs = self.find_abbreviations(kwargs) # Before we proceed, let's make sure that the user has passed a set of args+kwargs # that will lead to a valid call of the function. This protects against unspecified # and doubly-specified arguments. try: check_argspec(f) except TypeError: # if we can't inspect, we can't validate pass else: getcallargs(f, **{n:v for n,v,_ in new_kwargs}) # Now build the widgets from the abbreviations. self.kwargs_widgets = self.widgets_from_abbreviations(new_kwargs) # This has to be done as an assignment, not using self.children.append, # so that traitlets notices the update. We skip any objects (such as fixed) that # are not DOMWidgets. c = [w for w in self.kwargs_widgets if isinstance(w, DOMWidget)] # If we are only to run the function on demand, add a button to request this. if self.manual: self.manual_button = Button(description=self.manual_name) c.append(self.manual_button) self.out = Output() c.append(self.out) self.children = c # Wire up the widgets # If we are doing manual running, the callback is only triggered by the button # Otherwise, it is triggered for every trait change received # On-demand running also suppresses running the function with the initial parameters if self.manual: self.manual_button.on_click(self.update) # Also register input handlers on text areas, so the user can hit return to # invoke execution. for w in self.kwargs_widgets: if isinstance(w, Text): w.continuous_update = False w.observe(self.update, names='value') else: for widget in self.kwargs_widgets: widget.observe(self.update, names='value') self.update() # Callback function def update(self, *args): """ Call the interact function and update the output widget with the result of the function call. Parameters ---------- *args : ignored Required for this method to be used as traitlets callback. """ self.kwargs = {} if self.manual: self.manual_button.disabled = True try: show_inline_matplotlib_plots() with self.out: if self.clear_output: clear_output(wait=True) for widget in self.kwargs_widgets: value = widget.get_interact_value() self.kwargs[widget._kwarg] = value self.result = self.f(**self.kwargs) show_inline_matplotlib_plots() if self.auto_display and self.result is not None: display(self.result) except Exception as e: ip = get_ipython() if ip is None: self.log.warning("Exception in interact callback: %s", e, exc_info=True) else: ip.showtraceback() finally: if self.manual: self.manual_button.disabled = False # Find abbreviations def signature(self): return signature(self.f) def find_abbreviations(self, kwargs): """Find the abbreviations for the given function and kwargs. Return (name, abbrev, default) tuples. """ new_kwargs = [] try: sig = self.signature() except (ValueError, TypeError): # can't inspect, no info from function; only use kwargs return [ (key, value, value) for key, value in kwargs.items() ] for param in sig.parameters.values(): for name, value, default in _yield_abbreviations_for_parameter(param, kwargs): if value is empty: raise ValueError('cannot find widget or abbreviation for argument: {!r}'.format(name)) new_kwargs.append((name, value, default)) return new_kwargs # Abbreviations to widgets def widgets_from_abbreviations(self, seq): """Given a sequence of (name, abbrev, default) tuples, return a sequence of Widgets.""" result = [] for name, abbrev, default in seq: if isinstance(abbrev, Widget) and (not isinstance(abbrev, ValueWidget)): raise TypeError("{!r} is not a ValueWidget".format(abbrev)) widget = self.widget_from_abbrev(abbrev, default) if widget is None: raise ValueError("{!r} cannot be transformed to a widget".format(abbrev)) if not hasattr(widget, "description") or not widget.description: widget.description = name widget._kwarg = name result.append(widget) return result @classmethod def widget_from_abbrev(cls, abbrev, default=empty): """Build a ValueWidget instance given an abbreviation or Widget.""" if isinstance(abbrev, ValueWidget) or isinstance(abbrev, fixed): return abbrev if isinstance(abbrev, tuple): widget = cls.widget_from_tuple(abbrev) if default is not empty: try: widget.value = default except Exception: # ignore failure to set default pass return widget # Try type annotation if isinstance(abbrev, type): widget = cls.widget_from_annotation(abbrev) if widget is not None: return widget # Try single value widget = cls.widget_from_single_value(abbrev) if widget is not None: return widget # Something iterable (list, dict, generator, ...). Note that str and # tuple should be handled before, that is why we check this case last. if isinstance(abbrev, Iterable): widget = cls.widget_from_iterable(abbrev) if default is not empty: try: widget.value = default except Exception: # ignore failure to set default pass return widget # No idea... return None @staticmethod def widget_from_single_value(o): """Make widgets from single values, which can be used as parameter defaults.""" if isinstance(o, str): return Text(value=str(o)) elif isinstance(o, bool): return Checkbox(value=o) elif isinstance(o, Integral): min, max, value = _get_min_max_value(None, None, o) return IntSlider(value=o, min=min, max=max) elif isinstance(o, Real): min, max, value = _get_min_max_value(None, None, o) return FloatSlider(value=o, min=min, max=max) else: return None @staticmethod def widget_from_annotation(t): """Make widgets from type annotation and optional default value.""" if t is str: return Text() elif t is bool: return Checkbox() elif t in {int, Integral}: return IntText() elif t in {float, Real}: return FloatText() elif isinstance(t, EnumType): return Dropdown(options={option.name: option for option in t}) else: return None @staticmethod def widget_from_tuple(o): """Make widgets from a tuple abbreviation.""" if _matches(o, (Real, Real)): min, max, value = _get_min_max_value(o[0], o[1]) if all(isinstance(_, Integral) for _ in o): cls = IntSlider else: cls = FloatSlider return cls(value=value, min=min, max=max) elif _matches(o, (Real, Real, Real)): step = o[2] if step <= 0: raise ValueError("step must be >= 0, not %r" % step) min, max, value = _get_min_max_value(o[0], o[1], step=step) if all(isinstance(_, Integral) for _ in o): cls = IntSlider else: cls = FloatSlider return cls(value=value, min=min, max=max, step=step) @staticmethod def widget_from_iterable(o): """Make widgets from an iterable. This should not be done for a string or tuple.""" # Dropdown expects a dict or list, so we convert an arbitrary # iterable to either of those. if isinstance(o, (list, dict)): return Dropdown(options=o) elif isinstance(o, Mapping): return Dropdown(options=list(o.items())) else: return Dropdown(options=list(o)) # Return a factory for interactive functions @classmethod def factory(cls): options = dict(manual=False, auto_display=True, manual_name="Run Interact") return _InteractFactory(cls, options) class _InteractFactory: """ Factory for instances of :class:`interactive`. This class is needed to support options like:: >>> @interact.options(manual=True) ... def greeting(text="World"): ... print("Hello {}".format(text)) Parameters ---------- cls : class The subclass of :class:`interactive` to construct. options : dict A dict of options used to construct the interactive function. By default, this is returned by ``cls.default_options()``. kwargs : dict A dict of **kwargs to use for widgets. """ def __init__(self, cls, options, kwargs={}): self.cls = cls self.opts = options self.kwargs = kwargs def widget(self, f): """ Return an interactive function widget for the given function. The widget is only constructed, not displayed nor attached to the function. Returns ------- An instance of ``self.cls`` (typically :class:`interactive`). Parameters ---------- f : function The function to which the interactive widgets are tied. """ return self.cls(f, self.opts, **self.kwargs) def __call__(self, __interact_f=None, **kwargs): """ Make the given function interactive by adding and displaying the corresponding :class:`interactive` widget. Expects the first argument to be a function. Parameters to this function are widget abbreviations passed in as keyword arguments (``**kwargs``). Can be used as a decorator (see examples). Returns ------- f : __interact_f with interactive widget attached to it. Parameters ---------- __interact_f : function The function to which the interactive widgets are tied. The `**kwargs` should match the function signature. Passed to :func:`interactive()` **kwargs : various, optional An interactive widget is created for each keyword argument that is a valid widget abbreviation. Passed to :func:`interactive()` Examples -------- Render an interactive text field that shows the greeting with the passed in text:: # 1. Using interact as a function def greeting(text="World"): print("Hello {}".format(text)) interact(greeting, text="Jupyter Widgets") # 2. Using interact as a decorator @interact def greeting(text="World"): print("Hello {}".format(text)) # 3. Using interact as a decorator with named parameters @interact(text="Jupyter Widgets") def greeting(text="World"): print("Hello {}".format(text)) Render an interactive slider widget and prints square of number:: # 1. Using interact as a function def square(num=1): print("{} squared is {}".format(num, num*num)) interact(square, num=5) # 2. Using interact as a decorator @interact def square(num=2): print("{} squared is {}".format(num, num*num)) # 3. Using interact as a decorator with named parameters @interact(num=5) def square(num=2): print("{} squared is {}".format(num, num*num)) """ # If kwargs are given, replace self by a new # _InteractFactory with the updated kwargs if kwargs: kw = dict(self.kwargs) kw.update(kwargs) self = type(self)(self.cls, self.opts, kw) f = __interact_f if f is None: # This branch handles the case 3 # @interact(a=30, b=40) # def f(*args, **kwargs): # ... # # Simply return the new factory return self # positional arg support in: https://gist.github.com/8851331 # Handle the cases 1 and 2 # 1. interact(f, **kwargs) # 2. @interact # def f(*args, **kwargs): # ... w = self.widget(f) try: f.widget = w except AttributeError: # some things (instancemethods) can't have attributes attached, # so wrap in a lambda f = lambda *args, **kwargs: __interact_f(*args, **kwargs) f.widget = w show_inline_matplotlib_plots() display(w) return f def options(self, **kwds): """ Change options for interactive functions. Returns ------- A new :class:`_InteractFactory` which will apply the options when called. """ opts = dict(self.opts) for k in kwds: try: # Ensure that the key exists because we want to change # existing options, not add new ones. _ = opts[k] except KeyError: raise ValueError("invalid option {!r}".format(k)) opts[k] = kwds[k] return type(self)(self.cls, opts, self.kwargs) interact = interactive.factory() interact_manual = interact.options(manual=True, manual_name="Run Interact") class fixed(HasTraits): """A pseudo-widget whose value is fixed and never synced to the client.""" value = Any(help="Any Python object") description = Unicode('', help="Any Python object") def __init__(self, value, **kwargs): super().__init__(value=value, **kwargs) def get_interact_value(self): """Return the value for this widget which should be passed to interactive functions. Custom widgets can change this method to process the raw value ``self.value``. """ return self.value