def test_suggesting_values_for_edit_variables(): """Test suggesting values in different situations. """ # Suggest value for an edit variable entering a weak equality s = Solver() v1 = Variable('foo') s.addEditVariable(v1, 'medium') s.addConstraint((v1 == 1) | 'weak') s.suggestValue(v1, 2) s.updateVariables() assert v1.value() == 2 # Suggest a value for an edit variable entering multiple solver rows s.reset() v1 = Variable('foo') v2 = Variable('bar') s = Solver() s.addEditVariable(v2, 'weak') s.addConstraint(v1 + v2 == 0) s.addConstraint((v2 <= -1)) s.addConstraint((v2 >= 0) | 'weak') s.suggestValue(v2, 0) s.updateVariables() assert v2.value() <= -1
def test_handling_infeasible_constraints(): """Test that we properly handle infeasible constraints. We use the example of the cassowary paper to generate an infeasible situation after updating an edit variable which causes the solver to use the dual optimization. """ xm = Variable('xm') xl = Variable('xl') xr = Variable('xr') s = Solver() s.addEditVariable(xm, 'strong') s.addEditVariable(xl, 'weak') s.addEditVariable(xr, 'weak') s.addConstraint(2 * xm == xl + xr) s.addConstraint(xl + 20 <= xr) s.addConstraint(xl >= -10) s.addConstraint(xr <= 100) s.suggestValue(xm, 40) s.suggestValue(xr, 50) s.suggestValue(xl, 30) # First update causing a normal update. s.suggestValue(xm, 60) # Create an infeasible condition triggering a dual optimization s.suggestValue(xm, 90) s.updateVariables() assert xl.value() + xr.value() == 2 * xm.value() assert xl.value() == 80 assert xr.value() == 100
def test_handling_infeasible_constraints(): """Test that we properly handle infeasible constraints. We use the example of the cassowary paper to generate an infeasible situation after updating an edit variable which causes the solver to use the dual optimization. """ xm = Variable('xm') xl = Variable('xl') xr = Variable('xr') s = Solver() s.addEditVariable(xm, 'strong') s.addEditVariable(xl, 'weak') s.addEditVariable(xr, 'weak') s.addConstraint(2*xm == xl + xr) s.addConstraint(xl + 20 <= xr) s.addConstraint(xl >= -10) s.addConstraint(xr <= 100) s.suggestValue(xm, 40) s.suggestValue(xr, 50) s.suggestValue(xl, 30) # First update causing a normal update. s.suggestValue(xm, 60) # Create an infeasible condition triggering a dual optimization s.suggestValue(xm, 90) s.updateVariables() assert xl.value() + xr.value() == 2*xm.value() assert xl.value() == 80 assert xr.value() == 100
def test_basic_solver(): s = Solver() x0 = Variable('x0') x1 = Variable('x1') s.addConstraint(x0 >= 0) s.addConstraint(x1 >= 0) s.updateVariables() assert x0.value() == 0.0 assert x1.value() == 0.0
def test_variable_methods(): """Test the variable modification methods. """ v = Variable() assert v.name() == "" v.setName(u'γ') assert v.name() == 'γ' v.setName('foo') assert v.name() == 'foo' with pytest.raises(TypeError): v.setName(1) if sys.version_info >= (3,): with pytest.raises(TypeError): v.setName(b'r') assert v.value() == 0.0 assert v.context() is None ctx = object() v.setContext(ctx) assert v.context() is ctx assert str(v) == 'foo' with pytest.raises(TypeError): Variable(1)
def test_variable_methods(): """Test the variable modification methods. """ v = Variable() assert v.name() == "" v.setName(u'γ') assert v.name() == 'γ' v.setName('foo') assert v.name() == 'foo' with pytest.raises(TypeError): v.setName(1) if sys.version_info >= (3, ): with pytest.raises(TypeError): v.setName(b'r') assert v.value() == 0.0 assert v.context() is None ctx = object() v.setContext(ctx) assert v.context() is ctx assert str(v) == 'foo' with pytest.raises(TypeError): Variable(1)
def test_solver(): s = Solver() x0 = Variable('x0') x1 = Variable('x1') s.addConstraint(x0 >= 0) s.addConstraint(x1 >= 0) s.addConstraint(x1 == x0) s.updateVariables() assert x0.value() == 0.0 assert x1.value() == 0.0 with edit_context(s, [x1]): s.suggestValue(x1, 1.0) s.updateVariables() assert x0.value() == 1.0 assert x1.value() == 1.0
def test_solving_with_strength(): """Test solving a system with unstatisfiable non-required constraint. """ v1 = Variable('foo') v2 = Variable('bar') s = Solver() s.addConstraint(v1 + v2 == 0) s.addConstraint(v1 == 10) s.addConstraint((v2 >= 0) | 'weak') s.updateVariables() assert v1.value() == 10 and v2.value() == -10 s.reset() s.addConstraint(v1 + v2 == 0) s.addConstraint((v1 >= 10) | 'medium') s.addConstraint((v2 == 2) | 'strong') s.updateVariables() assert v1.value() == -2 and v2.value() == 2
class Grid(Widget): """Widget for proportionally dividing its internal area into a grid. This widget will automatically set the position and size of child widgets according to provided constraints. Parameters ---------- spacing : int Spacing between widgets. **kwargs : dict Keyword arguments to pass to `Widget`. """ def __init__(self, spacing=6, **kwargs): """Create solver and basic grid parameters.""" self._next_cell = [0, 0] # row, col self._cells = {} self._grid_widgets = {} self.spacing = spacing self._n_added = 0 self._default_class = ViewBox # what to add when __getitem__ is used self._solver = Solver() self._need_solver_recreate = True # width and height of the Rect used to place child widgets self._var_w = Variable("w_rect") self._var_h = Variable("h_rect") self._width_grid = None self._height_grid = None # self._height_stay = None # self._width_stay = None Widget.__init__(self, **kwargs) def __getitem__(self, idxs): """Return an item or create it if the location is available.""" if not isinstance(idxs, tuple): idxs = (idxs, ) if len(idxs) == 1: idxs = idxs + (slice(0, 1, None), ) elif len(idxs) != 2: raise ValueError('Incorrect index: %s' % (idxs, )) lims = np.empty((2, 2), int) for ii, idx in enumerate(idxs): if isinstance(idx, int): idx = slice(idx, idx + 1, None) if not isinstance(idx, slice): raise ValueError('indices must be slices or integers, not %s' % (type(idx), )) if idx.step is not None and idx.step != 1: raise ValueError('step must be one or None, not %s' % idx.step) start = 0 if idx.start is None else idx.start end = self.grid_size[ii] if idx.stop is None else idx.stop lims[ii] = [start, end] layout = self.layout_array existing = layout[lims[0, 0]:lims[0, 1], lims[1, 0]:lims[1, 1]] + 1 if existing.any(): existing = set(list(existing.ravel())) ii = list(existing)[0] - 1 if len(existing) != 1 or ( (layout == ii).sum() != np.prod(np.diff(lims))): raise ValueError('Cannot add widget (collision)') return self._grid_widgets[ii][-1] spans = np.diff(lims)[:, 0] item = self.add_widget(self._default_class(), row=lims[0, 0], col=lims[1, 0], row_span=spans[0], col_span=spans[1]) return item def add_widget(self, widget=None, row=None, col=None, row_span=1, col_span=1, **kwargs): """Add a new widget to this grid. This will cause other widgets in the grid to be resized to make room for the new widget. Can be used to replace a widget as well. Parameters ---------- widget : Widget | None The Widget to add. New widget is constructed if widget is None. row : int The row in which to add the widget (0 is the topmost row) col : int The column in which to add the widget (0 is the leftmost column) row_span : int The number of rows to be occupied by this widget. Default is 1. col_span : int The number of columns to be occupied by this widget. Default is 1. **kwargs : dict parameters sent to the new Widget that is constructed if widget is None Notes ----- The widget's parent is automatically set to this grid, and all other parent(s) are removed. """ if row is None: row = self._next_cell[0] if col is None: col = self._next_cell[1] if widget is None: widget = Widget(**kwargs) else: if kwargs: raise ValueError("cannot send kwargs if widget is given") _row = self._cells.setdefault(row, {}) _row[col] = widget self._grid_widgets[self._n_added] = (row, col, row_span, col_span, widget) self._n_added += 1 widget.parent = self self._next_cell = [row, col + col_span] widget._var_w = Variable("w-(row: %s | col: %s)" % (row, col)) widget._var_h = Variable("h-(row: %s | col: %s)" % (row, col)) # update stretch based on colspan/rowspan # usually, if you make something consume more grids or columns, # you also want it to actually *take it up*, ratio wise. # otherwise, it will never *use* the extra rows and columns, # thereby collapsing the extras to 0. stretch = list(widget.stretch) stretch[0] = col_span if stretch[0] is None else stretch[0] stretch[1] = row_span if stretch[1] is None else stretch[1] widget.stretch = stretch self._need_solver_recreate = True return widget def remove_widget(self, widget): """Remove a widget from this grid. Parameters ---------- widget : Widget The Widget to remove """ self._grid_widgets = dict((key, val) for (key, val) in self._grid_widgets.items() if val[-1] != widget) self._need_solver_recreate = True def resize_widget(self, widget, row_span, col_span): """Resize a widget in the grid to new dimensions. Parameters ---------- widget : Widget The widget to resize row_span : int The number of rows to be occupied by this widget. col_span : int The number of columns to be occupied by this widget. """ row = None col = None for (r, c, _rspan, _cspan, w) in self._grid_widgets.values(): if w == widget: row = r col = c break if row is None or col is None: raise ValueError("%s not found in grid" % widget) self.remove_widget(widget) self.add_widget(widget, row, col, row_span, col_span) self._need_solver_recreate = True def _prepare_draw(self, view): self._update_child_widget_dim() def add_grid(self, row=None, col=None, row_span=1, col_span=1, **kwargs): """ Create a new Grid and add it as a child widget. Parameters ---------- row : int The row in which to add the widget (0 is the topmost row) col : int The column in which to add the widget (0 is the leftmost column) row_span : int The number of rows to be occupied by this widget. Default is 1. col_span : int The number of columns to be occupied by this widget. Default is 1. **kwargs : dict Keyword arguments to pass to the new `Grid`. """ from .grid import Grid grid = Grid(**kwargs) return self.add_widget(grid, row, col, row_span, col_span) def add_view(self, row=None, col=None, row_span=1, col_span=1, **kwargs): """ Create a new ViewBox and add it as a child widget. Parameters ---------- row : int The row in which to add the widget (0 is the topmost row) col : int The column in which to add the widget (0 is the leftmost column) row_span : int The number of rows to be occupied by this widget. Default is 1. col_span : int The number of columns to be occupied by this widget. Default is 1. **kwargs : dict Keyword arguments to pass to `ViewBox`. """ view = ViewBox(**kwargs) return self.add_widget(view, row, col, row_span, col_span) def next_row(self): self._next_cell = [self._next_cell[0] + 1, 0] @property def grid_size(self): rvals = [ widget[0] + widget[2] for widget in self._grid_widgets.values() ] cvals = [ widget[1] + widget[3] for widget in self._grid_widgets.values() ] return max(rvals + [0]), max(cvals + [0]) @property def layout_array(self): locs = -1 * np.ones(self.grid_size, int) for key in self._grid_widgets.keys(): r, c, rs, cs = self._grid_widgets[key][:4] locs[r:r + rs, c:c + cs] = key return locs def __repr__(self): return (('<Grid at %s:\n' % hex(id(self))) + str(self.layout_array + 1) + '>') @staticmethod def _add_total_width_constraints(solver, width_grid, _var_w): for ws in width_grid: width_expr = ws[0] for w in ws[1:]: width_expr += w solver.addConstraint(width_expr == _var_w) @staticmethod def _add_total_height_constraints(solver, height_grid, _var_h): for hs in height_grid: height_expr = hs[0] for h in hs[1:]: height_expr += h solver.addConstraint(height_expr == _var_h) @staticmethod def _add_gridding_width_constraints(solver, width_grid): # access widths of one "y", different x for ws in width_grid.T: for w in ws[1:]: solver.addConstraint(ws[0] == w) @staticmethod def _add_gridding_height_constraints(solver, height_grid): # access heights of one "y" for hs in height_grid.T: for h in hs[1:]: solver.addConstraint(hs[0] == h) @staticmethod def _add_stretch_constraints(solver, width_grid, height_grid, grid_widgets, widget_grid): xmax = len(height_grid) ymax = len(width_grid) stretch_widths = [[] for _ in range(0, ymax)] stretch_heights = [[] for _ in range(0, xmax)] for (y, x, ys, xs, widget) in grid_widgets.values(): for ws in width_grid[y:y + ys]: total_w = np.sum(ws[x:x + xs]) for sw in stretch_widths[y:y + ys]: sw.append((total_w, widget.stretch[0])) for hs in height_grid[x:x + xs]: total_h = np.sum(hs[y:y + ys]) for sh in stretch_heights[x:x + xs]: sh.append((total_h, widget.stretch[1])) for (x, xs) in enumerate(widget_grid): for (y, widget) in enumerate(xs): if widget is None: stretch_widths[y].append((width_grid[y][x], 1)) stretch_heights[x].append((height_grid[x][y], 1)) for sws in stretch_widths: if len(sws) <= 1: continue comparator = sws[0][0] / sws[0][1] for (stretch_term, stretch_val) in sws[1:]: solver.addConstraint((comparator == stretch_term / stretch_val) | 'weak') for sws in stretch_heights: if len(sws) <= 1: continue comparator = sws[0][0] / sws[0][1] for (stretch_term, stretch_val) in sws[1:]: solver.addConstraint((comparator == stretch_term / stretch_val) | 'weak') @staticmethod def _add_widget_dim_constraints(solver, width_grid, height_grid, total_var_w, total_var_h, grid_widgets): assert (total_var_w is not None) assert (total_var_h is not None) for ws in width_grid: for w in ws: solver.addConstraint(w >= 0, ) for hs in height_grid: for h in hs: solver.addConstraint(h >= 0) for (_, val) in grid_widgets.items(): (y, x, ys, xs, widget) = val for ws in width_grid[y:y + ys]: total_w = np.sum(ws[x:x + xs]) # assert(total_w is not None) solver.addConstraint(total_w >= widget.width_min) if widget.width_max is not None: solver.addConstraint(total_w <= widget.width_max) else: solver.addConstraint(total_w <= total_var_w) for hs in height_grid[x:x + xs]: total_h = np.sum(hs[y:y + ys]) solver.addConstraint(total_h >= widget.height_min) if widget.height_max is not None: solver.addConstraint(total_h <= widget.height_max) else: solver.addConstraint(total_h <= total_var_h) def _recreate_solver(self): self._solver.reset() self._var_w = Variable("w_rect") self._var_h = Variable("h_rect") self._solver.addEditVariable(self._var_w, 'strong') self._solver.addEditVariable(self._var_h, 'strong') rect = self.rect.padded(self.padding + self.margin) ymax, xmax = self.grid_size self._solver.suggestValue(self._var_w, rect.width) self._solver.suggestValue(self._var_h, rect.height) self._solver.addConstraint(self._var_w >= 0) self._solver.addConstraint(self._var_h >= 0) # self._height_stay = None # self._width_stay = None # add widths self._width_grid = np.array([[ Variable("width(x: %s, y: %s)" % (x, y)) for x in range(0, xmax) ] for y in range(0, ymax)]) # add heights self._height_grid = np.array([[ Variable("height(x: %s, y: %s" % (x, y)) for y in range(0, ymax) ] for x in range(0, xmax)]) # setup stretch stretch_grid = np.zeros(shape=(xmax, ymax, 2), dtype=float) stretch_grid.fill(1) for (_, val) in self._grid_widgets.items(): (y, x, ys, xs, widget) = val stretch_grid[x:x + xs, y:y + ys] = widget.stretch # even though these are REQUIRED, these should never fail # since they're added first, and thus the slack will "simply work". Grid._add_total_width_constraints(self._solver, self._width_grid, self._var_w) Grid._add_total_height_constraints(self._solver, self._height_grid, self._var_h) try: # these are REQUIRED constraints for width and height. # These are the constraints which can fail if # the corresponding dimension of the widget cannot be fit in the # grid. Grid._add_gridding_width_constraints(self._solver, self._width_grid) Grid._add_gridding_height_constraints(self._solver, self._height_grid) except UnsatisfiableConstraint: self._need_solver_recreate = True # these are WEAK constraints, so these constraints will never fail # with a RequiredFailure. Grid._add_stretch_constraints(self._solver, self._width_grid, self._height_grid, self._grid_widgets, self._widget_grid) Grid._add_widget_dim_constraints(self._solver, self._width_grid, self._height_grid, self._var_w, self._var_h, self._grid_widgets) self._solver.updateVariables() def _update_child_widget_dim(self): # think in terms of (x, y). (row, col) makes code harder to read ymax, xmax = self.grid_size if ymax <= 0 or xmax <= 0: return rect = self.rect # .padded(self.padding + self.margin) if rect.width <= 0 or rect.height <= 0: return if self._need_solver_recreate: self._need_solver_recreate = False self._recreate_solver() # we only need to remove and add the height and width constraints of # the solver if they are not the same as the current value h_changed = abs(rect.height - self._var_h.value()) > 1e-4 w_changed = abs(rect.width - self._var_w.value()) > 1e-4 if h_changed: self._solver.suggestValue(self._var_h, rect.height) if w_changed: self._solver.suggestValue(self._var_w, rect.width) if h_changed or w_changed: self._solver.updateVariables() value_vectorized = np.vectorize(lambda x: x.value()) for (_, val) in self._grid_widgets.items(): (row, col, rspan, cspan, widget) = val width = np.sum( value_vectorized(self._width_grid[row][col:col + cspan])) height = np.sum( value_vectorized(self._height_grid[col][row:row + rspan])) if col == 0: x = 0 else: x = np.sum(value_vectorized(self._width_grid[row][0:col])) if row == 0: y = 0 else: y = np.sum(value_vectorized(self._height_grid[col][0:row])) if isinstance(widget, ViewBox): widget.rect = Rect(x, y, width, height) else: widget.size = (width, height) widget.pos = (x, y) @property def _widget_grid(self): ymax, xmax = self.grid_size widget_grid = np.array([[None for _ in range(0, ymax)] for _ in range(0, xmax)]) for (_, val) in self._grid_widgets.items(): (y, x, ys, xs, widget) = val widget_grid[x:x + xs, y:y + ys] = widget return widget_grid