Coverage for sparkle/platform/latex.py: 34%

1"""Helper classes/method for LaTeX and bibTeX.""" 

2 

3import math 

4 

5import numpy as np 

6import pandas as pd 

7import plotly 

8import plotly.express as px 

9import pylatex as pl 

10import kaleido 

11 

12kaleido.get_chrome_sync() # Ensure chrome is available for Kaleido 

13 

14 

15class AutoRef(pl.base_classes.CommandBase): 

16 """AutoRef command for PyLateX.""" 

17 

18 _latex_name = "autoref" 

19 packages = [pl.Package("hyperref")] 

20 

21 

22def comparison_plot( 

23 data_frame: pd.DataFrame, title: str = None 

24) -> plotly.graph_objects.Figure: 

25 """Creates a comparison plot from the given data frame. 

26 

27 The first column is used for the x axis, the second for the y axis. 

28 

29 Args: 

30 data_frame: The data frame with the data 

31 x_label: The label for the x axis 

32 y_label: The label for the y axis 

33 title: The title of the plot 

34 output_plot: The path where the plot should be written to 

35 

36 Returns: 

37 The plot object 

38 """ 

39 from scipy import stats 

40 

41 # Determine if data is log scale, linregress tells us how linear the data is 

42 linregress = stats.linregress( 

43 data_frame[data_frame.columns[0]].to_numpy(), 

44 data_frame[data_frame.columns[1]].to_numpy(), 

45 ) 

46 log_scale = not (linregress.rvalue > 0.65 and linregress.pvalue < 0.05) 

47 

48 if log_scale and (data_frame < 0).any(axis=None): 

49 # Log scale cannot deal with negative and zero values, set to smallest non zero 

50 data_frame[data_frame < 0] = np.nextafter(0, 1) 

51 

52 # Maximum value should come from the objective? 

53 min_value, max_value = data_frame.min(axis=None), data_frame.max(axis=None) 

54 # Slightly more than min/max for interpretability 

55 if log_scale: # Take next step on log scale 

56 max_value = 10 ** math.ceil(math.log(max_value, 10)) 

57 plot_range = (min_value, max_value) 

58 else: # Take previous/next step on linear scale 

59 order_magnitude = math.ceil(math.log(max_value, 10)) 

60 next_step_max = math.ceil(max_value / (10 ** (order_magnitude - 1))) * 10 ** ( 

61 order_magnitude - 1 

62 ) 

63 plot_range = (0, next_step_max) if min_value > 0 else (min_value, next_step_max) 

64 fig = px.scatter( 

65 data_frame=data_frame, 

66 x=data_frame.columns[0], 

67 y=data_frame.columns[1], 

68 range_x=plot_range, 

69 range_y=plot_range, 

70 title=title, 

71 log_x=log_scale, 

72 log_y=log_scale, 

73 width=1000, 

74 height=1000, 

75 ) 

76 # Add dividing diagonal 

77 fig.add_shape( 

78 type="line", 

79 x0=0, 

80 y0=0, 

81 x1=max_value, 

82 y1=max_value, 

83 line=dict(color="grey", dash="dot", width=1), 

84 ) 

85 # Add maximum lines 

86 fig.add_shape( 

87 type="line", 

88 opacity=0.7, 

89 x0=0, 

90 y0=max_value, 

91 x1=max_value, 

92 y1=max_value, 

93 line=dict(color="red", width=1.5, dash="longdash"), 

94 ) 

95 fig.add_shape( 

96 type="line", 

97 opacity=0.7, 

98 x0=max_value, 

99 y0=0, 

100 x1=max_value, 

101 y1=max_value, 

102 line=dict(color="red", width=0.5, dash="longdash"), 

103 ) 

104 fig.update_traces(marker=dict(color="RoyalBlue", symbol="x")) 

105 fig.update_layout(plot_bgcolor="white", autosize=False, width=1000, height=1000) 

106 minor = dict(ticks="inside", ticklen=6, showgrid=True) if log_scale else None 

107 # Tick every 10^(log(max)) / 10 for linear scale, log scale is resolved by plotly 

108 dtick = 1 if log_scale else 10 ** (math.ceil(math.log(max_value, 10)) - 1) 

109 fig.update_xaxes( 

110 mirror=True, 

111 tickmode="linear", 

112 ticks="outside", 

113 tick0=0, 

114 minor=minor, 

115 dtick=dtick, 

116 showline=True, 

117 linecolor="black", 

118 gridcolor="lightgrey", 

119 ) 

120 fig.update_yaxes( 

121 mirror=True, 

122 tickmode="linear", 

123 ticks="outside", 

124 tick0=0, 

125 minor=minor, 

126 dtick=dtick, 

127 showline=True, 

128 linecolor="black", 

129 gridcolor="lightgrey", 

130 ) 

131 return fig