hanhou/patchseq

1"""2Size mapping utilities for the scatter plot.3"""4 5from typing import Union6 7import pandas as pd8from bokeh.models import ColumnDataSource9 10 11class SizeMapping:12    """Handles size mapping for scatter plots."""13 14    def __init__(self, df_meta: pd.DataFrame):15        """Initialize with metadata dataframe."""16        self.df_meta = df_meta17 18    def determine_size_mapping(19        self,20        size_mapping: str,21        source: ColumnDataSource,22        min_size: int = 10,23        max_size: int = 20,24        gamma: float = 1,25    ) -> Union[int, str]:26        """27        Determine the size mapping for the scatter plot.28 29        Args:30            size_mapping: Column name to use for size mapping31            source: ColumnDataSource to add size values to32            min_size: Minimum marker size33            max_size: Maximum marker size34            gamma: Gamma value for nonlinear size scaling35 36        Returns:37            Either a fixed size or the name of the size column in the source38        """39        if size_mapping == "None":40            return 1041 42        if size_mapping in self.df_meta.columns:43            numeric_data = pd.Series(pd.to_numeric(self.df_meta[size_mapping], errors="coerce"))44            if not numeric_data.isna().all():45                # Get the min and max of the numeric data46                p5 = numeric_data.quantile(0.00)47                p95 = numeric_data.quantile(1.00)48 49                # Map the normalized values to sizes between min and max with50                # gamma control for nonlinearity51                normalized_values = ((numeric_data - p5) / (p95 - p5)).clip(0, 1)52                normalized_sizes = min_size + (normalized_values**gamma) * (max_size - min_size)53 54                # Replace NaN values with the minimum size55                normalized_sizes = normalized_sizes.fillna(5)  # Fixed size for NaN values56 57                # Add the size values to the source data58                source.data["size_values"] = normalized_sizes59                return "size_values"60 61        return 1062