Now cld4py can add letters to seaborn/matplotlib figures

cld4py/__init__.py

@@ -0,0 +1,2 @@
+from .cld4py import assign_letters
+from .cld4py import plot_letters

cld4py/cld4py.py

@@ -0,0 +1,168 @@
+import string
+import pandas as pd
+
+
+def assign_letters(df, G1, G2, P, alpha=.05,
+           order=None, data=None, vals=None, group=None):
+    """
+    Function to apply compact letter display for pairwise contrasts.
+    Groups with no significant differences share a letter.
+    
+    Parameters:
+    
+      Required:
+        df    - dataframe with contrasts (pairwise comparisons).
+        G1, G2 - columns in contrasts df with compared groups.
+        P      - column in contrasts df with p-value (adjusted, right?).
+        
+      Optional:
+        alpha  - sigificance level (default 0.05).
+        order  - None (default), list or ['ascending', 'descending'].
+                 This parameter will define the order of assigned letters.
+                 None - alphabetical order will be applied.
+                 List - order of groups will be defined by that list.
+                 String 'ascending' or 'descending' requires parameters
+                 'data', 'values' and 'group' to order groups by the mean.
+        data   - dataframe with values that were compared to get contrasts.
+        vals   - column in data with compared values.
+        group  - column in data with group information.
+    """
+
+    #helper function to check p
+    def is_sign(lv1, lv2, df=df, alpha=alpha):
+        return df.loc[(df[G1]==lv1)&(df[G2]==lv2)|(df[G1]==lv2)&(df[G2]==lv1),P].iloc[0] < alpha
+
+    letters = string.ascii_lowercase
+    df[P] = df[P].apply(pd.to_numeric)
+
+    #define order
+    if order == None:
+        order = sorted(set(df[G1].tolist() + df[G2].tolist()))
+    if order in ['ascending', 'descending']:
+        asc = order=='ascending'
+        data[vals] = data[vals].apply(pd.to_numeric)
+        order = pd.DataFrame(data.groupby(group)[vals].mean()).sort_values(vals, ascending=asc).index.tolist()
+
+    #assign letters
+    draft, sets = {}, []
+    for i, l1 in enumerate(order):
+        draft.update({i: {l1}})
+        for l2 in order: 
+            if l1 != l2 and not any([is_sign(l1,l2)]+[is_sign(l,l2) for l in draft[i]]):
+                    draft[i].add(l2)
+    [sets.append(v) for v in draft.values() if v not in sets]
+    cld = pd.DataFrame(columns=['Group', 'Letters'])
+    for i,l in enumerate(order):
+        cld.loc[i, ['Group', 'Letters']] = l, ''.join([letters[j] for j,s in enumerate(sets) if l in s])
+    return cld.set_index('Group')
+
+
+def plot_letters(cld, data, vals, group, figax, 
+                axis='x', plot='boxplot', pos='upper',
+                pad=1, c='black', fs=None, lim=0):
+    """
+    Function to plot CLD letters for sns boxplot, violinplot, barplot or swarmplot.
+    Groups with no significant differences share a letter.
+    
+    Parameters:
+    
+      Required:
+        cld    - dataframe or dictionary with groups and letters. If df then groups 
+                 should be in the index and letters in the "Letters" column.
+        data   - dataframe with values that were plotted.
+        vals   - column in data with compared values.
+        group  - column in data with group information.
+        figax  - matplotlib or sns figure or ax with plot.
+        
+      Optional:
+        axis   - axis with plotted groups: "x" or "y" (default "x").
+        plot   - plot type: "boxplot", "violinplot", "barplot" or "swarmplot"
+                 (default "boxplot").
+        pos    - letters position: "upper", "lower", "top" or "bottom" 
+                 (default: "upper"). If axis = "y", "upper" and "top" will be 
+                 plotted on the right side, "lower" and "bottom" - on the left.
+        pad    - distance (% of data range) to the plotted group object (default 1).     
+        c      - color of letters.
+        fs     - fontsize of letters.
+        lim    - increase axes limits, expressed in "pad" (see above) values (default 0)
+    """
+
+    #import
+    import seaborn as sns
+    import matplotlib.pyplot as plt
+    from matplotlib.cbook import boxplot_stats
+
+    types = ['boxplot', 'violinplot', 'barplot', 'swarmplot']
+    labels = figax.get_xticklabels() if axis=="x" else figax.get_yticklabels()
+    step = (data[vals].max() - data[vals].min())/100 * pad
+
+    #check parameters
+    if axis not in "xy":
+        raise ValueError('axis should by either "x" or "y"')
+    if plot not in types:
+        raise ValueError(f'"{plot}" is not in the list of supported types')
+    if pos not in ["upper", "lower", "top", "bottom"]:
+        raise ValueError('pos should by either "upper", "lower", "top" or "bottom"')
+    if not any([isinstance(cld, pd.DataFrame), isinstance(cld, dict)]):
+        raise ValueError('cld should by either dataframe or dictionary')
+    if fs ==  None:
+        fs = labels[0].get_fontsize()
+
+    #set limits
+    if axis=="y":
+        va = 'center'
+        lims = figax.get_xlim()
+        figax.set_xlim(min(lims)-lim*step, max(lims)+lim*step)
+    if axis=="x":
+        lims = figax.get_ylim()
+        figax.set_ylim(min(lims)-lim*step, max(lims)+lim*step)
+
+    #plot letters
+    pos_dict = {
+        'top': max(figax.get_ylim())-step if axis=='x' else max(figax.get_xlim())-step,
+        'bottom': min(figax.get_ylim())+step if axis=='x' else min(figax.get_xlim())+step}
+
+    if axis=='x':
+        ha = 'center'
+        va = 'bottom' if pos in ['upper', 'bottom'] else 'top'
+    if axis=='y':
+        va = 'center'
+        ha = 'left' if pos in ['upper', 'bottom'] else 'right'
+    if not isinstance(cld, dict):
+        cld = {i: cld.loc[i, 'Letters'] for i in cld.index}
+
+
+    for i, label in enumerate(labels):
+        df = data.loc[data[group]==label.get_text()]
+        rel_step = (df[vals].max() - df[vals].min())/1.7
+        x, y = label.get_position()
+
+        #boxplot
+        if plot == 'boxplot':
+            pos_dict.update({'upper': boxplot_stats(df[vals])[0]['whishi'] + step})
+            pos_dict.update({'lower': boxplot_stats(df[vals])[0]['whislo'] - step})
+
+        #violinplot
+        if plot == 'violinplot':
+            pos_dict.update({'upper': df[vals].max() + rel_step + step})
+            pos_dict.update({'lower': df[vals].min() - rel_step - step})
+
+        #barplot
+        if plot == 'barplot':
+            line = figax.lines[i]
+            gain = line.get_ydata() if axis=="x" else line.get_xdata()
+            pos_dict.update({'upper': step + abs(max(gain))})
+            pos_dict.update({'lower': -step + abs(min(gain))})
+
+        #swarmplot
+        if plot == 'swarmplot':
+            pos_dict.update({'upper': df[vals].max() + step})
+            pos_dict.update({'lower': df[vals].min() - step})
+
+        xpos = x if axis=="x" else pos_dict[pos]
+        ypos = y if axis=="y" else pos_dict[pos]   
+        if plot == 'barplot':
+            mn, mx = 0, max(lims)+lim*step
+            [figax.set_ylim(mn, mx) if axis=='x' else figax.set_xlim(mn, mx)]
+
+        figax.text(xpos, ypos, cld[label.get_text()], size=fs, color=c, ha=ha, va=va,)

pyproject.toml

@@ -4,11 +4,11 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "cld4py"
-version = "1.0.1"
+version = "1.1.0"
 authors = [
   { name="Timur Yergaliyev", email="[email protected]" },
 ]
-description = "Python 3 function to assign compact letter display letters"
+description = "Python 3 function to assign compact letter display letters and for plotting them with matplotlib/seaborn"
 readme = "README.md"
 requires-python = ">=3.7"
 classifiers = [