Update app.py

app.py

@@ -1,72 +1,68 @@
 import pandas as pd
 import panel as pn
 import numpy as np
+import hvplot.pandas
 from bokeh.models.widgets.tables import NumberFormatter, BooleanFormatter
 
-pn.extension("tabulator")
+import matplotlib.pyplot as plt
+import seaborn as sns
 
 
-### Beispiel DataFrame
+sns.set(style='whitegrid', context='notebook')
+#plt.rcParams["font.family"] = "Latin Modern Sans"
 
-df = pd.read_pickle(f"cossims_plural_slicewidth=1.pkl")
-if "boson" in df.index:
-    df = df.drop("boson")
-if "bosons" in df.index:
-    df = df.drop("bosons")
-if "boson_s" in df.index:
-    df = df.drop("boson_s")
 
-df["sum"] = df.sum(axis=1)
-df = df.loc[:, 1948:] #1948
-#df = df.fillna(0).round(3)
+### Data
 
-df = df.sort_values("sum", ascending=False)
-df = df.head(20)
-df.index.name = "token"
+def load_data(sel_token):
+    if sel_token == "virtual":
+        df = pd.read_pickle("server_data/virtual_fulltexts_cossim_all_slice=1.pkl")
+        df = df.drop("virtual")
+    elif sel_token == "boson":
+        df = pd.read_pickle("server_data/cossims_plural_slicewidth=1.pkl")
+    elif sel_token == "intermediate":
+        df = pd.read_pickle("server_data/intermediate_fulltexts_cossim_all_slice=1.pkl")
+        df = df.drop("intermediate")
 
+    topn = 1000
+    df["sum"] = df.sum(axis=1)
+    df = df.astype("float32")
+    df = df.round(3)
 
+    df = df.sort_values("sum", ascending=False)
+    df = df.head(topn)
+    df.index.name = "token"
+    return df
 
-outpath = "../results/bosons/cossims_plural_bosons.html"
-
-### functionality
-
-#tabulator_formatters = {"token" : {"title": "token", "type" : "textarea"}} # Formatiert die Zellen
-cell_formatters = {col : NumberFormatter(format='0.000') for col in df.columns} # Formatiert die Zellen
-header_filters={'token': {'type': 'input', 'func': 'like', 'placeholder': 'search'}} # Für Suche, etc.
+sel_token = pn.widgets.Select(
+    name="Select dataset",
+    value="virtual",
+    options=["virtual", "intermediate", "boson"],
+    #description="Select the base token ",
+)
 
-### style
 
-theme = "modern" # 'default', 'site', 'simple', 'midnight', 'modern', 'bootstrap', 'bootstrap4', 'materialize', 'bulma', 'semantic-ui', or 'fast'
-page_size = 50 # Number of rows
-page = 1 # Start on page x
-#configuration={"rowHeight" : 10*24}
-#widths = {'index' : 40} or {'index': '5%'}
+df = pn.rx(load_data)(sel_token=sel_token)
 
-#frozen_rows = list
-#frozen_columns = list
 
+### Table
 
-tabs = pn.widgets.Tabulator(
+table = pn.widgets.Tabulator(
     df,
 
     ### functionality
 
-    formatters= cell_formatters, #tabulator_formatters,
-    header_filters = header_filters,
-    selectable="toggle",
+    #formatters= {col : NumberFormatter(format='0.000') for col in df.columns}, #tabulator_formatters,
+    header_filters = {'token': {'type': 'input', 'func': 'like', 'placeholder': 'search'}},
+    selectable='checkbox',
 
     ### style
 
-    theme = theme,
-    page_size = page_size,
-    page = page,
+    theme = "modern", # 'default', 'site', 'simple', 'midnight', 'modern', 'bootstrap', 'bootstrap4', 'materialize', 'bulma', 'semantic-ui', or 'fast'
+    page_size = 8,
+    page = 1,
     frozen_columns = {"token" : "left", "sum" : "right"}, # Must give width, otherwise doesn't work!
     width=1800,
-    height=950,
-    #frozen_rows = [1],
-    #widths
-    #configuration = configuration,
-
     
 
     ### other
@@ -76,13 +72,66 @@ tabs = pn.widgets.Tabulator(
 )
 
 
+### Plot
+
+def make_fig():
+    fig, ax = plt.subplots(figsize=(12,4))
+    df_temp = load_data(sel_token.value)
+    if len(table.selection) > 0:
+        for i in table.selection:
+            df_temp.iloc[i][:-1].plot(ax=ax, label=df_temp.iloc[i].name, lw=2.2, marker=".")
+    #else:
+        #df.loc["particle"][:-1].plot(ax=ax, label="particle", lw=2.2, marker=".")
+        #plt.hist(np.random.random(10))
+    plt.ylabel("Cosine Similarity", fontsize=12)
+    plt.xlim()
+    plt.legend()
+    plt.close()
+
+    return fig
+
+def plot_data(event):
+
+    # selected rows as indices in table.selection
+    #token = df.iloc[table.selection[0]].name
+    #values = df.iloc[table.selection[0]][:-1]
+    canvas.loading = True
+    fig = make_fig()
+    canvas.object = fig
+    canvas.loading = False
+
+button = pn.widgets.Button(
+    name='Plot', 
+    button_type='primary', 
+    align="center",
+    width=100,
+    icon="snowman",
+)
+button.on_click(plot_data)
+
+
+canvas = pn.pane.Matplotlib(
+    make_fig(),
+    format="svg", 
+    #width=1000,
+    sizing_mode='stretch_width',
+    height=400,
+    tight=True)
+
+
+### Serve
+
+
 ACCENT = "teal"
 
 pn.template.FastListTemplate(
-    #title="Wind Turbine Dashboard",
-    #sidebar=[image, t_manu, p_year],
-    #main=[pn.Column(indicators, tabs, sizing_mode="stretch_both")],
-    main=[pn.Column(tabs, sizing_mode="stretch_both")],
-    #main_layout=None,
+    title="Cosine Similarity for selected tokens",
+    sidebar=[],
+    main=[pn.Column(
+            pn.Row(sel_token),
+            table, 
+            button, 
+            canvas)],
+    main_layout=None,
     accent=ACCENT,
 ).servable()