Improved data handling in memory

Still suboptimal but performs better than previous implementation

src/funki/app.py

@@ -23,8 +23,6 @@
 app.layout = html.Div(
     children=[
         dcc.Store(id='data', storage_type=storage_type),
-        #dcc.Store(id='ann-data', storage_type=storage_type),
-        #dcc.Store(id='proc-data', storage_type=storage_type),
         html.Div(
             html.Img(
                 src='assets/logos/funki_logo.svg',

src/funki/pages/clustering.py

@@ -149,7 +149,7 @@ def update_param_panel(embedding, data):
         children.extend(['Select embedding parameters: ', html.Br(), html.Br()])
 
     if embedding == 'tsne':
-        max_per = len(data['index']) - 1 if data else 50
+        max_per = len(data['obs_names']) - 1 if data else 50
         min_per = 1 if max_per < 10 else 5
         step = 1 if max_per < 10 else 5
 
@@ -220,7 +220,7 @@ def apply_clustering(n_clicks, data, algorithm, resolution):
 )
 def update_dropdown(data):
     try:
-        options = list(data['obs']['records'][0].keys())
+        options = list(data['obs']['var_names'])
 
     except (KeyError, TypeError):
         options = []
@@ -248,7 +248,6 @@ def plot_embedding(n_clicks, data, embedding, param_panel, color):
     elif embedding == 'tsne':
         # TODO: There is probably a more elegant way to do this
         perplexity = param_panel[-1]['props']['value']
-        print(perplexity)
         fig = fpl.plot_tsne(dset, perplexity=perplexity, color=color)
 
     elif embedding == 'umap':

src/funki/utils/utils.py

@@ -32,13 +32,18 @@ def parse_contents(content, filename):
     return df
 
 def dataframe_to_serial(df):
-    return {'index': df.index, 'records': df.to_dict('records')}
+    return {
+        'obs_names': df.index,
+        'var_names': df.columns,
+        'X': df.values.tolist()
+    }
 
 def serial_to_dataframe(data):
-    df = pd.DataFrame(data['records'])
-
-    if 'index' in data.keys():
-        df.index = data['index']
+    df = pd.DataFrame(
+        np.array(data['X']),
+        index=data['obs_names'] if 'obs_names' in data.keys() else None,
+        columns=data['var_names'] if 'var_names' in data.keys() else None,
+    )
 
     return df
 
@@ -60,6 +65,7 @@ def serial_to_dataset(data):
 
 def dataset_to_serial(dset):
     data = dataframe_to_serial(dset.to_df())
+
     data.update({
         k: dataframe_to_serial(getattr(dset, k))
         for k in ('obs', 'var')