Bugfix data-explorer images (#382)

PR that fixes the issue with rendering images in the explorer. Images
currently have type `binary` and not `object`. This was changed to
render them properly.

Caveat: not all `binary` types will correspond to images so we will need
to implement another logic for image detection but this fix should do
for now.

Fixed another small bug related to errors in rendering the numerical
explorer when no numeric columns are present

data_explorer/app/data.py

@@ -1,7 +1,7 @@
 """This file contains data loading logic"""
 import json
 import logging
-from typing import List, Tuple
+from typing import List
 from urllib.parse import urlparse
 
 import dask.dataframe as dd

data_explorer/app/main.py

@@ -1,13 +1,16 @@
 """Main file of the data explorer interface"""
 import logging
 
+import dask
 import streamlit as st
 from data import load_dataframe
 from table import get_image_fields, get_numeric_fields
 from widgets import (build_explorer_table, build_image_explorer,
                      build_numeric_analysis_plots,
                      build_numeric_analysis_table, build_sidebar)
 
+dask.config.set({"dataframe.convert-string": False})
+
 LOGGER = logging.getLogger(__name__)
 # streamlit wide
 st.set_page_config(layout="wide")
@@ -31,7 +34,6 @@
         # extract image and numeric columns
         image_fields = get_image_fields(fields)
         numeric_fields = get_numeric_fields(fields)
-
         # build tabs
         tab_explorer, tab_numeric, tab_images = st.tabs(
             ["Data explorer", "Numerical analysis", "Image explorer"]

data_explorer/app/table.py

@@ -45,7 +45,7 @@ def get_image_fields(fields: Dict[str, str]) -> List[str]:
     # check which of the columns contain byte data
     image_fields = []
     for k, v in fields.items():
-        if v == "object":
+        if v == "binary":
             image_fields.append(k)
     return image_fields
 

data_explorer/app/widgets.py

@@ -78,7 +78,7 @@ def build_sidebar() -> Tuple[Optional[str], Optional[str], Optional[Dict]]:
 
 
 def build_explorer_table(
-    dataframe: Union[dd.DataFrame, pd.DataFrame], image_fields: List[str]
+        dataframe: Union[dd.DataFrame, pd.DataFrame], image_fields: List[str]
 ) -> None:
     """Build the dataframe explorer table.
 
@@ -129,7 +129,7 @@ def build_explorer_table(
 
 
 def build_numeric_analysis_table(
-    dataframe: Union[dd.DataFrame, pd.DataFrame], numeric_fields: List[str]
+        dataframe: Union[dd.DataFrame, pd.DataFrame], numeric_fields: List[str]
 ) -> None:
     """Build the numeric analysis table.
 
@@ -138,7 +138,9 @@ def build_numeric_analysis_table(
         numeric_fields (List[str]): list of numeric fields
     """
     # check if there are numeric fields
-    if len(numeric_fields) > 0:
+    if len(numeric_fields) == 0:
+        st.warning("There are no numeric fields in this subset")
+    else:
         st.write("## Numerical statistics")
 
         # make numeric statistics table
@@ -159,25 +161,29 @@ def build_numeric_analysis_table(
 
 
 def build_numeric_analysis_plots(
-    dataframe: Union[dd.DataFrame, pd.DataFrame], numeric_fields: List[str]
+        dataframe: Union[dd.DataFrame, pd.DataFrame], numeric_fields: List[str]
 ) -> None:
     """Build the numeric analysis plots.
 
     Args:
         dataframe (Union[dd.DataFrame, pd.DataFrame]): dataframe to explore
         numeric_fields (List[str]): list of numeric fields
     """
-    st.write("## Show numeric distributions")
+    # check if there are numeric fields
+    if len(numeric_fields) == 0:
+        st.warning("There are no numeric fields in this subset")
+    else:
+        st.write("## Show numeric distributions")
 
-    # choose a numeric field in dropdown
-    cols = st.columns(2)
-    with cols[0]:
-        numeric_field = st.selectbox("Field", numeric_fields)
-    with cols[1]:
-        plot_type = st.selectbox("Plot type",
-                                 ["histogram", "violin", "density", "categorical"])
+        # choose a numeric field in dropdown
+        cols = st.columns(2)
+        with cols[0]:
+            numeric_field = st.selectbox("Field", numeric_fields)
+        with cols[1]:
+            plot_type = st.selectbox("Plot type",
+                                     ["histogram", "violin", "density", "categorical"])
 
-    make_numeric_plot(dataframe, numeric_field, plot_type)
+        make_numeric_plot(dataframe, numeric_field, plot_type)
 
 
 def build_image_explorer(dataframe: dd.DataFrame, image_fields: List[str]):
@@ -188,12 +194,13 @@ def build_image_explorer(dataframe: dd.DataFrame, image_fields: List[str]):
         dataframe (dd.DataFrame): dataframe to explore
         image_fields (List[str]): list of image fields
     """
-    st.write("## Image explorer")
-    st.write("In this table, you can explore the images")
 
     if len(image_fields) == 0:
         st.warning("There are no image fields in this subset")
     else:
+        st.write("## Image explorer")
+        st.write("In this table, you can explore the images")
+
         image_field = st.selectbox("Image field", image_fields)
 
         images = dataframe[image_field].compute()