Dataset viewer issues

I was looking through the dataset viewer and I like it a lot. Version numbers, citation information, everything’s there! I’ve spotted a few issues/bugs though:

• the URL is still under nlp, perhaps an alias for datasets can be made
• when I remove a feature (and the feature list is empty), I get an error. This is probably expected, but perhaps a better error message can be shown to the user

IndexError: list index out of range
Traceback:
File "/home/sasha/streamlit/lib/streamlit/ScriptRunner.py", line 322, in _run_script
    exec(code, module.__dict__)
File "/home/sasha/nlp-viewer/run.py", line 316, in <module>
    st.table(style)
File "/home/sasha/streamlit/lib/streamlit/DeltaGenerator.py", line 122, in wrapped_method
    return dg._enqueue_new_element_delta(marshall_element, delta_type, last_index)
File "/home/sasha/streamlit/lib/streamlit/DeltaGenerator.py", line 367, in _enqueue_new_element_delta
    rv = marshall_element(msg.delta.new_element)
File "/home/sasha/streamlit/lib/streamlit/DeltaGenerator.py", line 120, in marshall_element
    return method(dg, element, *args, **kwargs)
File "/home/sasha/streamlit/lib/streamlit/DeltaGenerator.py", line 2944, in table
    data_frame_proto.marshall_data_frame(data, element.table)
File "/home/sasha/streamlit/lib/streamlit/elements/data_frame_proto.py", line 54, in marshall_data_frame
    _marshall_styles(proto_df.style, df, styler)
File "/home/sasha/streamlit/lib/streamlit/elements/data_frame_proto.py", line 73, in _marshall_styles
    translated_style = styler._translate()
File "/home/sasha/.local/share/virtualenvs/lib-ogGKnCK_/lib/python3.7/site-packages/pandas/io/formats/style.py", line 351, in _translate
    * (len(clabels[0]) - len(hidden_columns))

• there seems to be an encoding issue in the default view, the dataset examples are shown as raw monospace text, without a decent encoding. That makes it hard to read for languages that use a lot of special characters. Take for instance the cs-en WMT19 set. This problem goes away when you enable “List view”, because then some syntax highlighteris used, and the special characters are coded correctly.

[lhoestq]

Thanks for reporting ! cc @srush for the empty feature list issue and the encoding issue cc @julien-c maybe we can update the url and just have a redirection from the old url to the new one ?

[mmkarlosmm]

‏⠀‏‏‏⠀‏‏‏⠀ ‏⠀