KeyError `warnings`

anorkos · April 11, 2020, 7:49pm

Hi everyone!

I have half a day streamlit experience and I have the following issue.
The first time the server runs everything looks good. But then, when I make a change and the watcher reloads the page, I get this error

The code line is like this
df = df.apply(lambda x: robust_scale(x), axis=1, raw=True)

Outside streamlit code runs fine.

Thank you in advance.

Jonathan_Rhone · April 14, 2020, 6:35pm

Hi @anorkos, welcome to the forum

Could you share sample code that we can copy/paste and run to reproduce your error?

anorkos · April 14, 2020, 8:04pm

@Jonathan_Rhone thank you! Of course.

import streamlit as st
import pandas as pd
import numpy as np
from sklearn.preprocessing import RobustScaler

df = pd.DataFrame(np.random.rand(10, 7))
df

data_scaled = RobustScaler().fit_transform(df.to_numpy())
df_scaled = pd.DataFrame(data_scaled)
df_scaled

Run streamlit. it should display both tables at first. Then make a change in the file, let the watcher update the page, and you should see the error.

Jonathan_Rhone · April 14, 2020, 9:20pm

I wasn’t able to reproduce the issue.

Could you share the following?


- Streamlit version: (get it with `$ streamlit version`)
- Python version: (get it with `$ python --version`)
- Using Conda? PipEnv? PyEnv? Pex?
- OS version:
- Browser version:

Also, is warnings.py your own code? Could you try renaming that file to something else?

anorkos · April 15, 2020, 6:57pm

Streamlit Version: 0.57.3
Python Version: 3.7.6
PipEnv
Windows 10 Pro 1909
Chrome Version 81.0.4044.113

No warnings.py is from the Python library https://docs.python.org/3/library/warnings.html. It’s inside env/Lib.

It seems that it’s a warning thrown by the scikit library. The question is, why is it handled differently by streamlit at first load and subsequent loads?

matthew_segal · May 4, 2020, 4:03am

The streamlit library is doing some funny business with sys.modules

At some point “warnings” is getting deleted from sys.modules

This is the bottom of the stack trace for the same issue as OP

File "/home/matt/code/monash/autumn/env/lib/python3.6/site-packages/matplotlib/cbook/deprecation.py", line 407, in _suppress_matplotlib_deprecation_warning
    with warnings.catch_warnings():File "/home/matt/code/monash/autumn/env/lib/python3.6/warnings.py", line 437, in __init__
    self._module = sys.modules['warnings'] if module is None else module

Something in streamlit is deleting “warnings” from sys.modules

matthew_segal · May 5, 2020, 5:47am

10 bucks says it’s this line of code

github.com

streamlit/streamlit/blob/018f49f5ddb0fa0119c17aaf2bff1bc7f25f70e1/lib/streamlit/watcher/LocalSourcesWatcher.py#L114


    # In principle, for reloading a given module, we only need to unload
    # the module itself and all of the modules which import it (directly
    # or indirectly) such that when we exec the application code, the
    # changes are reloaded and reflected in the running application.
    #
    # However, determining all import paths for a given loaded module is
    # non-trivial, and so as a workaround we simply unload all watched
    # modules.
    for wm in self._watched_modules.values():
        if wm.module_name is not None and wm.module_name in sys.modules:
            del sys.modules[wm.module_name]


    self._on_file_changed()


def close(self):
    for wm in self._watched_modules.values():
        wm.watcher.close()
    self._watched_modules = {}
    self._is_closed = True


def _register_watcher(self, filepath, module_name):

matthew_segal · May 5, 2020, 5:49am

I think this fixes it on my machine:

        for wm in self._watched_modules.values():
        if wm.module_name is not None and wm.module_name in sys.modules:
            if wm.module_name != "warnings":
                del sys.modules[wm.module_name]

randyzwitch · May 5, 2020, 1:39pm

Hi @matthew_segal, welcome to the community!

Thanks for taking a look at this bug. Would you mind submitting this as a PR to https://github.com/streamlit/streamlit?

matthew_segal · May 5, 2020, 11:26pm

Hi @randyzwitch. My fix is really a situation-specific hack that would not be acceptible in the Streamlit codebase since it does not cover all bases. I believe that more investigation of this issue will be required before implementing a fix. If it’s helpful I can move this discussion to a GitHub issue and describe my hacky fix. (and link to issue from here)

randyzwitch · May 6, 2020, 1:09pm

That’s fair. Moving this to an issue would be great, along with your hack, so that one of our engineers can evaluate why the line that you modified is there in the first place.

Thanks!

matthew_segal · May 9, 2020, 1:18am

Raised as an issue here: http://github.com/streamlit/streamlit/issues/1430

matthew_segal · May 9, 2020, 1:19am

By the way @randyzwitch the forum is blocking links to github as “spam” which is very counterproductive

Tiago_Augusto_Ferrei · June 5, 2020, 1:15pm

Hello!

I had the same problem here and applied the workaround below and it worked for me.

import warnings
warnings.filterwarnings(‘ignore’)

Regards.

robokage · July 8, 2021, 5:01am

Worked for me thanks a lot!!

leb_dev · April 30, 2022, 6:19am

this can be a great way to solve this issue

tagugo5 · December 15, 2022, 1:17pm

Hi everyone!!

I have an strange issue about Key Error: 'Warnings" as below

I tried making a chart as below and succeeded first time.

However, I revise a code and reload the page after saving a code, it happens this error as below.

I found that I can succeed to reflect a chart when I “run” through new Terminal

The sample code as below.

import streamlit as st
import numpy as np
import pandas as pd
st.title(‘Streamlit begining’)
st.write(‘DataFrame’)
df = pd.DataFrame(
np.random.rand(20, 3),
columns=[‘a’, ‘b’, ‘c’]
)
st.bar_chart(df)

Thank you in advance.

macronbell · December 19, 2022, 7:48am

Pandas KeyError occurs when we try to access some column/row label in our DataFrame that doesn’t exist. Usually, this error occurs when you misspell a column/row name or include an unwanted space before or after the column/row name… Before doing anything with the data frame, use print(df.columns) to see dataframe column exist or not.

print(df.columns)

I was getting a similar kind of error in one of my codes. Turns out, that particular index was missing from my data frame as I had dropped the empty dataframe 2 rows. If this is the case, you can do df.reset_index(inplace=True) and the error should be resolved.

Topic		Replies	Views
Error Dissabling Using Streamlit	5	12239	February 7, 2022
From the docs: Get Started: Main Concepts, pandas styler does not work? Using Streamlit	8	986	December 4, 2023
RuntimeError (watchdog ): dictionary changed size during iteration Using Streamlit debugging	0	83	July 4, 2025
How to solve this error while using use a streamlit Community Cloud	2	792	April 24, 2023
St.cache_data warning on AWS Deployment cache , pandas	6	442	September 7, 2023

KeyError `warnings`

Related topics