Pandas import error #2433
Comments
|
I can make the error go away by adding |
|
@alexcjohnson when did you get the error, when running the script or when triggering a specific callback from the UI? I cannot reproduce (I tried changing the value of the |
|
The error appears for me when first loading the app in the browser: If I reload the page everything is fine thereafter - so it's just the first time these callbacks are called. after starting the app Maybe the two callbacks are being called in parallel threads and this is a race condition? One callback starts importing |
|
Interesting. Hum maybe this is indeed what happens. Probably if you initialize the |
|
I have the same error when I try to update a graph using callback |
|
Not sure if this is 100% related, but I put up a PR recently for a similar issue (#2391) involving ValueErrors thrown with numpy imports. Not a dash rendering issue, but I wonder if there's a more more efficient way we can do type checks involving different imports with libraries like numpy/pandas? Interesting that OP has a "partially initialized module error". That seems to support the race condition theory and why it might only fail on initialization. |
|
I'm not able to reproduce locally either, and wonder if it's dependent on the machine itself (perhaps low memory/cpu when running the app? so loading pandas is taking longer than it takes to make that initial callback) |
I was able to get the error with just this block of simple code. I tried fixing multiple python installations as well. Creating a virtual environment also doesn't help. Will be really grateful if you could fix this! Here is the error log |
|
Still haven't been able to reproduce locally with my machine, with any of these provided examples... Makes me think this could potentially be hardware/machine specific and some timing issues loading the library. Anyone have any other thoughts on how we might be able to reproduce better or locate a good place to start making some improvements? |
|
I have a scenario that seems to reproduce this type of error fairly consistently: from uuid import uuid4
import flask
import dash
from dash import html, dcc
from dash.dependencies import Output, Input
import plotly.graph_objects as go
server = flask.Flask('app')
app = dash.Dash('app', server=server)
n_comps = 10
buttton_id = "click_me"
n_clicks_counter_id = str(uuid4())
figids = [str(uuid4()) for _ in range(n_comps)]
@app.callback(
Output(n_clicks_counter_id, "children"),
Input(buttton_id, "n_clicks")
)
def update_count(n_clicks):
return str(n_clicks)
for figid in figids:
@app.callback(
Output(figid, "figure"),
Input(buttton_id, "n_clicks")
)
def plot(val):
if val is None:
val = 0
figure = go.Figure(
data=[go.Bar(x=[1+val, 2, 3], y=[1, 3, 2])],
layout=go.Layout(
title=go.layout.Title(text="A Figure Specified By A Graph Object")
)
)
return figure
app.layout = html.Div(
id="main-title",
children=[html.Button(id=buttton_id, children="click me"), html.Div(children="0", id=n_clicks_counter_id)] +
[
dcc.Graph(id=figid) for figid in figids
]
)
if __name__ == '__main__':
app.run_server()Instead of running the script directly, start the server with gunicorn set to do semi-frequent worker reboots; e.g. Then, click the button repeatedly; it should trigger an internal error fairly frequently. With dash 2.0 + orjson installed, it seems to first trigger a bad import of orjson: if orjson is pre-emptively imported, it will instead break on a numpy import: I tried this on a couple of machines and it seems to reproduce consistently. |
A Dash user is seeing an error from
pandaswithout ever directly importingpandas, justdash,plotly, andnumpy. https://community.plotly.com/t/callback-error-when-plotting-multiple-graph-objects/38756The error occurs inside
plotlywhen the Dash app tries to render one of the plots in a callback:That's the error I see - the OP's error message is a little more extensive, which is a little funny because we both report pandas v1.0.3
I can reproduce locally with this app:
The text was updated successfully, but these errors were encountered: