Jupyter Notebook displays Series VERY slowly

[JoshuaC3]

Code Sample

import pandas as pd
from IPython.display import display
df = pd.DataFrame(
    pd.np.random.random((10000000, 1)), 
    columns=['a'], 
    index=pd.date_range(start='2001-01-01', freq='min', periods=10000000)
)
display(df.a)

import pandas as pd
from IPython.display import display

# read some data for size 200,000 x 4

display(data.loc[:, ['reading']])  #25ms
display(data.reading.to_frame())  #25ms
display(data.reading) #3.53s
s = data.reading
display(s) #3.32s possibly cached
print(data.reading) #9ms
print(data.loc[:, ['reading']]) # 15ms

Problem description

In the Jupyter notebook, displaying a pd.Series is VERY slow. It is displays quicker when kept as a pd.DataFrame but is of course more verbose.

Expected Output

A pd.Series display/print out at reasonable speeds.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.4.final.0
python-bits: 64
OS: Windows
OS-release: 7
machine: AMD64
processor: Intel64 Family 6 Model 60 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.21.0
pytest: 3.2.1
pip: 9.0.1
setuptools: 36.2.4
Cython: 0.26
numpy: 1.13.3
scipy: 0.19.1
pyarrow: None
xarray: None
IPython: 5.3.0
sphinx: 1.6.3
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2017.3
blosc: None
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.2
feather: None
matplotlib: 2.0.0
openpyxl: 2.5.0a2
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: 0.9.8
lxml: 3.8.0
bs4: 4.6.0
html5lib: 0.9999999
sqlalchemy: 1.1.13
pymysql: 0.7.9.None
psycopg2: None
jinja2: 2.9.6
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: 0.4.0

[TomAugspurger]

Could you make a reproducible by generating the data randomly?

[JoshuaC3]

Apparently not! When I try to replicate this with randomise data, np.random.rand, this prints and displays efficiently. This leads me to believe that the bug is caused by a specific issue with my data. I will investigate further...

[JoshuaC3]

I ran this in the same Jupyter Notebook

import numpy as np
import pandas as pd
from IPython.display import display

data = pd.DataFrame({'reading': np.random.rand(200000), 
                   'strings': ['string']*200000,
                   'ints': range(200000),
                   'dates': pd.date_range(start='2017', periods=200000, freq='MIN')})

display(data.loc[:, ['reading']])  #25ms
display(data.reading.to_frame())  #25ms
display(data.reading) #3.53s
s = data.reading
display(s) #3.32s possibly cached
print(data.reading) #9ms
print(data.loc[:, ['reading']]) # 15ms

All prints/displays performed well.

Are there any attributes/functions/info that I can run on the two DataFrames to compare them and debug?

FYI: I have experienced this issue before in other notebooks, with other, but similar types of data.

[JoshuaC3]

Accidentally closed. My apologies.

[JoshuaC3]

Is there anything else I should do to help debug this?

[TomAugspurger]

Nothing comes to mind.

…

________________________________ From: JoshuaC3 <notifications@github.com> Sent: Wednesday, November 8, 2017 6:19:21 AM To: pandas-dev/pandas Cc: Tom Augspurger; Comment Subject: Re: [pandas-dev/pandas] Jupyter Notebook displays Series VERY slowly (#18137) Is there anything else I should do to help debug this? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#18137 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ABQHIqzu1YnUYFwBM_Z1z_zNEsBjpo0Sks5s0ZxJgaJpZM4QS-S3>.

[jreback]

do a prun on the problematic and ok ones

[JoshuaC3]

A notebook restart has solved this issue. I have had this problem before in other notebooks however. I will run the above when I see this problem arise again. Is it ok to leave this open until then or close and re-open when I re-discover it?

[Strateus]

I have same issue. How to reproduce:

import pandas as pd
from IPython.display import display
df = pd.DataFrame(
    pd.np.random.random((10000000, 1)), 
    columns=['a'], 
    index=pd.date_range(start='2001-01-01', freq='min', periods=10000000)
)
display(df.a)

if you just run display(df) - it works well.

[TomAugspurger]

Thanks for the reproducible example. I've added to the first post.

Anyone have time to look into where things are slow?

[TomAugspurger]

Fixed by #20834 I think. LMK if not.