Fixed force_terminal on Jupyter

[verdimrc]

Type of changes

• Bug fix
• New feature
• Documentation / docstrings
• Tests
• Other

Checklist

• I've run the latest black with default args on new code.
• I've updated CHANGELOG.md and CONTRIBUTORS.md where appropriate.
• I've added tests for new code.
• I accept that @willmcgugan may be pedantic in the code review.

Description

Please describe your changes here. If this fixes a bug, please link to the issue, if possible.

Fixes behavior after running rich.pretty.install() Jupyter Lab:

• Do not show <Figure size 432x288 with 1 Axes> on Matplotlib figures
• With force_terminal=True. and force_jupyter=None (or True), pandas dataframe must still be rendered once (i.e., the html repr).

Screenshot below (from this notebook) compares the four behaviors: (1) vanilla Python, (2 and 3) rich, and (4) this PR.

[willmcgugan]

Why would you force_terminal in Jupyter lab?

Please don't ask me to run something in Jupyter when you can paste a screenshot.

[verdimrc]

Thank you for taking a look. Added a screenshot to the PR description, with markers on the behaviors that this PR tries to fix.

[willmcgugan]

Why would you force_terminal in Jupyter lab?

[codecov-commenter]

Codecov Report

Merging #1918 (9cf8499) into master (e839bfb) will decrease coverage by 0.02%.
The diff coverage is 98.49%.

@@            Coverage Diff             @@
##           master    #1918      +/-   ##
==========================================
- Coverage   99.82%   99.80%   -0.03%     
==========================================
  Files          71       71              
  Lines        6943     7057     +114     
==========================================
+ Hits         6931     7043     +112     
- Misses         12       14       +2     

Flag	Coverage Δ
unittests	99.80% <98.49%> (-0.03%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
rich/default_styles.py	100.00% <ø> (ø)
rich/markdown.py	100.00% <ø> (ø)
rich/pretty.py	99.43% <97.61%> (-0.57%)	⬇️
rich/syntax.py	99.27% <97.72%> (-0.34%)	⬇️
rich/__main__.py	100.00% <100.00%> (ø)
rich/_inspect.py	100.00% <100.00%> (ø)
rich/color.py	100.00% <100.00%> (ø)
rich/console.py	100.00% <100.00%> (ø)
rich/progress.py	98.43% <100.00%> (+0.31%)	⬆️
rich/segment.py	99.34% <100.00%> (+0.05%)	⬆️
... and 9 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 880693d...9cf8499. Read the comment docs.

[willmcgugan]

I'm going to need an answer to my question, and some code I can cut and paste (or a notebook).

[verdimrc]

Hi, I thought the quoted question was posted accidentally, my apology.

The motivation to let Jupyter display cell outputs as colored ASCII rather than HTML are:

1. to use the same font as the code cells. Otherwise, HTML output will be displayed with a noticeably different font.
2. to automatically wrap long ASCII texts (which is how Jupyter behaves without rich). Otherwise, long HTML output will need horizontal scrolling, which makes rich notebook to behave differently than stock Jupyter.

Below are the Python fragments to demonstrate the effect of this PR. You'll need to pip install pandas to run this example in Jupyter.

This fragment shows the effect on Jupyter "display_data".

import pandas as pd
import rich
import rich.pretty

# Stuffs to display
d = {"A" * 200, "B" * 200}
df = pd.DataFrame({"a": range(3), "b": range(3)})

# Render "display_data" as colored text.
rich.reconfigure(force_terminal=True)
rich.pretty.install()
display(d, df, df["a"])

# Jupyter may surpress certain "display_data". For example, the semicolon makes Jupyter hide the
# "<AxesSubplot:...>" message.
df.plot();

Here's the side-by-side comparison:

One thing to also note is this PR changes only the ipython display hook, and makes no attempt to change the Console behavior. In the next example, the Jupyter "stream_output" still defaults to HTML, and will have to be specifically instructed for colored ASCII.

# By default, render "stream_output" (e.g., stdout) as HTML.
rich.get_console().out("a", d)

# However, "stream_output" can be force rendered as colored text.
rich.console.Console(force_terminal=True, force_jupyter=False).out("a", d)

[olivierphi]

@verdimrc Note that in order to automatically wrap long ASCII texts it seems that we just have to specify a fold "overflow" to Rich, without having to use the force_* hammers: i.e. with rich.pretty.install(overflow="fold") the text is wrapped when the Jupyter notebook displays it 🙂