Add static_columns param for interactive dataframes

[hannahblair]

Description

Adds a static_columns param, which takes a list of ints or strs to reference the columns. The use of static rows does feel potentially quite niche so we'll see if there are any requests for that.

Closes: #2020

🎯 PRs Should Target Issues

Before your create a PR, please check to see if there is an existing issue for this change. If not, please create an issue before you create this PR, unless the fix is very small.

Not adhering to this guideline will result in the PR being closed.

Testing and Formatting Your Code

1. PRs will only be merged if tests pass on CI. We recommend at least running the backend tests locally, please set up your Gradio environment locally and run the backed tests: bash scripts/run_backend_tests.sh

2. Please run these bash scripts to automatically format your code: bash scripts/format_backend.sh, and (if you made any changes to non-Python files) bash scripts/format_frontend.sh

[hannahblair]

I am interested to know whether we need a little UI/UX touch to indicate the column isn't editable. I tried a not-allowed cursor but it's a bit too aggressive.

[abidlabs]

Thanks @hannahblair! LGTM with two minor suggestions:

• I do think some kind of visual feedback would be helpful. Perhaps we can apply a slight background shade or opacity to cells that are static?

Adds a static_columns param, which takes a list of ints or strs to reference the columns

• I see the ergonomic benefit of being able to refer to columns by either the ints or strs, but I think it may lead to some ambiguity since we create 1-indexed headers by default. For example, if I adapt the matrix_transpose demo to add a static column:

import numpy as np

import gradio as gr

def transpose(matrix):
    return matrix.T

demo = gr.Interface(
    transpose,
    gr.Dataframe(type="numpy", datatype="number", row_count=5, col_count=3, show_fullscreen_button=True, static_columns=[1]),
    "numpy",
    examples=[
        [np.zeros((3, 3)).tolist()],
        [np.ones((2, 2)).tolist()],
        [np.random.randint(0, 10, (3, 10)).tolist()],
        [np.random.randint(0, 10, (10, 3)).tolist()],
        [np.random.randint(0, 10, (10, 10)).tolist()],
    ],
    cache_examples=False
)

if __name__ == "__main__":
    demo.launch()

the developer may be referring to either the column whose header is 1 or the second column in the dataframe. To avoid ambiguity, perhaps we should only refer to columns by numerical indices.

[abidlabs]

While I was testing this, I noticed an strange but unrelated issue which is that when we create column headers automatically, they can be either 0-indexed or 1-indexed (I think depending on whether the dataframe initially has values):

import gradio as gr
import pandas as pd


data = [[25, 45, 35],
[88, 92, 95]]

df = pd.DataFrame(data)


with gr.Blocks() as blocks:
    m = gr.Dataframe()
    m = gr.Dataframe(df)


blocks.launch()

[hannahblair]

@abidlabs update the datatype!

i also think a fairly straightforward disabled indicator is actually the cursor not-allowed on the actual input element. so right now clicking on the cell will show the input, but with this cursor change the input is disabled and the cursor is not-allowed. here's a visual:

[hannahblair]

I was thinking a padlock somewhere in the cell but it just looks too crowded.

[abidlabs]

sgtm

[hysts]

The current behavior doesn't seem to work well with adding/deleting columns.

It depends on the use case, but for example, even if there is a column that the app developer wants to prevent users from changing, users can make it editable by adding a new column to its left.

import gradio as gr
import numpy as np

with gr.Blocks() as demo:
    gr.Dataframe(value=np.arange(15).reshape(3, 5), static_columns=[1], interactive=True)
demo.launch()

0.mp4

[abidlabs]

great catch @hysts. Perhaps we'll need to add some logic that if static_columns is specified, then col_count is fixed and columns cannot be inserted/deleted at all

[hannahblair]

hmm I see. alternatively, we could remove the specified static column from static_columns if it's removed. but adding columns will also cause some confusing behaviour. i've updated the PR to set it to "fixed" if static columns are specified. thanks for reviewing @hysts!

[abidlabs]

I reran @hysts's repro and I'm seeing a different issue with cell / header alignment being off:

import gradio as gr
import numpy as np

with gr.Blocks() as demo:
    gr.Dataframe(value=np.arange(15).reshape(3, 5), static_columns=[1], interactive=True)
demo.launch()

[hannahblair]

@abidlabs fixed! my bad python code was the culprit

[abidlabs]

Nice @hannahblair lgtm! We could consider adding a padlock symbol just in the header cell to designate a static column, but we don't have to do it now (or ever if it's a bad ui)