save: investigate use of base64 encoded urls in headless browser

[machow]

In PR #499 we replaced saving to a temporary file with inputting base64 encoded rendered html into a headless browser.

Browser url limits are a bit tricky, but they do exist. Let's investigate potential issues we might hit by passing long urls to the headless browser.

Alternatively, it might save us time to revert back to saving a temporary file, since the gt R package has used that approach for ~6 years with little issue.

I would vote for reverting, just to align with gt R. It's likely modern browsers have large url limits, but it's tricky to verify.

[jrycw]

@machow , I wasn’t aware of this issue. Please feel free to revert to the temporary file approach if needed.

[phimad]

Related to this, I am having problems with special character encodings which I believe is due to the base64 encoding.

Below I provide a MRE. I have a table with special Danish characters such as Ø, Æ, Å. It shows up correctly with GT.show method, but special characters are lost when using save()-method.

import pandas as pd
from great_tables import GT

df = pd.DataFrame(
{
"Stubhead with Å": [
"Rowlabel with Ø",
"Rowlabel with Æ",
],
"2023": [
3.2,
4.3,
],
"2024": [
5.7,
7.8,
],
"2025": [
8.9,
5.4,
],
}
)

gt_tbl = GT(df)

gt_tbl.save("test.png")

This outputs the attached image:

which does not show the special characters correctly. Seems it would be prudent to use the same approach for the show()-method as for the save()-method, but I am not an expert on this.

Am hoping this will be fixed soon as I really like this library :)

[jrycw]

@phimad, thanks so much for reporting this issue. I’ve reformatted your code for better readability:

import pandas as pd
from great_tables import GT

df = pd.DataFrame(
    {
        "Stubhead with Å": [
            "Rowlabel with Ø",
            "Rowlabel with Æ",
        ],
        "2023": [
            3.2,
            4.3,
        ],
        "2024": [
            5.7,
            7.8,
        ],
        "2025": [
            8.9,
            5.4,
        ],
    }
)

gt_tbl = GT(df)

gt_tbl.save("test.png")

@machow and @rich-iannone, it looks like our old friend is making a return. Interestingly, the code runs fine on my Windows 11 machine.

[phimad]

To be clear, the output in the jupyter notebook is correct, but the saved image file is incorrect.

I am on Windows 10. Is there any sort of test I can run on my machine to help you?

EDIT:
It works when I revert to saving a temporary file to feed into the browser.

[jrycw]

Woah! Thanks so much for the clarification.

I can confirm that the issue also occurs on my machine: