Autocompletions not working with fish shell after #877 was merged

[goraje]

First Check

• I added a very descriptive title here.
• I used the GitHub search to find a similar question and didn't find it.
• I searched the Typer documentation, with the integrated search.
• I already searched in Google "How to X in Typer" and didn't find any information.
• I already read and followed all the tutorials in the docs and didn't find an answer.
• I already checked if it is not related to Typer but to Click.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from typing import Annotated

import typer
from rich.markup import escape

app = typer.Typer(no_args_is_help=True, rich_markup_mode="rich")

@app.command(name="example1", help="Just [bold]some[/] help for example1")
def example1(name: Annotated[str, typer.Option(help="test1  name")]) -> None:
    pass

@app.command(name="example2", help=f"Just some help for example2 {escape('[I'm just text in brackets]')}")
def example2() -> None:
    pass

def main() -> None:
    app()

Description

• open a terminal using fish shell
• hit tab completion
• observe the following output instead of a valid tab completion

testapp example1\ \ \ Just\ some\ help\ for\ example1

What I think happens here is that changes introduced in #877 which make use of typer.rich_utils.print_with_rich to print out the autocompletion yield malformed completions.

To be more precise rich.Console.print does no preserve "\t" which acts as the separator between the item and item help in the completion, so invoking console.print(text) in the aforementioned method inadvertently replaces the "\t". Rich has made it pretty clear that they do not intend to implement behaviour that preserves tabs here. This means that the current code completions are properly generated only for some shells. This could be circumvented eg. by implementing something to the likes of

def _render_remove_markup(text: str) -> str:
    try:
        import rich

    except ImportError:  # pragma: no cover
        rich = None  # type: ignore

    if rich:
        from rich.console import Console

        console = Console()
        rendered_text = "".join(
            segment.text for segment in console.render(text)
        ).rstrip("\n")
        return rendered_text

    return text

and then using it directly to format the help text for the completion in the designated format_completion method of a given shell completion class eg. for fish shell

def format_completion(self, item: click.shell_completion.CompletionItem) -> str:
        # TODO: Explore replicating the new behavior from Click, pay attention to
        # the difference with and without formatted help
        # if item.help:
        #     return f"{item.type},{item.value}\t{item.help}"

        # return f"{item.type},{item.value}
        if item.help:
            formatted_help = re.sub(r"\s", " ", item.help)
            formatted_help = _render_remove_markup(formatted_help)
            return f"{item.value}\t{formatted_help}"
        else:
            return f"{item.value}"

This approach:

• still removes the markup tags without touching the final output format
• goes around the problems with console.print so changes from ✨ Remove Rich tags when showing completion text #877 in typer.completions.py could be reverted to:

 # Typer override to print the completion help msg with Rich
    if instruction == "complete":
        click.echo(comp.complete())
        return 0

which preserves the proper output formatting.

Final thoughts:

• removing the inline markup accepted by rich should be done before final completion formatting
• the current solution should be fixed for fish shell

Operating System

macOS

Operating System Details

No response

Typer Version

0.14.0

Python Version

Python 3.12.7

Additional Context

No response

[bckohan]

I confirm that I see this behavior >=0.13.1. I also started seeing ansi color control sequences come through on some systems.

[svlandeg]

Thanks for the report @goraje and @bckohan for the confirmation! I'm looking into this issue now.

I also started seeing ansi color control sequences come through on some systems.

Could you open a separate discussion about this, with some more information?

[svlandeg]

The original issue as mentioned by @goraje here, should be fixed by PR #1069.

[YuriiMotov]

Since #1069 has been merged, I think this can be closed, right?