`fig.update_traces(xaxis=..)` will destroy `hline` or `vline` objects on subplots

Summary

When the figure is created by make_subplots, updating traces' xaxis or yaxis will destroy hline or vlines from user's view.

Details

The following picture represents my figure created with subplots of 3 rows and 1 column. It has horizontal line added by figure.add_hline(y=0.0, ...).

def get_historical_charts(
    tournaments: list[TournamentSummary],
    max_data_points: int = 2000,
    window_sizes: tuple[int, ...] = DEFAULT_WINDOW_SIZES,
):
    # .... some dirty codes..

    figure.add_hline(    # <-------------------------------- ADDED hline
        y=0.0,
        line_color="red",
        line_dash="dash",
        row=1,
        col=1,
        label={
            "text": "Break-even",
            "textposition": "end",
            "font": {"color": "red"},
            "yanchor": "top",
        },
    )
    figure.update_layout(
        title="Historical Performance",
        hovermode="x unified",
        yaxis1={"tickformat": "$"},
        yaxis2={"tickformat": "%"},
        yaxis3={"tickformat": "$"},
    )
    figure.update_yaxes(row=3, col=1, patch={"type": "log"})
    # figure.update_traces(xaxis="x3")  # <------------------------------------- WILL CHANGE HERE
    return figure

However, I wanted to display vertical dash line on all subplots regardless of which graph you are focusing. I googled and found figure.update_traces(xaxis="x3") will make this work, but it destroyed hline. Notice that red dash line is disappeared on the second picture.

def get_historical_charts(
    tournaments: list[TournamentSummary],
    max_data_points: int = 2000,
    window_sizes: tuple[int, ...] = DEFAULT_WINDOW_SIZES,
):
    # .... some dirty codes..

    figure.add_hline(    # <-------------------------------- ADDED hline
        y=0.0,
        line_color="red",
        line_dash="dash",
        row=1,
        col=1,
        label={
            "text": "Break-even",
            "textposition": "end",
            "font": {"color": "red"},
            "yanchor": "top",
        },
    )
    figure.update_layout(
        title="Historical Performance",
        hovermode="x unified",
        yaxis1={"tickformat": "$"},
        yaxis2={"tickformat": "%"},
        yaxis3={"tickformat": "$"},
    )
    figure.update_yaxes(row=3, col=1, patch={"type": "log"})
    figure.update_traces(xaxis="x3")  # <------------- UPDATED XAXIS AND hline IS DESTROYED
    return figure

Note that there is only 1 line of difference from above two codes.

For full codes, you can refer my repository's code.

For bandaid, I can do figure.update_traces(xaxis="x1") instead, but it will destroy custom objects on second/third subplot. So this is not a fundamental resolution.

Personally I think showing horizontal or vertical dash line to indicate which x or y are you hovering is important. However, with current status you can potentially destroy custom plot objects like hline if you do that.

Thanks for reading this.

Start by reproducing the issue with make_subplots, add_hline, and figure.update_traces(xaxis="x3"); the linked visualize.py provides the full example. Trace how update_traces changes axis assignments and how subplot hline objects are represented. Done means updating trace axes preserves every hline and vline across the subplots, with the reported example still displaying them.