Thank you!

[floringogianu]

I keep postponing this message for over a year now. Just wanted to say thank you to Hassan an the rest of the contributors for how amazing plotnine feels in day to day use. I've been wrangling plots in matplotlib, seaborn and altair for a while now but never reached this level of enjoyment.

Not my best plot, just the one I'm currently working on.

_agg = iqm

(
    p9.ggplot(aql_topK)
    + p9.geom_col(p9.aes(x="eps", y="Gt_nrm", fill="factor(lr)"), position="dodge", stat="summary", fun_y=_agg)
    + p9.geom_point(p9.aes(x="eps", y="Gt_nrm"), stat="summary", fun_y=_agg, size=2)
    + p9.geom_line(p9.aes(x="eps", y="Gt_nrm"), stat="summary", fun_y=_agg, size=1)
    + p9.scale_x_continuous(breaks=aql_topK["eps"].unique())
    + p9.facet_grid("~ resetable", scales="free_x", labeller="label_both")
    + p9.scale_fill_manual(AQL_PAL)
    + p9.scale_color_manual(AQL_PAL)
    # + p9.scale_x_log10(breaks=aql_topK["eps"].unique())
    # + p9.geom_hline(p9.aes(yintercept="best_score"), data=_best_score, linetype="dotted", alpha=0.5)
    + p9.theme(
        figure_size=(7, 3),
        strip_background=p9.themes.elements.element_blank(),
        panel_grid_major_y = p9.element_line(alpha=0.33, size=0.33)
    )
    + p9.labs(
        x=r"$\varepsilon$ value", y="IQM score",
        color=r"$\eta$ value:", fill=r"$\eta$ value:",
        subtitle=r"AQ($\lambda$) aggregate performance",
        caption=(
            r"Grid search over $\eta, \varepsilon$ and $\beta_0$ values. Lines are aggregates over $\varepsilon$."
            # + "\n" + r"Dotted line is the maximum of the averages of all runs in an $(\eta,\varepsilon$) group."
        ),
        # tag="A",
    )
).draw()