feat(plotnine): implement bar-stacked-percent

plots/bar-stacked-percent/implementations/python/plotnine.py

@@ -1,114 +1,154 @@
 """ anyplot.ai
 bar-stacked-percent: 100% Stacked Bar Chart
-Library: plotnine 0.15.4 | Python 3.13.13
-Quality: 86/100 | Updated: 2026-05-08
+Library: plotnine 0.15.8 | Python 3.13.15
+Quality: 88/100 | Updated: 2026-08-18
 """
 
 import os
 import sys
 
+
 sys.path = [p for p in sys.path if os.path.abspath(p) != os.getcwd()]
 
 import pandas as pd  # noqa: E402
+from mizani.formatters import percent_format  # noqa: E402
 from plotnine import (  # noqa: E402
     aes,
     element_blank,
     element_line,
     element_rect,
     element_text,
     geom_bar,
+    geom_text,
     ggplot,
     labs,
     position_fill,
+    scale_color_identity,
     scale_fill_manual,
+    scale_y_continuous,
     theme,
     theme_minimal,
 )
 
 
-# Theme tokens
+# Theme tokens (see prompts/default-style-guide.md "Background" + "Theme-adaptive Chrome")
 THEME = os.getenv("ANYPLOT_THEME", "light")
 PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
 ELEVATED_BG = "#FFFDF6" if THEME == "light" else "#242420"
 INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
 INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"
+MUTED = "#6B6A63" if THEME == "light" else "#A8A79F"  # Imprint semantic anchor: other/rest
 
-# Okabe-Ito palette
-IMPRINT = ["#009E73", "#C475FD", "#4467A3", "#BD8233"]
+# Imprint palette — named competitors take positions 1-3 in canonical order;
+# "Others" uses the muted semantic anchor since it is literally the aggregate
+# rest-of-market bucket, not a distinct company.
+IMPRINT = ["#009E73", "#C475FD", "#4467A3"]
 
-# Data - Market share by quarter for tech companies
+# Data - smartphone market share by quarter
 quarters = ["Q1 2023", "Q2 2023", "Q3 2023", "Q4 2023", "Q1 2024", "Q2 2024"]
 companies_ordered = ["Others", "Xiaomi", "Samsung", "Apple"]
+apple_share = [23, 21, 20, 22, 21, 20]
+samsung_share = [22, 21, 20, 19, 20, 19]
+xiaomi_share = [12, 13, 14, 14, 15, 16]
+others_share = [43, 45, 46, 45, 44, 45]
 data = {
     "Quarter": quarters * 4,
     "Company": (["Apple"] * 6 + ["Samsung"] * 6 + ["Xiaomi"] * 6 + ["Others"] * 6),
-    "Share": [
-        # Apple
-        23,
-        21,
-        20,
-        22,
-        21,
-        20,
-        # Samsung
-        22,
-        21,
-        20,
-        19,
-        20,
-        19,
-        # Xiaomi
-        12,
-        13,
-        14,
-        14,
-        15,
-        16,
-        # Others
-        43,
-        45,
-        46,
-        45,
-        44,
-        45,
-    ],
+    "Share": apple_share + samsung_share + xiaomi_share + others_share,
 }
 df = pd.DataFrame(data)
 
 # Set categorical ordering for proper display
 df["Quarter"] = pd.Categorical(df["Quarter"], categories=quarters, ordered=True)
 df["Company"] = pd.Categorical(df["Company"], categories=companies_ordered, ordered=True)
 
-# Color mapping with Okabe-Ito palette
-color_map = {"Others": IMPRINT[3], "Xiaomi": IMPRINT[2], "Samsung": IMPRINT[1], "Apple": IMPRINT[0]}
+# Color mapping: Apple/Samsung/Xiaomi in canonical Imprint order, Others muted
+color_map = {"Others": MUTED, "Xiaomi": IMPRINT[2], "Samsung": IMPRINT[1], "Apple": IMPRINT[0]}
+
+# In-segment percentage labels (DE-03): pick whichever ink extreme has higher
+# WCAG contrast against each segment's own fill color, so labels stay legible
+# on both the mid-tone brand hues and the theme-adaptive "Others" gray.
+LIGHT_INK = "#F0EFE8"
+DARK_INK = "#1A1A17"
+
+
+def _relative_luminance(hex_color):
+    r, g, b = (int(hex_color[i : i + 2], 16) / 255 for i in (1, 3, 5))
+
+    def _linearize(c):
+        return c / 12.92 if c <= 0.03928 else ((c + 0.055) / 1.055) ** 2.4
+
+    r, g, b = _linearize(r), _linearize(g), _linearize(b)
+    return 0.2126 * r + 0.7152 * g + 0.0722 * b
+
+
+def _contrast_ratio(l1, l2):
+    lighter, darker = max(l1, l2), min(l1, l2)
+    return (lighter + 0.05) / (darker + 0.05)
+
+
+def _label_color(fill_hex):
+    fill_l = _relative_luminance(fill_hex)
+    light_contrast = _contrast_ratio(fill_l, _relative_luminance(LIGHT_INK))
+    dark_contrast = _contrast_ratio(fill_l, _relative_luminance(DARK_INK))
+    return LIGHT_INK if light_contrast >= dark_contrast else DARK_INK
+
+
+df["Label"] = df["Share"].astype(str) + "%"
+df["LabelColor"] = df["Company"].map(color_map).map(_label_color)
+
+# Precompute label y-positions explicitly (SC-03/VQ-02 fix): geom_bar and
+# geom_text each calling their own independent position_fill() can derive
+# mismatched per-group cumulative offsets when the source rows are grouped
+# by company rather than interleaved per quarter. Instead, compute the exact
+# fill-stack midpoint per (Quarter, Company) ourselves -- stacked bottom-to-top
+# as Apple/Samsung/Xiaomi/Others, i.e. the reverse of the legend/factor order,
+# matching plotnine's default stacking -- and feed it to geom_text via
+# position="identity" so both layers are guaranteed to agree.
+stack_order_bottom_to_top = ["Apple", "Samsung", "Xiaomi", "Others"]
+stack_rank = {company: rank for rank, company in enumerate(stack_order_bottom_to_top)}
+# .map() on a Categorical column returns a Categorical result that inherits
+# the *original* category order, so sorting by it would sort by category
+# position rather than by the mapped rank value -- cast to plain strings
+# first so the mapped ranks are ordinary integers.
+df["StackRank"] = df["Company"].astype(str).map(stack_rank)
+df = df.sort_values(["Quarter", "StackRank"]).reset_index(drop=True)
+df["Fraction"] = df["Share"] / 100
+cum_top = df.groupby("Quarter", observed=True)["Fraction"].cumsum()
+cum_bottom = cum_top - df["Fraction"]
+df["LabelY"] = (cum_top + cum_bottom) / 2
 
-# Theme-adaptive colors for legend and text
+# Theme-adaptive chrome
 anyplot_theme = theme(
     plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
     panel_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
-    panel_grid_major=element_line(color=INK, size=0.3, alpha=0.10),
+    panel_border=element_blank(),
+    panel_grid_major_x=element_blank(),
+    panel_grid_major_y=element_line(color=INK, size=0.3, alpha=0.10),
     panel_grid_minor=element_blank(),
-    panel_border=element_rect(color=INK_SOFT, fill=None, size=0.5),
-    axis_title=element_text(color=INK, size=20),
-    axis_text=element_text(color=INK_SOFT, size=16),
+    axis_title=element_text(color=INK, size=10),
+    axis_text=element_text(color=INK_SOFT, size=8),
     axis_line=element_line(color=INK_SOFT, size=0.5),
-    plot_title=element_text(color=INK, size=24, weight="bold"),
+    plot_title=element_text(color=INK, size=12),
     legend_background=element_rect(fill=ELEVATED_BG, color=INK_SOFT),
-    legend_title=element_text(color=INK, size=16),
-    legend_text=element_text(color=INK_SOFT, size=16),
+    legend_title=element_text(color=INK, size=8),
+    legend_text=element_text(color=INK_SOFT, size=8),
     legend_position="right",
-    figure_size=(16, 9),
+    figure_size=(8, 4.5),
 )
 
-# Create 100% stacked bar chart
+# Create 100% stacked bar chart with in-segment percentage labels
 plot = (
     ggplot(df, aes(x="Quarter", y="Share", fill="Company"))
-    + geom_bar(stat="identity", position=position_fill(), width=0.65)
+    + geom_bar(stat="identity", position=position_fill(), width=0.7)
+    + geom_text(aes(y="LabelY", label="Label", color="LabelColor"), position="identity", size=2.8, show_legend=False)
     + scale_fill_manual(values=color_map)
-    + labs(title="bar-stacked-percent · plotnine · anyplot.ai", x="Quarter", y="Market Share (%)", fill="Company")
+    + scale_color_identity()
+    + scale_y_continuous(labels=percent_format())
+    + labs(title="bar-stacked-percent · python · plotnine · anyplot.ai", x="Quarter", y="Market Share", fill="Company")
     + theme_minimal()
     + anyplot_theme
 )
 
 # Save
-plot.save(f"plot-{THEME}.png", dpi=300, verbose=False)
+plot.save(f"plot-{THEME}.png", dpi=400, width=8, height=4.5, units="in", verbose=False)