import matplotlib.pyplot as plt
import numpy as np

# Data (lang, num_articles, words_per_article, depth)
data = [
    ("English",7085151,709.86,1336),
    ("Cebuano",6115898,216.86,209),
    ("German",3065898,551.76,742),
    ("French",2718357,675.29,1145),
    ("Swedish",2619342,186.88,232),
    ("Dutch",2201228,247.84,281),
    ("Spanish",2071945,634.49,1016),
    ("Russian",2069963,559.71,892),
    ("Italian",1942793,546.34,889),
    ("Polish",1673719,323.17,394),
    ("Ukrainian",1396184,395.25,603),
    ("Vietnamese",1296630,277.79,535),
    ("Portuguese",1159159,538.26,918),
    ("Catalan",783388,520.06,662),
    ("Finnish",606606,297.57,384),
    ("Czech",579646,486.58,660),
    ("Hungarian",562443,432.84,593),
    ("Serbo-Croatian",461167,264.15,503),
    ("Esperanto",377333,245.25,288),
    ("Lithuanian",223935,225.02,285),
    ("Latin",140669,154.71,169),
    ("Ido",59986,131.23,82),
    ("Volapük",45855,71.86,109),
    ("Scots",34282,225.50,359),
    ("Interlingua",30146,110.45,80),
    ("Kotava",29896,91.94,34),
    ("Interlingue",13358,301.30,155),
    ("Sardinian",7728,306.27,362),
    ("Kashubian",5495,80.62,65),
    ("Lingua Franca Nova",4490,357.24,283),
    ("Pennsylvania German",2039,75.91,106),
    ("Novial",1877,82.86,107),
    ("Tetum",1380,195.23,269),
    ("Lojban",1348,344.33,559),
    ("Gothic",976,101.75,162),
    ("Dinka",338,268.36,397),
    ("Cree",14,162.29,341),
]

langs, num_articles, words_per_article, depth = zip(*data)
num_articles = np.array(num_articles)
words_per_article = np.array(words_per_article)
depth = np.array(depth)

# Scale marker sizes (so they're visible but not huge)
area = (words_per_article - words_per_article.min()) / (np.ptp(words_per_article) + 1e-9) * 800 + 50

plt.figure(figsize=(12,8))
scatter = plt.scatter(num_articles, depth, s=area, c=words_per_article, cmap='viridis', alpha=0.85, edgecolor='k', linewidth=0.4)

# Use linear x-axis (remove log scale)
# plt.xscale('log')  # removed
plt.xlabel("Number of articles")
plt.ylabel("Depth*")
plt.title("Languages: Number of articles vs Depth (marker size = words per article, color = words per article)")

# Optionally set x limits for better visualization (uncomment and adjust if desired)
# plt.xlim(0, num_articles.max()*1.05)

cbar = plt.colorbar(scatter)
cbar.set_label("Words per article")

# Annotate a subset to avoid overcrowding: annotate top 25 by number of articles
idxs = np.argsort(-num_articles)[:25]
for i in idxs:
    plt.annotate(langs[i], (num_articles[i], depth[i]),
                 xytext=(5,3), textcoords='offset points', fontsize=9)

plt.grid(axis='y', alpha=0.3)
plt.tight_layout()
plt.show()