File size: 4,786 Bytes
6773486
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
"""
Pre-compute force-directed layout for Steam co-review network.

Uses networkx spring_layout (Fruchterman-Reingold) with scipy acceleration.
26K nodes, filtered edges. Outputs steam_force_layout.json.

Usage:
    python3 compute_layout.py
"""

import json
import time
import sys
from pathlib import Path

try:
    import networkx as nx
except ImportError:
    print("ERROR: networkx required. pip install networkx")
    sys.exit(1)

# ── Config ──────────────────────────────────────────────────────────
HERE = Path(__file__).parent
NETWORK_JSON = HERE / 'steam_network.json'
OUTPUT = HERE / 'steam_force_layout.json'

# Edge weight threshold — only include edges with weight >= this
# Higher = fewer edges = faster layout. 100 gives ~9K nodes, ~219K edges.
MIN_WEIGHT = 100
# Layout iterations (more = better convergence, slower)
ITERATIONS = 80
# Spring constant (lower = more spread out)
K_SPRING = None  # None = auto (1/sqrt(n))
# Seed for reproducibility
SEED = 42

def main():
    t0 = time.time()
    print("=" * 60)
    print("Steam Network Force Layout Generator")
    print("=" * 60)

    # 1. Load network data
    print("\n[1/4] Loading network data...")
    with open(NETWORK_JSON) as f:
        data = json.load(f)

    nodes = data['nodes']
    links = data['links']
    print(f"  {len(nodes):,} nodes, {len(links):,} links")

    # 2. Build graph with filtered edges
    print(f"\n[2/4] Building graph (min weight={MIN_WEIGHT})...")
    G = nx.Graph()

    # Add all nodes with metadata
    for i, node in enumerate(nodes):
        G.add_node(i, title=node['title'], reviews=node.get('reviews', 0))

    # Add edges above threshold
    added = 0
    skipped = 0
    for link in links:
        if link['weight'] >= MIN_WEIGHT:
            G.add_edge(link['source'], link['target'], weight=link['weight'])
            added += 1
        else:
            skipped += 1

    # Remove isolated nodes (no edges after filtering)
    isolates = list(nx.isolates(G))
    G.remove_nodes_from(isolates)

    print(f"  {G.number_of_nodes():,} nodes, {G.number_of_edges():,} edges")
    print(f"  ({len(isolates):,} isolated nodes removed, {skipped:,} edges below threshold)")

    # 3. Compute layout
    print(f"\n[3/4] Computing spring layout ({ITERATIONS} iterations)...")
    print(f"  This may take a few minutes for {G.number_of_nodes():,} nodes...")

    t1 = time.time()

    # Use weight for layout — higher weight = stronger attraction
    pos = nx.spring_layout(
        G,
        k=K_SPRING,
        iterations=ITERATIONS,
        seed=SEED,
        weight='weight',
    )

    layout_time = time.time() - t1
    print(f"  Layout computed in {layout_time:.1f}s")

    # 4. Build output — map node index back to node ID
    print("\n[4/4] Writing output...")
    output = {
        'positions': {},
        'meta': {
            'node_count': G.number_of_nodes(),
            'edge_count': G.number_of_edges(),
            'min_weight': MIN_WEIGHT,
            'iterations': ITERATIONS,
            'layout_time_seconds': round(layout_time, 1),
        }
    }

    # Normalize positions to [0, 1] using percentile-based scaling.
    # Min/max normalization is ruined by outlier nodes — a few extreme
    # positions compress the main cluster into a tiny area. Percentile
    # scaling maps the 1st-99th percentile range to [0.02, 0.98] and
    # clamps outliers to the edges.
    all_x = sorted(float(xy[0]) for xy in pos.values())
    all_y = sorted(float(xy[1]) for xy in pos.values())
    n = len(all_x)
    p_lo, p_hi = 0.01, 0.99
    x_lo = all_x[int(n * p_lo)]
    x_hi = all_x[int(n * p_hi)]
    y_lo = all_y[int(n * p_lo)]
    y_hi = all_y[int(n * p_hi)]
    x_range = x_hi - x_lo or 1
    y_range = y_hi - y_lo or 1
    margin = 0.02

    for node_idx, (x, y) in pos.items():
        node_id = nodes[node_idx]['id']
        nx = (float(x) - x_lo) / x_range * (1 - 2 * margin) + margin
        ny = (float(y) - y_lo) / y_range * (1 - 2 * margin) + margin
        nx = max(0.0, min(1.0, nx))
        ny = max(0.0, min(1.0, ny))
        output['positions'][node_id] = [round(nx, 6), round(ny, 6)]

    with open(OUTPUT, 'w') as f:
        json.dump(output, f, separators=(',', ':'))

    size_mb = OUTPUT.stat().st_size / (1024 * 1024)
    elapsed = time.time() - t0

    print(f"\n{'=' * 60}")
    print(f"Saved: {OUTPUT}")
    print(f"  {output['meta']['node_count']:,} nodes with positions")
    print(f"  {output['meta']['edge_count']:,} edges used for layout")
    print(f"  Size: {size_mb:.1f} MB")
    print(f"  Total time: {elapsed:.0f}s ({elapsed/60:.1f} min)")
    print("=" * 60)


if __name__ == '__main__':
    main()