harmorp 0.1.2

Enhanced Nazief-Adriani Indonesian stemmer: iterative ECS, nasal-assimilation restoration, phonotactic guards, FST dictionary, zero-alloc hot path
Documentation 
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

use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
use harmorp::IndonesianStemmer;
use sastrawi::{Dictionary, Stemmer as SastrawiStemmer};

// ── Representative word sets ──────────────────────────────────────────────────

const ROOTS: &[&str] = &[
    "buku", "makan", "rumah", "anak", "kerja", "besar", "kecil", "tulis", "baca", "jual", "beli",
    "ajar", "pikir", "dapat",
];

const SINGLE_PREFIX: &[&str] = &[
    "membaca", "menulis", "berjalan", "terbang", "sekolah", "bermain", "bekerja", "pelajar",
    "pekerja", "pembaca",
];

const MULTI_AFFIX: &[&str] = &[
    "pembelajaran",
    "membukakan",
    "pengembangan",
    "mempertimbangkan",
    "penyelenggaraan",
    "mengepulangkan",
    "pembangunan",
    "pemberitahuan",
    "menyampaikan",
    "mengembangkan",
    "menyelenggarakan",
    "mempertahankan",
];

fn corpus() -> Vec<String> {
    let mut v = Vec::with_capacity(ROOTS.len() + SINGLE_PREFIX.len() + MULTI_AFFIX.len());
    v.extend(ROOTS.iter().map(|s| s.to_string()));
    v.extend(SINGLE_PREFIX.iter().map(|s| s.to_string()));
    v.extend(MULTI_AFFIX.iter().map(|s| s.to_string()));
    v
}

fn create_sastrawi_dict() -> Dictionary {
    // Create a basic dictionary with common root words
    Dictionary::custom(ROOTS)
}

// ── Comparison Benchmarks ───────────────────────────────────────────────────────

fn compare_single_words(c: &mut Criterion) {
    let harmorp = IndonesianStemmer::new();
    let dict = create_sastrawi_dict();
    let sastrawi = SastrawiStemmer::new(&dict);

    let mut group = c.benchmark_group("compare_single_cold");

    for word in SINGLE_PREFIX.iter().chain(MULTI_AFFIX) {
        group.throughput(Throughput::Elements(1));

        group.bench_with_input(BenchmarkId::new("harmorp", word), word, |b, w| {
            b.iter(|| {
                harmorp.clear_cache();
                harmorp.stem(black_box(w))
            });
        });

        group.bench_with_input(BenchmarkId::new("sastrawi", word), word, |b, w| {
            b.iter(|| {
                let mut word_mut = w.to_string();
                sastrawi.stem_word(&mut word_mut);
                word_mut
            });
        });
    }
    group.finish();
}

fn compare_batch(c: &mut Criterion) {
    let harmorp = IndonesianStemmer::new();
    let dict = create_sastrawi_dict();
    let sastrawi = SastrawiStemmer::new(&dict);
    let corp = corpus();

    let mut group = c.benchmark_group("compare_batch");
    group.throughput(Throughput::Elements(corp.len() as u64));

    group.bench_function("harmorp_cold", |b| {
        b.iter(|| {
            harmorp.clear_cache();
            harmorp.stem_batch(black_box(&corp))
        });
    });

    group.bench_function("harmorp_hot", |b| {
        // Warm cache
        harmorp.stem_batch(&corp);
        b.iter(|| harmorp.stem_batch(black_box(&corp)));
    });

    group.bench_function("sastrawi", |b| {
        b.iter(|| {
            corp.iter()
                .map(|w| {
                    let mut w_mut = w.clone();
                    sastrawi.stem_word(&mut w_mut);
                    w_mut
                })
                .collect::<Vec<_>>()
        });
    });

    group.finish();
}

fn compare_throughput(c: &mut Criterion) {
    let harmorp = IndonesianStemmer::new();
    let dict = create_sastrawi_dict();
    let sastrawi = SastrawiStemmer::new(&dict);
    let large: Vec<String> = corpus().into_iter().cycle().take(10_000).collect();

    let mut group = c.benchmark_group("compare_throughput");
    group.throughput(Throughput::Elements(large.len() as u64));

    group.bench_function("harmorp_10k", |b| {
        harmorp.stem_batch(&large);
        b.iter(|| harmorp.stem_batch(black_box(&large)));
    });

    group.bench_function("sastrawi_10k", |b| {
        b.iter(|| {
            large
                .iter()
                .map(|w| {
                    let mut w_mut = w.clone();
                    sastrawi.stem_word(&mut w_mut);
                    w_mut
                })
                .collect::<Vec<_>>()
        });
    });

    group.finish();
}

criterion_group!(
    benches,
    compare_single_words,
    compare_batch,
    compare_throughput
);
criterion_main!(benches);