Skip to content

Refactor hash calculation and token emission logic for improved clarity and performance #20

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Refactor hash calculation and token emission logic for improved clarity and performance #20

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
123 changes: 55 additions & 68 deletions src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,90 +35,82 @@ fn murmurhash2(mut k: u32, seed: u32) -> u32 {
}

impl Feature {
#[inline(always)]
pub fn to_hash(&self) -> u32 {
match self {
Feature::AsciiNGram(ngram) => murmurhash2(*ngram, SEED),
Feature::Unicode(chr) => murmurhash2(*chr as u32 / 128, SEED ^ 2),
Feature::UnicodeClass(chr) => murmurhash2(classify_codepoint(*chr), SEED ^ 4),
match *self {
Feature::AsciiNGram(ngram) => murmurhash2(ngram, SEED),
Feature::Unicode(chr) => murmurhash2((chr as u32) >> 7, SEED ^ 2),
Feature::UnicodeClass(chr) => murmurhash2(classify_codepoint(chr), SEED ^ 4),
}
}
}

pub fn detect_language(text: &str) -> Lang {
let mut scores: [f32; NUM_LANGUAGES] = Default::default();
let mut num_features: u32 = 0;
emit_tokens(
text,
#[inline(always)]
|token| {
num_features += 1u32;
let bucket = token.to_hash() % DIMENSION as u32;
let idx = bucket as usize * NUM_LANGUAGES;
let per_language_scores = &weights::WEIGHTS[idx..idx + NUM_LANGUAGES];
for i in 0..NUM_LANGUAGES {
scores[i] += per_language_scores[i];
}
},
);
emit_tokens(text, |token| {
num_features += 1;
let bucket = token.to_hash() % DIMENSION as u32;
let idx = bucket as usize * NUM_LANGUAGES;
scores
.iter_mut()
.zip(&weights::WEIGHTS[idx..idx + NUM_LANGUAGES])
.for_each(|(score, weight)| *score += weight);
});
if num_features == 0 {
// By default, we return English
return Lang::Eng;
}

let sqrt_inv_num_features = 1.0f32 / (num_features as f32).sqrt();
#[allow(clippy::needless_range_loop)]
for i in 0..NUM_LANGUAGES {
// Ok so the sqrt(num_features) is not really the norm, but whatever.
scores[i] = scores[i] * sqrt_inv_num_features + weights::INTERCEPTS[i];
let sqrt_inv = 1.0 / (num_features as f32).sqrt();
scores
.iter_mut()
.zip(weights::INTERCEPTS.iter())
.for_each(|(score, intercept)| *score = *score * sqrt_inv + intercept);
let mut lang_id = 0;
let mut max_score = scores[0];
for (i, &score) in scores.iter().enumerate().skip(1) {
if score > max_score {
max_score = score;
lang_id = i;
}
}

let lang_id = scores
.iter()
.enumerate()
.max_by(|(_, &score_left), (_, &score_right)| score_left.partial_cmp(&score_right).unwrap())
.map(|(pos, _val)| pos)
.unwrap();
weights::LANGUAGES[lang_id]
}

#[doc(hidden)]
pub fn emit_tokens(text: &str, mut listener: impl FnMut(Feature)) {
let mut prev = ' ' as u32;
let mut num_previous_ascii_chr = 1;
let mut num_prev_ascii = 1;
for chr in text.chars() {
let code = chr.to_ascii_lowercase() as u32;
if !chr.is_ascii() {
listener(Feature::Unicode(chr));
listener(Feature::UnicodeClass(chr));
num_previous_ascii_chr = 0;
continue;
}
prev = prev << 8 | code;
match num_previous_ascii_chr {
0 => {
num_previous_ascii_chr = 1;
num_prev_ascii = 0;
} else {
let code = chr.to_ascii_lowercase() as u32;
prev = prev << 8 | code;
match num_prev_ascii {
0 => {
num_prev_ascii = 1;
}
1 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
num_prev_ascii = 2;
}
2 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
listener(Feature::AsciiNGram(prev & TRIGRAM_MASK));
num_prev_ascii = 3;
}
3 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
listener(Feature::AsciiNGram(prev & TRIGRAM_MASK));
listener(Feature::AsciiNGram(prev));
}
_ => unreachable!(),
}
1 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
num_previous_ascii_chr = 2;
if !chr.is_alphanumeric() {
prev = ' ' as u32;
}
2 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
listener(Feature::AsciiNGram(prev & TRIGRAM_MASK));
num_previous_ascii_chr = 3;
}
3 => {
listener(Feature::AsciiNGram(prev & BIGRAM_MASK));
listener(Feature::AsciiNGram(prev & TRIGRAM_MASK));
listener(Feature::AsciiNGram(prev));
}
_ => {
unreachable!();
}
}
if !chr.is_alphanumeric() {
prev = ' ' as u32;
}
}
}
Expand Down Expand Up @@ -195,15 +187,10 @@ fn classify_codepoint(chr: char) -> u32 {

#[cfg(test)]
mod tests {
use crate::detect_language;
use crate::emit_tokens;
use crate::Feature;
use crate::Lang;
use crate::{detect_language, emit_tokens, Feature, Lang};

fn ascii_ngram_feature(text: &str) -> Feature {
assert!(text.is_ascii());
let mut bytes: [u8; 4] = [0u8; 4];
assert!(text.len() <= 4);
let mut bytes = [0; 4];
bytes[4 - text.len()..].copy_from_slice(text.as_bytes());
Feature::AsciiNGram(u32::from_be_bytes(bytes))
}
Expand All @@ -213,8 +200,8 @@ mod tests {
let mut tokens = Vec::new();
emit_tokens("hello　こん！", |token| tokens.push(token));
assert_eq!(
&tokens,
&[
tokens,
vec![
ascii_ngram_feature(" h"),
ascii_ngram_feature("he"),
ascii_ngram_feature(" he"),
Expand Down