MeiliSearch/src/bin/raptor-cli.rs
2018-08-25 17:00:05 +02:00

70 lines
2.0 KiB
Rust

// TODO make the raptor binary expose multiple subcommand
// make only one binary
extern crate fst;
extern crate raptor;
extern crate serde_json;
#[macro_use] extern crate serde_derive;
use std::collections::HashSet;
use std::fs::File;
use std::io::{BufReader, BufRead};
use serde_json::from_str;
use raptor::{MapBuilder, Map};
#[derive(Debug, Deserialize)]
struct Product {
product_id: u64,
title: String,
ft: String,
}
fn main() {
let data = File::open("products.json_lines").unwrap();
let data = BufReader::new(data);
let common_words = {
// TODO don't break if doesn't exist
let file = File::open("fr.stopwords.txt").unwrap();
let file = BufReader::new(file);
let mut set = HashSet::new();
for line in file.lines() {
let words = line.unwrap();
for word in words.split_whitespace() {
set.insert(word.to_owned());
}
}
set
};
let mut builder = MapBuilder::new();
for line in data.lines() {
let line = line.unwrap();
// TODO if possible remove String allocation of Product here...
let product: Product = from_str(&line).unwrap();
let title = product.title.split_whitespace();
let description = product.ft.split_whitespace().filter(|&s| s != "Description");
let words = title.chain(description)
.filter(|&s| s.chars().any(|c| c.is_alphabetic())) // remove that ?
.map(|s| s.trim_matches(|c: char| !c.is_alphabetic()).to_lowercase())
.filter(|s| !common_words.contains(s));
for word in words {
builder.insert(word, product.product_id);
}
}
let map = File::create("map.fst").unwrap();
let values = File::create("values.vecs").unwrap();
let (map, values) = builder.build(map, values).unwrap();
eprintln!("Checking the dump consistency...");
unsafe { Map::<u64>::from_paths("map.fst", "values.vecs").unwrap() };
}