TextureSync/server/texture-sync-server/src/search/mod.rs

use crate::model::*;

mod query_filter;
pub use self::query_filter::QueryFilterSyntaxError;
use self::query_filter::*;

pub struct Query {
    filters: Vec<QueryFilter>,
}

impl Query {
    pub fn search(&self, input: &mut Iterator<Item = &Texture>) -> Vec<Texture> {
        let mut results: Vec<(i64, &Texture)> = Vec::new();

        // We use pseudo decimal fixed point numbers here.
        // 1.000_000 = 1_000_000;

        // This is done, since algorithms like quicksort can fail on floats.
        let required_score = self.required_score() * 1_000_000i64;

        'texture_loop: for texture in input {
            let mut score = 0i64;

            for (pos, filter) in self.filters.iter().enumerate() {
                match filter.score(texture) {
                    Score::RequiredMatch(true) => (),
                    Score::RequiredMatch(false) => {
                        // skip this texture
                        continue 'texture_loop;
                    }
                    Score::Match(true) => {
                        score += 1_000_000i64 + ((100_000.0 / f64::sqrt(pos as f64 + 1.0)) as i64);
                    }
                    Score::Match(false) => (),
                }
            }

            if score >= required_score {
                results.push((score, texture))
            }
        }

        results.sort_by_key(|(score, _)| score * -1);

        results
            .iter()
            .map(|(_, texture)| Texture::clone(texture))
            .collect::<Vec<Texture>>()
    }

    fn required_score(&self) -> i64 {
        let non_special = self.filters.iter().filter(|f| !f.is_special()).count() as i64;

        // ceil(non_special / 2)
        (non_special + 1) / 2
    }

    pub fn parse(input: &[String]) -> Result<Query, QueryFilterSyntaxError> {
        let mut result = Query { filters: vec![] };

        for fltr in input {
            let qryfltr = fltr.parse::<QueryFilter>()?;
            result.filters.push(qryfltr);
        }

        Ok(result)
    }
}

#[cfg(test)]
/// These tests check against the example section in the search design document.
mod test {
    use super::*;
    use std::str::FromStr;

    /// just a shorthand
    fn tex(id: i32, name: &str, tags: &str, added_on: &str, resolution: u64) -> Texture {
        Texture {
            id: format!("{}", id), // Id should actaly be a uuid, but for testing this is fine.
            name: name.to_string(),
            tags: tags.split(",").map(|s| s.trim().to_string()).collect(),
            added_on: Date::from_str(added_on).unwrap(),
            resolution: (resolution, resolution),
            format: TextureFormat::JPEG,
            texture_hash: Sha256::from_data(b"Some Hash"),
        }
    }

    // data of example section of search document.
    fn test_data() -> Vec<Texture> {
        vec![
            tex(
                1,
                "wood_185841",
                "Holz, Dunkel, Rot, Edel",
                "2019-05-15",
                4 * 1024,
            ),
            tex(2, "wood_84846", "Holz, Hell", "2019-05-13", 2 * 1024),
            tex(3, "silk_large", "Stoff, Rot, Edel", "2018-01-01", 8 * 1024),
            tex(4, "cotton_xxx", "Stoff, Rot, Rau", "2018-02-01", 2048),
            tex(5, "green_frabric", "Grün, Stoff", "2018-03-01", 1024),
            tex(6, "tin54_45", "Metall, Hell", "2018-03-01", 4 * 1024),
            tex(7, "copper4_1k", "Rot, Metall", "2016-03-01", 1024),
            tex(
                8,
                "rusty_metall",
                "Rot, Metall, Rost, Dunkel",
                "2015-03-01",
                8 * 1024,
            ),
        ]
    }

    fn assert_query(query: &str, expected: Vec<u8>) {
        let q = Query::parse(
            &query
                .split_whitespace()
                .map(|s| s.to_string())
                .collect::<Vec<String>>(),
        )
        .unwrap();

        let data = test_data();

        let result = q.search(&mut data.iter());

        let only_ids = result
            .into_iter()
            .map(|tex| tex.id.parse::<u8>().unwrap())
            .collect::<Vec<u8>>();

        assert_eq!(only_ids, expected)
    }

    #[test]
    fn test_examples_from_doc_1() {
        assert_query("Holz Dunkel", vec![1, 2, 8]);
    }

    #[test]
    fn test_examples_from_doc_2() {
        assert_query("n:wood_", vec![1, 2]);
    }

    #[test]
    fn test_examples_from_doc_3() {
        assert_query("before:2019-05-31 after:2019-04-30", vec![1, 2]);
    }

    #[test]
    fn test_examples_from_doc_4() {
        assert_query("Stoff Rot Edel", vec![3, 4, 1]);
    }

    #[test]
    fn test_examples_from_doc_5() {
        assert_query("Stoff Rot !Edel", vec![4, 3, 5, 7, 8]);
    }

    #[test]
    fn test_examples_from_doc_6() {
        assert_query("Metall Dunkel res:4k", vec![8, 6, 1]);
    }
}