brittling/src/literature.rs

use ammonia::Builder;
use html_escape::decode_html_entities;
use reqwest::Error;
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use unicode_general_category::{GeneralCategory, get_general_category};

#[derive(Serialize, Deserialize, Debug, Clone)]
pub struct Authorship {
    pub author_position: String,
    pub raw_author_name: String,
}

// TODO: Handle duplicates by having vectors of ids
#[derive(Serialize, Deserialize, Debug, Clone)]
pub struct Publication {
    pub id: String,
    pub display_name: Option<String>,
    pub authorships: Vec<Authorship>,
    pub publication_year: Option<u32>,
    pub abstract_inverted_index: Option<HashMap<String, Vec<u32>>>,
    pub referenced_works: Vec<String>,
}

#[derive(Serialize, Deserialize, Default)]
pub enum SnowballingStep {
    #[default]
    Backward,
    Forward,
}

impl ToString for SnowballingStep {
    fn to_string(&self) -> String {
        match self {
            SnowballingStep::Forward => String::from("forward"),
            SnowballingStep::Backward => String::from("backward"),
        }
    }
}

// TODO: Only store IDs of excluded publications?
#[derive(Serialize, Deserialize, Default)]
pub struct SnowballingIteration {
    pub included_publications: Vec<Publication>,
    pub excluded_publications: Vec<Publication>,
    pub step: SnowballingStep,
}

#[derive(Serialize, Deserialize, Default)]
pub struct SnowballingHistory {
    pub seed: Vec<Publication>,
    pub current_iteration: SnowballingIteration,
    pub previoius_iterations: Vec<SnowballingIteration>,
    pub pending_publications: Vec<Publication>,
}

impl SnowballingHistory {
    pub fn get_all_included(&self) -> Vec<Publication> {
        vec![self.current_iteration.included_publications.clone()]
            .into_iter()
            .chain(
                self.previoius_iterations
                    .iter()
                    .map(|iter| iter.included_publications.clone()),
            )
            .flatten()
            .collect()
    }

    pub fn get_all_pending(&self) -> Vec<Publication> {
        self.pending_publications.clone()
    }
}

impl Publication {
    pub fn get_title(&self) -> Option<String> {
        self.display_name.clone()
    }

    pub fn get_year(&self) -> Option<u32> {
        self.publication_year
    }

    pub fn get_author_text(&self) -> String {
        let mut author_str = self
            .authorships
            .first()
            .map(|authorship| authorship.raw_author_name.clone())
            .expect("Papers are required to always have at least one author");

        if self.authorships.len() > 1 {
            author_str.push_str(" et al.");
        }

        author_str
    }

    pub fn get_abstract(&self) -> Option<String> {
        self.abstract_inverted_index.clone().map(|content| {
            let mut words_with_pos: Vec<(u32, &String)> = Vec::new();

            for (word, positions) in &content {
                for pos in positions {
                    words_with_pos.push((*pos, word));
                }
            }

            words_with_pos.sort_by_key(|k| k.0);

            let raw_text = words_with_pos
                .into_iter()
                .map(|(_, word)| word.as_str())
                .collect::<Vec<_>>()
                .join(" ");

            let cleaner = Builder::empty().clean(&raw_text).to_string();
            let decoded = decode_html_entities(&cleaner);

            let cleaned: String = decoded
                .chars()
                .filter(|&c| {
                    let cat = get_general_category(c);
                    !matches!(
                        cat,
                        GeneralCategory::Control
                            | GeneralCategory::Format
                            | GeneralCategory::Surrogate
                            | GeneralCategory::PrivateUse
                            | GeneralCategory::Unassigned
                    ) || c.is_whitespace()
                })
                .collect();

            cleaned.split_whitespace().collect::<Vec<_>>().join(" ")
        })
    }
}

// #[derive(Serialize, Deserialize, Debug)]
// pub struct OpenAlexResponse {
//     pub results: Vec<Publication>,
// }

pub async fn get_publication_by_id(
    api_link: &str,
    email: &str,
) -> Result<Publication, Error> {
    let url = format!("{}&mailto={}", api_link, email);

    let client = reqwest::Client::new();
    let response = client
        .get(url)
        .header("User-Agent", "Rust-OpenAlex-Client/1.0")
        .send()
        .await?
        .json::<Publication>()
        .await?;

    Ok(response)
}

// // TODO: Get all papers, not just the first page
// pub async fn get_citing_papers(
//     target_id: &str,
//     email: &str,
// ) -> Result<Vec<Publication>, Error> {
//     let url = format!(
//         "https://api.openalex.org/works?filter=cites:{}&mailto={}",
//         target_id, email
//     );
//
//     let client = reqwest::Client::new();
//     let response = client
//         .get(url)
//         .header("User-Agent", "Rust-OpenAlex-Client/1.0")
//         .send()
//         .await?
//         .json::<OpenAlexResponse>()
//         .await?;
//
//     Ok(response.results)
// }