1use crate::{Result, arxiv, paper::Paper, provider};
2use chrono::NaiveDate;
3use serde::Deserialize;
4use std::{collections::HashMap, time::Duration};
5
6const BASE_URL: &str = "https://api.openalex.org/works";
7
8#[derive(Debug, Clone)]
9pub struct SearchQuery {
10 pub query: Option<String>,
11 pub from: Option<NaiveDate>,
12 pub to: Option<NaiveDate>,
13 pub limit: usize,
14}
15
16pub fn search_url(query: &SearchQuery, email: Option<&str>) -> String {
17 let mut url = format!(
18 "{BASE_URL}?per-page={}&sort=publication_date:desc",
19 query.limit
20 );
21 if let Some(search) = query
22 .query
23 .as_deref()
24 .filter(|query| !query.trim().is_empty())
25 {
26 url.push_str("&search=");
27 url.push_str(&urlencoding::encode(search));
28 }
29 let mut filters = Vec::new();
30 if let Some(from) = query.from {
31 filters.push(format!("from_publication_date:{from}"));
32 }
33 if let Some(to) = query.to {
34 filters.push(format!("to_publication_date:{to}"));
35 }
36 if !filters.is_empty() {
37 url.push_str("&filter=");
38 url.push_str(&urlencoding::encode(&filters.join(",")));
39 }
40 if let Some(email) = email.filter(|email| !email.trim().is_empty()) {
41 url.push_str("&mailto=");
42 url.push_str(&urlencoding::encode(email));
43 }
44 url
45}
46
47pub async fn search(query: &SearchQuery) -> Result<Vec<Paper>> {
48 let client = reqwest::Client::builder()
49 .user_agent("paper-gap/0.2")
50 .timeout(Duration::from_secs(20))
51 .build()?;
52 let email = std::env::var("OPENALEX_EMAIL").ok();
53 let body = client.get(search_url(query, email.as_deref()));
54 let body = provider::send_with_retry(body)
55 .await?
56 .error_for_status()?
57 .text()
58 .await?;
59 parse(&body)
60}
61
62pub fn parse(json: &str) -> Result<Vec<Paper>> {
63 let response: Response = serde_json::from_str(json)?;
64 Ok(response.results.into_iter().map(normalize).collect())
65}
66
67#[derive(Deserialize)]
68struct Response {
69 results: Vec<Work>,
70}
71
72#[derive(Deserialize)]
73struct Work {
74 id: String,
75 title: String,
76 doi: Option<String>,
77 publication_date: Option<String>,
78 updated_date: Option<String>,
79 abstract_inverted_index: Option<HashMap<String, Vec<usize>>>,
80 authorships: Vec<Authorship>,
81 ids: HashMap<String, String>,
82 primary_location: Option<Location>,
83 #[serde(default)]
84 topics: Vec<Topic>,
85}
86
87#[derive(Deserialize)]
88struct Authorship {
89 author: Author,
90}
91
92#[derive(Deserialize)]
93struct Author {
94 display_name: String,
95}
96
97#[derive(Deserialize)]
98struct Location {
99 landing_page_url: Option<String>,
100}
101
102#[derive(Deserialize)]
103struct Topic {
104 display_name: String,
105}
106
107fn normalize(work: Work) -> Paper {
108 let abstract_text = reconstruct_abstract(work.abstract_inverted_index.as_ref());
109 let arxiv_id = work
110 .ids
111 .get("arxiv")
112 .and_then(|url| url.rsplit('/').next())
113 .map(str::to_string);
114 let doi = work
115 .doi
116 .as_deref()
117 .map(|doi| doi.trim_start_matches("https://doi.org/").to_string());
118 let url = work
119 .primary_location
120 .and_then(|location| location.landing_page_url)
121 .unwrap_or_else(|| work.id.clone());
122 let extracted_methods = arxiv::extract_acronyms(&format!("{} {}", work.title, abstract_text));
123
124 Paper {
125 id: work.id,
126 title: work.title,
127 abstract_text,
128 authors: work
129 .authorships
130 .into_iter()
131 .map(|entry| entry.author.display_name)
132 .collect(),
133 published_date: work.publication_date,
134 updated_date: work.updated_date,
135 doi,
136 arxiv_id,
137 url,
138 categories: work
139 .topics
140 .into_iter()
141 .map(|topic| topic.display_name)
142 .collect(),
143 extracted_methods,
144 source_provider: "openalex".into(),
145 }
146}
147
148fn reconstruct_abstract(index: Option<&HashMap<String, Vec<usize>>>) -> String {
149 let Some(index) = index else {
150 return String::new();
151 };
152 let Some(max) = index.values().flatten().max().copied() else {
153 return String::new();
154 };
155 let mut words = vec![String::new(); max + 1];
156 for (word, positions) in index {
157 for &position in positions {
158 words[position] = word.clone();
159 }
160 }
161 words
162 .into_iter()
163 .filter(|word| !word.is_empty())
164 .collect::<Vec<_>>()
165 .join(" ")
166}