When creating the search index, omit words longer than 80 characters

This avoids creating deeply nested objects in searchindex.json
This commit is contained in:
Matthew Woodcraft 2022-05-22 13:37:19 +01:00
parent f8782666ba
commit d65ce55453

View file

@ -13,6 +13,8 @@ use crate::utils;
use serde::Serialize;
const MAX_WORD_LENGTH_TO_INDEX: usize = 80;
/// Creates all files required for search.
pub fn create_files(search_config: &Search, destination: &Path, book: &Book) -> Result<()> {
let mut index = Index::new(&["title", "body", "breadcrumbs"]);
@ -44,6 +46,15 @@ pub fn create_files(search_config: &Search, destination: &Path, book: &Book) ->
Ok(())
}
/// Tokenizes in the same way as elasticlunr-rs (for English), but also drops long tokens.
fn tokenize(text: &str) -> Vec<String> {
text.split(|c: char| c.is_whitespace() || c == '-')
.filter(|s| !s.is_empty())
.map(|s| s.trim().to_lowercase())
.filter(|s| s.len() <= MAX_WORD_LENGTH_TO_INDEX)
.collect()
}
/// Uses the given arguments to construct a search document, then inserts it to the given index.
fn add_doc(
index: &mut Index,
@ -62,7 +73,7 @@ fn add_doc(
doc_urls.push(url.into());
let items = items.iter().map(|&x| utils::collapse_whitespace(x.trim()));
index.add_doc(&doc_ref, items);
index.add_doc_with_tokenizer(&doc_ref, items, tokenize);
}
/// Renders markdown into flat unformatted text and adds it to the search index.