magmaus3/continuwuity
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

factor search tokenization out into a function

Browse source 
This ensures that the tokenization algorithm will remain in sync between
querying, indexing, and deindexing. The existing code had slightly
different behavior for querying, because it did not discard words with
>50 bytes. This was inconsequential, because >50 byte tokens are never
present in the index.

Signed-off-by: strawberry <strawberry@puppygock.gay>
This commit is contained in:
Benjamin Lee 2024-06-12 00:22:49 -04:00 • committed by June 🍓🦴
parent 81cd677b4e
commit 20a54aacd6
1 changed files with 19 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

36
src/service/rooms/search/data.rs
View file

@ -10,20 +10,26 @@ pub trait Data: Send + Sync {
fn search_pdus<'a>(&'a self, room_id: &RoomId, search_string: &str) -> SearchPdusResult<'a>; fn search_pdus<'a>(&'a self, room_id: &RoomId, search_string: &str) -> SearchPdusResult<'a>;
} }
/// Splits a string into tokens used as keys in the search inverted index
///
/// This may be used to tokenize both message bodies (for indexing) or search
/// queries (for querying).
fn tokenize(body: &str) -> impl Iterator<Item = String> + '_ {
body.split_terminator(|c: char| !c.is_alphanumeric())
.filter(|s| !s.is_empty())
.filter(|word| word.len() <= 50)
.map(str::to_lowercase)
}
impl Data for KeyValueDatabase { impl Data for KeyValueDatabase {
fn index_pdu(&self, shortroomid: u64, pdu_id: &[u8], message_body: &str) -> Result<()> { fn index_pdu(&self, shortroomid: u64, pdu_id: &[u8], message_body: &str) -> Result<()> {
let mut batch = message_body let mut batch = tokenize(message_body).map(|word| {
.split_terminator(|c: char| !c.is_alphanumeric()) let mut key = shortroomid.to_be_bytes().to_vec();
.filter(|s| !s.is_empty()) key.extend_from_slice(word.as_bytes());
.filter(|word| word.len() <= 50) key.push(0xFF);
.map(str::to_lowercase) key.extend_from_slice(pdu_id); // TODO: currently we save the room id a second time here
.map(|word| { (key, Vec::new())
let mut key = shortroomid.to_be_bytes().to_vec(); });
key.extend_from_slice(word.as_bytes());
key.push(0xFF);
key.extend_from_slice(pdu_id); // TODO: currently we save the room id a second time here
(key, Vec::new())
});
self.tokenids.insert_batch(&mut batch) self.tokenids.insert_batch(&mut batch)
} }
@ -37,11 +43,7 @@ impl Data for KeyValueDatabase {
.to_be_bytes() .to_be_bytes()
.to_vec(); .to_vec();
let words: Vec<_> = search_string let words: Vec<_> = tokenize(search_string).collect();
.split_terminator(|c: char| !c.is_alphanumeric())
.filter(|s| !s.is_empty())
.map(str::to_lowercase)
.collect();
let iterators = words.clone().into_iter().map(move |word| { let iterators = words.clone().into_iter().map(move |word| {
let mut prefix2 = prefix.clone(); let mut prefix2 = prefix.clone();