dfir_rs/compiled/pull/half_join_state/
set.rs

1use std::borrow::Cow;
2use std::collections::VecDeque;
3use std::collections::hash_map::Entry;
4
5use super::HalfJoinState;
6
7type HashMap<K, V> = rustc_hash::FxHashMap<K, V>;
8
9use smallvec::{SmallVec, smallvec};
10
11/// [`HalfJoinState`] with set semantics.
12#[derive(Debug)]
13pub struct HalfSetJoinState<Key, ValBuild, ValProbe> {
14    // Here a smallvec with inline storage of 1 is chosen.
15    // The rationale for this decision is that, I speculate, that joins possibly have a bimodal distribution with regards to how much key contention they have.
16    // That is, I think that there are many joins that have 1 value per key on LHS/RHS, and there are also a large category of joins that have multiple values per key.
17    // For the category of joins that have multiple values per key, it's not clear why they would only have 2, 3, 4, or N specific number of values per key. So there's no good number to set the smallvec storage to.
18    // Instead we can just focus on the first group of joins that have 1 value per key and get benefit there without hurting the other group too much with excessive memory usage.
19    /// Table to probe, vec val contains all matches.
20    table: HashMap<Key, SmallVec<[ValBuild; 1]>>,
21    /// Not-yet emitted matches.
22    current_matches: VecDeque<(Key, ValProbe, ValBuild)>,
23    len: usize,
24}
25
26impl<Key, ValBuild, ValProbe> Default for HalfSetJoinState<Key, ValBuild, ValProbe> {
27    fn default() -> Self {
28        Self {
29            table: HashMap::default(),
30            current_matches: VecDeque::default(),
31            len: 0,
32        }
33    }
34}
35
36impl<Key, ValBuild, ValProbe> HalfJoinState<Key, ValBuild, ValProbe>
37    for HalfSetJoinState<Key, ValBuild, ValProbe>
38where
39    Key: Clone + Eq + std::hash::Hash,
40    ValBuild: Clone + Eq,
41    ValProbe: Clone,
42{
43    fn build(&mut self, k: Key, v: Cow<'_, ValBuild>) -> bool {
44        let entry = self.table.entry(k);
45
46        match entry {
47            Entry::Occupied(mut e) => {
48                let vec = e.get_mut();
49
50                if !vec.contains(v.as_ref()) {
51                    vec.push(v.into_owned());
52                    self.len += 1;
53                    return true;
54                }
55            }
56            Entry::Vacant(e) => {
57                e.insert(smallvec![v.into_owned()]);
58                self.len += 1;
59                return true;
60            }
61        };
62
63        false
64    }
65
66    fn probe(&mut self, k: &Key, v: &ValProbe) -> Option<(Key, ValProbe, ValBuild)> {
67        // TODO: We currently don't free/shrink the self.current_matches vecdeque to save time.
68        // This mean it will grow to eventually become the largest number of matches in a single probe call.
69        // Maybe we should clear this memory at the beginning of every tick/periodically?
70        let mut iter = self
71            .table
72            .get(k)?
73            .iter()
74            .map(|valbuild| (k.clone(), v.clone(), valbuild.clone()));
75
76        let first = iter.next();
77
78        self.current_matches.extend(iter);
79
80        first
81    }
82
83    fn full_probe(&self, k: &Key) -> std::slice::Iter<'_, ValBuild> {
84        let Some(sv) = self.table.get(k) else {
85            return [].iter();
86        };
87
88        sv.iter()
89    }
90
91    fn pop_match(&mut self) -> Option<(Key, ValProbe, ValBuild)> {
92        self.current_matches.pop_front()
93    }
94
95    fn len(&self) -> usize {
96        self.len
97    }
98
99    fn iter(&self) -> std::collections::hash_map::Iter<'_, Key, SmallVec<[ValBuild; 1]>> {
100        #[expect(clippy::disallowed_methods, reason = "FxHasher is deterministic")]
101        self.table.iter()
102    }
103
104    fn clear(&mut self) {
105        self.table.clear();
106        self.current_matches.clear();
107        self.len = 0;
108    }
109}
dfir_rs/compiled/pull/half_join_state/set.rs

dfir_rs/compiled/pull/half_join_state/
set.rs
