rust - Rayon 折叠成一个 HashMap

Question

我有一个 Vec<Result<MyStruct, ()>> ，我想将它们放入 HashMap 中的桶中，其中键是 MyStruct.b field 和 values 是一个包含所有 MyStructs 的 Vec MyStruct.b领域(见下面的 Playground 链接，可能更清楚)。我正在使用折叠，它适用于普通的 .fold

let my_array = vec![Ok(m1), Ok(m2), Ok(m3), Err(())];

let result: HashMap<String, Vec<&MyStruct>> = my_array
    .iter()
    .filter_map(|value| value.as_ref().ok())
    .fold(HashMap::new(), |mut acc, value| {
        acc.entry(value.b.clone()).or_insert(vec![]).push(&value);
        acc
    });

但是，我无法用人造丝实现同样的效果

let my_array = vec![Ok(m1), Ok(m2), Ok(m3), Err(())];

let result = my_array
    .par_iter()
    .filter_map(|value| value.as_ref().ok())
    .fold(
        ||HashMap::new(),
        |mut acc, value| {
            acc.entry(value.b.clone()).or_insert(vec![]).push(&value);
            acc
        }
    )
    .collect::<HashMap<String, Vec<&MyStruct>>>();

我得到的错误是

   Compiling playground v0.0.1 (/playground)
error[E0277]: the trait bound `std::collections::HashMap<std::string::String, std::vec::Vec<&MyStruct>>: rayon::iter::FromParallelIterator<std::collections::HashMap<std::string::String, std::vec::Vec<&&MyStruct>>>` is not satisfied
  --> src/main.rs:39:10
   |
39 |         .collect::<HashMap<String, Vec<&MyStruct>>>();
   |          ^^^^^^^ the trait `rayon::iter::FromParallelIterator<std::collections::HashMap<std::string::String, std::vec::Vec<&&MyStruct>>>` is not implemented for `std::collections::HashMap<std::string::String, std::vec::Vec<&MyStruct>>`
   |
   = help: the following implementations were found:
             <std::collections::HashMap<K, V, S> as rayon::iter::FromParallelIterator<(K, V)>>

Playground

Answer 1

根据 parallel fold documentation :

The parallel fold works similarly except that it first breaks up your list into sublists, and hence instead of yielding up a single sum at the end, it yields up multiple sums. The number of results is nondeterministic, as is the point where the breaks occur.

所以你想reduce将fold的结果变成一个值作为最终结果，而不是collect:

let result = my_array
    .par_iter()
    .filter_map(|value| value.as_ref().ok())
    .fold(
        || HashMap::new(),
        |mut acc, value| {
            acc.entry(value.b.clone()).or_insert(vec![]).push(value);
            acc
        },
    )
    .reduce(
        || HashMap::new(),
        |m1, m2| {
            m2.iter().fold(m1, |mut acc, (k, vs)| {
                acc.entry(k.clone()).or_insert(vec![]).extend(vs);
                acc
            })
        },
    );

OTOH :

reduce() requires that the identity function has the same type as the things you are iterating over, and it fully reduces the list of items into a single item

所以你可以一次性reduce到最终结果:

let result = my_array
    .par_iter()
    .filter_map(|value| value.as_ref().ok())
    .map(|s| {
        let mut m = HashMap::new();
        m.insert(s.b.clone(), vec![s]);
        m
    })
    .reduce(
        || HashMap::new(),
        |m1, m2| {
            m2.iter().fold(m1, |mut acc, (k, vs)| {
                acc.entry(k.clone()).or_insert(vec![]).extend(vs);
                acc
            })
        },
    );

选择fold or map/reduce由您决定。