summary_backlog.rs

 1use collections::HashMap;
 2use fs::MTime;
 3use std::{path::Path, sync::Arc};
 4
 5const MAX_FILES_BEFORE_RESUMMARIZE: usize = 4;
 6const MAX_BYTES_BEFORE_RESUMMARIZE: u64 = 1_000_000; // 1 MB
 7
 8#[derive(Default, Debug)]
 9pub struct SummaryBacklog {
10    /// Key: path to a file that needs summarization, but that we haven't summarized yet. Value: that file's size on disk, in bytes, and its mtime.
11    files: HashMap<Arc<Path>, (u64, Option<MTime>)>,
12    /// Cache of the sum of all values in `files`, so we don't have to traverse the whole map to check if we're over the byte limit.
13    total_bytes: u64,
14}
15
16impl SummaryBacklog {
17    /// Store the given path in the backlog, along with how many bytes are in it.
18    pub fn insert(&mut self, path: Arc<Path>, bytes_on_disk: u64, mtime: Option<MTime>) {
19        let (prev_bytes, _) = self
20            .files
21            .insert(path, (bytes_on_disk, mtime))
22            .unwrap_or_default(); // Default to 0 prev_bytes
23
24        // Update the cached total by subtracting out the old amount and adding the new one.
25        self.total_bytes = self.total_bytes - prev_bytes + bytes_on_disk;
26    }
27
28    /// Returns true if the total number of bytes in the backlog exceeds a predefined threshold.
29    pub fn needs_drain(&self) -> bool {
30        self.files.len() > MAX_FILES_BEFORE_RESUMMARIZE ||
31        // The whole purpose of the cached total_bytes is to make this comparison cheap.
32        // Otherwise we'd have to traverse the entire dictionary every time we wanted this answer.
33        self.total_bytes > MAX_BYTES_BEFORE_RESUMMARIZE
34    }
35
36    /// Remove all the entries in the backlog and return the file paths as an iterator.
37    #[allow(clippy::needless_lifetimes)] // Clippy thinks this 'a can be elided, but eliding it gives a compile error
38    pub fn drain<'a>(&'a mut self) -> impl Iterator<Item = (Arc<Path>, Option<MTime>)> + 'a {
39        self.total_bytes = 0;
40
41        self.files
42            .drain()
43            .map(|(path, (_size, mtime))| (path, mtime))
44    }
45
46    pub fn len(&self) -> usize {
47        self.files.len()
48    }
49}