RomanNum3ral
/
zoxide
mirror of https://github.com/ajeetdsouza/zoxide.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge 5b88d3a6b3 into 67ca1bc959

This commit is contained in:
lyj 2026-03-24 09:54:33 +09:00 committed by GitHub
parent 67ca1bc959 5b88d3a6b3
commit db25e08617
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
7 changed files with 410 additions and 271 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
CHANGELOG.md
View File

@ -12,6 +12,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
### Added ### Added
- POSIX: support for non-Cygwin Windows environments (e.g. Busybox). - POSIX: support for non-Cygwin Windows environments (e.g. Busybox).
- Automatically migrate data from the legacy bincode-backed `db.zo` file
when upgrading to the sqlite backend (only performed if no
`db.sqlite3` exists).
### Fixed ### Fixed

92
Cargo.lock generated
View File

@ -2,6 +2,18 @@
# It is not intended for manual editing. # It is not intended for manual editing.
version = 4 version = 4
[[package]]
name = "ahash"
version = "0.8.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5a15f179cd60c4584b8a8c596927aadc462e27f2ca70c04e0071964a73ba7a75"
dependencies = [
"cfg-if",
"once_cell",
"version_check",
"zerocopy",
]
[[package]] [[package]]
name = "aho-corasick" name = "aho-corasick"
version = "1.1.3" version = "1.1.3"
@ -17,6 +29,12 @@ version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "250f629c0161ad8107cf89319e990051fae62832fd343083bea452d93e2205fd" checksum = "250f629c0161ad8107cf89319e990051fae62832fd343083bea452d93e2205fd"
[[package]]
name = "allocator-api2"
version = "0.2.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "683d7910e743518b0e34f1186f92494becacb047c7b6bf616c96772180fef923"
[[package]] [[package]]
name = "anstream" name = "anstream"
version = "0.6.18" version = "0.6.18"
@ -321,6 +339,18 @@ dependencies = [
"windows-sys", "windows-sys",
] ]
[[package]]
name = "fallible-iterator"
version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2acce4a10f12dc2fb14a218589d4f1f62ef011b2d0cc4b3cb1bba8e94da14649"
[[package]]
name = "fallible-streaming-iterator"
version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7360491ce676a36bf9bb3c56c1aa791658183a54d2744120f27285738d90465a"
[[package]] [[package]]
name = "fastrand" name = "fastrand"
version = "2.3.0" version = "2.3.0"
@ -356,6 +386,25 @@ version = "0.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a8d1add55171497b4705a648c6b583acafb01d58050a51727785f0b2c8e0a2b2" checksum = "a8d1add55171497b4705a648c6b583acafb01d58050a51727785f0b2c8e0a2b2"
[[package]]
name = "hashbrown"
version = "0.14.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e5274423e17b7c9fc20b6e7e208532f9b19825d82dfd615708b70edd83df41f1"
dependencies = [
"ahash",
"allocator-api2",
]
[[package]]
name = "hashlink"
version = "0.8.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e8094feaf31ff591f651a2664fb9cfd92bba7a60ce3197265e9482ebe753c8f7"
dependencies = [
"hashbrown",
]
[[package]] [[package]]
name = "heck" name = "heck"
version = "0.4.1" version = "0.4.1"
@ -396,6 +445,16 @@ dependencies = [
"libc", "libc",
] ]
[[package]]
name = "libsqlite3-sys"
version = "0.27.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cf4e226dcd58b4be396f7bd3c20da8fdee2911400705297ba7d2d7cc2c30f716"
dependencies = [
"pkg-config",
"vcpkg",
]
[[package]] [[package]]
name = "linux-raw-sys" name = "linux-raw-sys"
version = "0.9.4" version = "0.9.4"
@ -484,6 +543,12 @@ version = "2.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e3148f5046208a5d56bcfc03053e3ca6334e51da8dfb19b6cdc8b306fae3283e" checksum = "e3148f5046208a5d56bcfc03053e3ca6334e51da8dfb19b6cdc8b306fae3283e"
[[package]]
name = "pkg-config"
version = "0.3.32"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7edddbd0b52d732b21ad9a5fab5c704c14cd949e5e9a1ec5929a24fded1b904c"
[[package]] [[package]]
name = "ppv-lite86" name = "ppv-lite86"
version = "0.2.21" version = "0.2.21"
@ -670,6 +735,20 @@ dependencies = [
"syn", "syn",
] ]
[[package]]
name = "rusqlite"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a78046161564f5e7cd9008aff3b2990b3850dc8e0349119b98e8f251e099f24d"
dependencies = [
"bitflags",
"fallible-iterator",
"fallible-streaming-iterator",
"hashlink",
"libsqlite3-sys",
"smallvec",
]
[[package]] [[package]]
name = "rustc-hash" name = "rustc-hash"
version = "2.1.1" version = "2.1.1"
@ -742,6 +821,12 @@ dependencies = [
"serde", "serde",
] ]
[[package]]
name = "smallvec"
version = "1.15.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "67b1b7a3b5fe4f1376887184045fcf45c69e92af734b7aaddc05fb777b6fbd03"
[[package]] [[package]]
name = "static_assertions" name = "static_assertions"
version = "1.1.0" version = "1.1.0"
@ -816,6 +901,12 @@ version = "0.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821" checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"
[[package]]
name = "vcpkg"
version = "0.2.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
[[package]] [[package]]
name = "version_check" name = "version_check"
version = "0.9.5" version = "0.9.5"
@ -1002,6 +1093,7 @@ dependencies = [
"ouroboros", "ouroboros",
"rstest", "rstest",
"rstest_reuse", "rstest_reuse",
"rusqlite",
"serde", "serde",
"tempfile", "tempfile",
"which", "which",

3
Cargo.toml
View File

@ -21,7 +21,7 @@ askama = { version = "0.14.0", default-features = false, features = [
"derive", "derive",
"std", "std",
] } ] }
bincode = "1.3.1" rusqlite = "0.30.0"
clap = { version = "4.3.0", features = ["derive"] } clap = { version = "4.3.0", features = ["derive"] }
color-print = "0.3.4" color-print = "0.3.4"
dirs = "6.0.0" dirs = "6.0.0"
@ -30,6 +30,7 @@ fastrand = "2.0.0"
glob = "0.3.0" glob = "0.3.0"
ouroboros = "0.18.3" ouroboros = "0.18.3"
serde = { version = "1.0.116", features = ["derive"] } serde = { version = "1.0.116", features = ["derive"] }
bincode = "1.3.1"
[target.'cfg(unix)'.dependencies] [target.'cfg(unix)'.dependencies]
nix = { version = "0.30.1", default-features = false, features = [ nix = { version = "0.30.1", default-features = false, features = [

16
src/cmd/import.rs
View File

@ -40,7 +40,7 @@ fn import_autojump(db: &mut Database, buffer: &str) -> Result<()> {
// take a while to normalize. // take a while to normalize.
rank = sigmoid(rank); rank = sigmoid(rank);
db.add_unchecked(path, rank, 0); db.add_unchecked_merge(path, rank, 0);
} }
if db.dirty() { if db.dirty() {
@ -65,7 +65,7 @@ fn import_z(db: &mut Database, buffer: &str) -> Result<()> {
let path = split.next().with_context(|| format!("invalid entry: {line}"))?; let path = split.next().with_context(|| format!("invalid entry: {line}"))?;
db.add_unchecked(path, rank, last_accessed); db.add_unchecked_merge(path, rank, last_accessed);
} }
if db.dirty() { if db.dirty() {
@ -107,12 +107,12 @@ mod tests {
println!("got: {:?}", &db.dirs()); println!("got: {:?}", &db.dirs());
let exp = [ let exp = [
Dir { path: "/baz".into(), rank: sigmoid(7.0), last_accessed: 0 },
Dir { path: "/corge/grault/garply".into(), rank: 6.0, last_accessed: 600 },
Dir { path: "/foo/bar".into(), rank: 9.0 + sigmoid(2.0), last_accessed: 900 },
Dir { path: "/quux/quuz".into(), rank: 1.0 + sigmoid(5.0), last_accessed: 100 }, Dir { path: "/quux/quuz".into(), rank: 1.0 + sigmoid(5.0), last_accessed: 100 },
Dir { path: "/corge/grault/garply".into(), rank: 6.0, last_accessed: 600 },
Dir { path: "/waldo/fred/plugh".into(), rank: 3.0, last_accessed: 300 }, Dir { path: "/waldo/fred/plugh".into(), rank: 3.0, last_accessed: 300 },
Dir { path: "/xyzzy/thud".into(), rank: 8.0, last_accessed: 800 }, Dir { path: "/xyzzy/thud".into(), rank: 8.0, last_accessed: 800 },
Dir { path: "/foo/bar".into(), rank: 9.0 + sigmoid(2.0), last_accessed: 900 },
Dir { path: "/baz".into(), rank: sigmoid(7.0), last_accessed: 0 },
]; ];
println!("exp: {exp:?}"); println!("exp: {exp:?}");
@ -148,12 +148,12 @@ mod tests {
println!("got: {:?}", &db.dirs()); println!("got: {:?}", &db.dirs());
let exp = [ let exp = [
Dir { path: "/baz".into(), rank: 7.0, last_accessed: 700 },
Dir { path: "/corge/grault/garply".into(), rank: 6.0, last_accessed: 600 },
Dir { path: "/foo/bar".into(), rank: 11.0, last_accessed: 900 },
Dir { path: "/quux/quuz".into(), rank: 10.0, last_accessed: 500 }, Dir { path: "/quux/quuz".into(), rank: 10.0, last_accessed: 500 },
Dir { path: "/corge/grault/garply".into(), rank: 6.0, last_accessed: 600 },
Dir { path: "/waldo/fred/plugh".into(), rank: 3.0, last_accessed: 300 }, Dir { path: "/waldo/fred/plugh".into(), rank: 3.0, last_accessed: 300 },
Dir { path: "/xyzzy/thud".into(), rank: 8.0, last_accessed: 800 }, Dir { path: "/xyzzy/thud".into(), rank: 8.0, last_accessed: 800 },
Dir { path: "/foo/bar".into(), rank: 11.0, last_accessed: 900 },
Dir { path: "/baz".into(), rank: 7.0, last_accessed: 700 },
]; ];
println!("exp: {exp:?}"); println!("exp: {exp:?}");

439
src/db/mod.rs
View File

@ -1,24 +1,64 @@
mod dir; mod dir;
mod stream; mod stream;
use std::fs;
use std::path::{Path, PathBuf}; use std::path::{Path, PathBuf};
use std::{fs, io};
use anyhow::{Context, Result, bail}; use anyhow::{Context, Result};
use rusqlite::{Connection, OptionalExtension, params};
use bincode::Options; use bincode::Options;
use ouroboros::self_referencing;
use crate::config;
pub use crate::db::dir::{Dir, Epoch, Rank}; pub use crate::db::dir::{Dir, Epoch, Rank};
pub use crate::db::stream::{Stream, StreamOptions}; pub use crate::db::stream::{Stream, StreamOptions};
use crate::{config, util};
#[self_referencing] /// Attempt to read an old bincode-formatted database and write its entries into
/// the provided SQLite connection.
fn migrate_from_bincode(conn: &mut Connection, old_path: &Path) -> Result<()> {
// The on-disk format used by the legacy database is very similar to the
// struct definitions we already use. We simply serialise a version number
// followed by a `Vec<Dir>`.
const MAX_SIZE: u64 = 32 << 20; // 32 MiB
let data = fs::read(old_path)
.with_context(|| format!("could not read legacy database: {}", old_path.display()))?;
let deserializer = &mut bincode::options()
.with_fixint_encoding()
.with_limit(MAX_SIZE);
let version_size = deserializer.serialized_size(&Database::VERSION)? as usize;
if data.len() < version_size {
anyhow::bail!("legacy database is corrupted");
}
let (bytes_version, bytes_dirs) = data.split_at(version_size);
let version: u32 = deserializer.deserialize(bytes_version)?;
if version != Database::VERSION {
anyhow::bail!(
"unsupported legacy database version (got {}, expected {})",
version,
Database::VERSION
);
}
let dirs: Vec<Dir<'_>> = deserializer.deserialize(bytes_dirs)?;
let tx = conn.transaction()?;
for dir in dirs {
let path_s: String = dir.path.into_owned();
tx.execute(
"INSERT OR REPLACE INTO dirs (path, rank, last_accessed) VALUES (?1, ?2, ?3)",
params![&path_s, dir.rank, dir.last_accessed],
)?;
}
tx.commit()?;
Ok(())
}
pub struct Database { pub struct Database {
path: PathBuf, conn: Connection,
bytes: Vec<u8>,
#[borrows(bytes)]
#[covariant]
pub dirs: Vec<Dir<'this>>,
dirty: bool, dirty: bool,
} }
@ -32,203 +72,267 @@ impl Database {
pub fn open_dir(data_dir: impl AsRef<Path>) -> Result<Self> { pub fn open_dir(data_dir: impl AsRef<Path>) -> Result<Self> {
let data_dir = data_dir.as_ref(); let data_dir = data_dir.as_ref();
let path = data_dir.join("db.zo"); let path = data_dir.join("db.sqlite3");
let path = fs::canonicalize(&path).unwrap_or(path); let path = fs::canonicalize(&path).unwrap_or(path);
match fs::read(&path) { fs::create_dir_all(data_dir)
Ok(bytes) => Self::try_new(path, bytes, |bytes| Self::deserialize(bytes), false), .with_context(|| format!("unable to create data directory: {}", data_dir.display()))?;
Err(e) if e.kind() == io::ErrorKind::NotFound => {
// Create data directory, but don't create any file yet. The file will be // Open or create sqlite database file.
// created later by [`Database::save`] if any data is modified. let existed = path.exists();
fs::create_dir_all(data_dir).with_context(|| {
format!("unable to create data directory: {}", data_dir.display()) let mut conn = Connection::open(&path)
})?; .with_context(|| format!("could not open database: {}", path.display()))?;
Ok(Self::new(path, Vec::new(), |_| Vec::new(), false))
} // Enable WAL for better concurrency and durability.
Err(e) => { conn.pragma_update(None, "journal_mode", &"WAL").ok();
Err(e).with_context(|| format!("could not read from database: {}", path.display()))
// Create table if it doesn't exist.
conn.execute_batch(
"CREATE TABLE IF NOT EXISTS dirs (
path TEXT PRIMARY KEY,
rank REAL NOT NULL,
last_accessed INTEGER NOT NULL
);",
)?;
// If the sqlite database didn't previously exist, attempt to migrate data
// from the legacy bincode-backed file. This keeps behaviour identical to
// older versions of zoxide while ensuring users transparently upgrade.
if !existed {
let old_path = data_dir.join("db.zo");
if old_path.exists() {
// Migration errors shouldn't prevent the program from running;
// just print a warning so users can investigate.
if let Err(e) = migrate_from_bincode(&mut conn, &old_path) {
eprintln!(
"warning: failed to migrate legacy database ({}): {}",
old_path.display(),
e
);
}
} }
} }
Ok(Database { conn, dirty: false })
} }
pub fn save(&mut self) -> Result<()> { pub fn save(&mut self) -> Result<()> {
// Only write to disk if the database is modified. // For SQLite, write operations are applied immediately via transactions.
if !self.dirty() { // Keep save() for compatibility; do nothing.
return Ok(()); self.dirty = false;
}
let bytes = Self::serialize(self.dirs())?;
util::write(self.borrow_path(), bytes).context("could not write to database")?;
self.with_dirty_mut(|dirty| *dirty = false);
Ok(()) Ok(())
} }
/// Increments the rank of a directory, or creates it if it does not exist. /// Increments the rank of a directory, or creates it if it does not exist.
pub fn add(&mut self, path: impl AsRef<str> + Into<String>, by: Rank, now: Epoch) { pub fn add(&mut self, path: impl AsRef<str> + Into<String>, by: Rank, now: Epoch) {
self.with_dirs_mut(|dirs| match dirs.iter_mut().find(|dir| dir.path == path.as_ref()) { let path_s: String = path.into();
Some(dir) => dir.rank = (dir.rank + by).max(0.0), let tx = match self.conn.transaction() {
None => { Ok(t) => t,
dirs.push(Dir { path: path.into().into(), rank: by.max(0.0), last_accessed: now }) Err(_) => return,
};
let existing: Option<(f64, u64)> = tx
.query_row(
"SELECT rank, last_accessed FROM dirs WHERE path = ?1",
params![&path_s],
|row| Ok((row.get(0)?, row.get(1)?)),
)
.optional()
.unwrap_or(None);
match existing {
Some((rank, _last)) => {
let new_rank = (rank + by).max(0.0);
let _ = tx.execute(
"UPDATE dirs SET rank = ?1 WHERE path = ?2",
params![new_rank, &path_s],
);
} }
}); None => {
self.with_dirty_mut(|dirty| *dirty = true); let _ = tx.execute(
"INSERT INTO dirs (path, rank, last_accessed) VALUES (?1, ?2, ?3)",
params![&path_s, by.max(0.0), now],
);
}
}
let _ = tx.commit();
self.dirty = true;
} }
/// Creates a new directory. This will create a duplicate entry if this /// Creates a new directory. This will create a duplicate entry if this
/// directory is already in the database, it is expected that the user /// directory is already in the database, it is expected that the user
/// either does a check before calling this, or calls `dedup()` /// either does a check before calling this, or calls `dedup()`
/// afterward. /// afterward.
#[cfg(test)]
pub fn add_unchecked(&mut self, path: impl AsRef<str> + Into<String>, rank: Rank, now: Epoch) { pub fn add_unchecked(&mut self, path: impl AsRef<str> + Into<String>, rank: Rank, now: Epoch) {
self.with_dirs_mut(|dirs| { let path_s: String = path.into();
dirs.push(Dir { path: path.into().into(), rank, last_accessed: now }) let _ = self.conn.execute(
}); "INSERT OR REPLACE INTO dirs (path, rank, last_accessed) VALUES (?1, ?2, ?3)",
self.with_dirty_mut(|dirty| *dirty = true); params![&path_s, rank, now],
);
self.dirty = true;
}
/// choose the max `now`
/// sum `rank`
pub fn add_unchecked_merge(&mut self, path: impl AsRef<str> + Into<String>, rank: Rank, now: Epoch) {
let path_s: String = path.into();
let _ = self.conn.execute(
"INSERT INTO dirs (path, rank, last_accessed) VALUES (?1, ?2, ?3)
ON CONFLICT(path) DO UPDATE SET
rank = dirs.rank + excluded.rank,
last_accessed = MAX(dirs.last_accessed, excluded.last_accessed)",
params![&path_s, rank, now],
);
self.dirty = true;
} }
/// Increments the rank and updates the last_accessed of a directory, or /// Increments the rank and updates the last_accessed of a directory, or
/// creates it if it does not exist. /// creates it if it does not exist.
pub fn add_update(&mut self, path: impl AsRef<str> + Into<String>, by: Rank, now: Epoch) { pub fn add_update(&mut self, path: impl AsRef<str> + Into<String>, by: Rank, now: Epoch) {
self.with_dirs_mut(|dirs| match dirs.iter_mut().find(|dir| dir.path == path.as_ref()) { let path_s: String = path.into();
Some(dir) => { let tx = match self.conn.transaction() {
dir.rank = (dir.rank + by).max(0.0); Ok(t) => t,
dir.last_accessed = now; Err(_) => return,
};
let existing: Option<(f64, u64)> = tx
.query_row(
"SELECT rank, last_accessed FROM dirs WHERE path = ?1",
params![&path_s],
|row| Ok((row.get(0)?, row.get(1)?)),
)
.optional()
.unwrap_or(None);
match existing {
Some((rank, _)) => {
let new_rank = (rank + by).max(0.0);
let _ = tx.execute(
"UPDATE dirs SET rank = ?1, last_accessed = ?2 WHERE path = ?3",
params![new_rank, now, &path_s],
);
} }
None => { None => {
dirs.push(Dir { path: path.into().into(), rank: by.max(0.0), last_accessed: now }) let _ = tx.execute(
"INSERT INTO dirs (path, rank, last_accessed) VALUES (?1, ?2, ?3)",
params![&path_s, by.max(0.0), now],
);
} }
}); }
self.with_dirty_mut(|dirty| *dirty = true);
let _ = tx.commit();
self.dirty = true;
} }
/// Removes the directory with `path` from the store. This does not preserve /// Removes the directory with `path` from the store. Returns true if an
/// ordering, but is O(1). /// entry was deleted.
pub fn remove(&mut self, path: impl AsRef<str>) -> bool { pub fn remove(&mut self, path: impl AsRef<str>) -> bool {
match self.dirs().iter().position(|dir| dir.path == path.as_ref()) { let path_s = path.as_ref();
Some(idx) => { match self.conn.execute("DELETE FROM dirs WHERE path = ?1", params![path_s]) {
self.swap_remove(idx); Ok(count) => {
true if count > 0 {
self.dirty = true;
true
} else {
false
}
} }
None => false, Err(_) => false,
} }
} }
pub fn swap_remove(&mut self, idx: usize) { pub fn swap_remove(&mut self, _idx: usize) {
self.with_dirs_mut(|dirs| dirs.swap_remove(idx)); // In the sqlite-backed implementation we don't maintain an in-memory
self.with_dirty_mut(|dirty| *dirty = true); // vector, so this is a no-op. Higher-level code that relies on
// indices shouldn't be calling this directly except within the
// streaming logic which uses Database::dirs(). For compatibility, keep
// the method but do nothing.
self.dirty = true;
} }
pub fn age(&mut self, max_age: Rank) { pub fn age(&mut self, max_age: Rank) {
let mut dirty = false; // Apply the aging algorithm to all rows.
self.with_dirs_mut(|dirs| { // Collect entries first to avoid holding a Statement borrow while starting
let total_age = dirs.iter().map(|dir| dir.rank).sum::<Rank>(); // a transaction on the connection.
if total_age > max_age { let mut entries = Vec::new();
let factor = 0.9 * max_age / total_age; if let Ok(mut stmt) = self.conn.prepare("SELECT path, rank FROM dirs") {
for idx in (0..dirs.len()).rev() { if let Ok(rows) =
let dir = &mut dirs[idx]; stmt.query_map([], |row| Ok((row.get::<_, String>(0)?, row.get::<_, f64>(1)?)))
dir.rank *= factor; {
if dir.rank < 1.0 { for r in rows {
dirs.swap_remove(idx); if let Ok((path, rank)) = r {
entries.push((path, rank));
} }
} }
dirty = true;
} }
}); }
self.with_dirty_mut(|dirty_prev| *dirty_prev |= dirty);
let total_age: f64 = entries.iter().map(|(_, rank)| *rank).sum();
if total_age > max_age {
let factor = 0.9 * max_age / total_age;
if let Ok(tx) = self.conn.transaction() {
for (path, rank) in entries {
let new_rank = rank * factor;
if new_rank < 1.0 {
let _ = tx.execute("DELETE FROM dirs WHERE path = ?1", params![path]);
} else {
let _ = tx.execute(
"UPDATE dirs SET rank = ?1 WHERE path = ?2",
params![new_rank, path],
);
}
}
let _ = tx.commit();
self.dirty = true;
}
}
} }
pub fn dedup(&mut self) { pub fn dedup(&mut self) {
// Sort by path, so that equal paths are next to each other. // Using path as PRIMARY KEY ensures uniqueness, nothing to do here.
self.sort_by_path();
let mut dirty = false;
self.with_dirs_mut(|dirs| {
for idx in (1..dirs.len()).rev() {
// Check if curr_dir and next_dir have equal paths.
let curr_dir = &dirs[idx];
let next_dir = &dirs[idx - 1];
if next_dir.path != curr_dir.path {
continue;
}
// Merge curr_dir's rank and last_accessed into next_dir.
let rank = curr_dir.rank;
let last_accessed = curr_dir.last_accessed;
let next_dir = &mut dirs[idx - 1];
next_dir.last_accessed = next_dir.last_accessed.max(last_accessed);
next_dir.rank += rank;
// Delete curr_dir.
dirs.swap_remove(idx);
dirty = true;
}
});
self.with_dirty_mut(|dirty_prev| *dirty_prev |= dirty);
} }
#[cfg(test)]
pub fn sort_by_path(&mut self) { pub fn sort_by_path(&mut self) {
self.with_dirs_mut(|dirs| dirs.sort_unstable_by(|dir1, dir2| dir1.path.cmp(&dir2.path))); // Sorting is done at query time in the sqlite-backed implementation.
self.with_dirty_mut(|dirty| *dirty = true);
} }
pub fn sort_by_score(&mut self, now: Epoch) { pub fn sort_by_score(&mut self, _now: Epoch) {
self.with_dirs_mut(|dirs| { // Sorting is done at query time in the sqlite-backed implementation.
dirs.sort_unstable_by(|dir1: &Dir, dir2: &Dir| {
dir1.score(now).total_cmp(&dir2.score(now))
})
});
self.with_dirty_mut(|dirty| *dirty = true);
} }
pub fn dirty(&self) -> bool { pub fn dirty(&self) -> bool {
*self.borrow_dirty() self.dirty
} }
pub fn dirs(&self) -> &[Dir<'_>] { pub fn dirs(&self) -> Vec<Dir<'static>> {
self.borrow_dirs() // Load all dirs from the database into an owned Vec.
} let mut stmt = match self.conn.prepare("SELECT path, rank, last_accessed FROM dirs") {
Ok(s) => s,
fn serialize(dirs: &[Dir<'_>]) -> Result<Vec<u8>> { Err(_) => return Vec::new(),
(|| -> bincode::Result<_> {
// Preallocate buffer with combined size of sections.
let buffer_size =
bincode::serialized_size(&Self::VERSION)? + bincode::serialized_size(&dirs)?;
let mut buffer = Vec::with_capacity(buffer_size as usize);
// Serialize sections into buffer.
bincode::serialize_into(&mut buffer, &Self::VERSION)?;
bincode::serialize_into(&mut buffer, &dirs)?;
Ok(buffer)
})()
.context("could not serialize database")
}
fn deserialize(bytes: &[u8]) -> Result<Vec<Dir<'_>>> {
// Assume a maximum size for the database. This prevents bincode from throwing
// strange errors when it encounters invalid data.
const MAX_SIZE: u64 = 32 << 20; // 32 MiB
let deserializer = &mut bincode::options().with_fixint_encoding().with_limit(MAX_SIZE);
// Split bytes into sections.
let version_size = deserializer.serialized_size(&Self::VERSION).unwrap() as _;
if bytes.len() < version_size {
bail!("could not deserialize database: corrupted data");
}
let (bytes_version, bytes_dirs) = bytes.split_at(version_size);
// Deserialize sections.
let version = deserializer.deserialize(bytes_version)?;
let dirs = match version {
Self::VERSION => {
deserializer.deserialize(bytes_dirs).context("could not deserialize database")?
}
version => {
bail!("unsupported version (got {version}, supports {})", Self::VERSION)
}
}; };
Ok(dirs) let rows = stmt.query_map([], |row| {
Ok(Dir {
path: row.get::<_, String>(0)?.into(),
rank: row.get::<_, f64>(1)?,
last_accessed: row.get::<_, u64>(2)?,
})
});
let mut out = Vec::new();
if let Ok(map) = rows {
for r in map {
if let Ok(dir) = r {
out.push(dir);
}
}
}
out
} }
} }
@ -253,7 +357,8 @@ mod tests {
let db = Database::open_dir(data_dir.path()).unwrap(); let db = Database::open_dir(data_dir.path()).unwrap();
assert_eq!(db.dirs().len(), 1); assert_eq!(db.dirs().len(), 1);
let dir = &db.dirs()[0]; let dirs = db.dirs();
let dir = &dirs[0];
assert_eq!(dir.path, path); assert_eq!(dir.path, path);
assert!((dir.rank - 2.0).abs() < 0.01); assert!((dir.rank - 2.0).abs() < 0.01);
assert_eq!(dir.last_accessed, now); assert_eq!(dir.last_accessed, now);
@ -285,4 +390,34 @@ mod tests {
db.save().unwrap(); db.save().unwrap();
} }
} }
#[test]
fn migrate_from_bincode() {
let data_dir = tempfile::tempdir().unwrap();
let old_path = data_dir.path().join("db.zo");
// prepare a legacy file with one entry
let dirs = vec![Dir {
path: "/foo".into(),
rank: 1.0,
last_accessed: 12345,
}];
let mut bytes = Vec::new();
let mut serializer = bincode::options().with_fixint_encoding();
serializer.serialize_into(&mut bytes, &Database::VERSION).unwrap();
serializer.serialize_into(&mut bytes, &dirs).unwrap();
fs::write(&old_path, &bytes).unwrap();
// opening should automatically migrate the data
let db = Database::open_dir(data_dir.path()).unwrap();
let dirs = db.dirs();
assert_eq!(dirs.len(), 1);
assert_eq!(dirs[0].path, "/foo");
assert!((dirs[0].rank - 1.0).abs() < f64::EPSILON);
assert_eq!(dirs[0].last_accessed, 12345);
// sqlite file should exist after opening
assert!(data_dir.path().join("db.sqlite3").exists());
// old file is left intact so future runs are no-ops
assert!(old_path.exists());
}
} }

32
src/db/stream.rs
View File

@ -1,5 +1,3 @@
use std::iter::Rev;
use std::ops::Range;
use std::path::Path; use std::path::Path;
use std::{fs, path}; use std::{fs, path};
@ -10,20 +8,25 @@ use crate::util::{self, MONTH};
pub struct Stream<'a> { pub struct Stream<'a> {
db: &'a mut Database, db: &'a mut Database,
idxs: Rev<Range<usize>>, entries: Vec<Dir<'static>>,
pos: usize,
options: StreamOptions, options: StreamOptions,
} }
impl<'a> Stream<'a> { impl<'a> Stream<'a> {
pub fn new(db: &'a mut Database, options: StreamOptions) -> Self { pub fn new(db: &'a mut Database, options: StreamOptions) -> Self {
db.sort_by_score(options.now); // Load entries and sort by score.
let idxs = (0..db.dirs().len()).rev(); let mut entries = db.dirs();
Stream { db, idxs, options } entries.sort_unstable_by(|a, b| a.score(options.now).total_cmp(&b.score(options.now)));
// iterate from highest to lowest
entries.reverse();
Stream { db, entries, pos: 0, options }
} }
pub fn next(&mut self) -> Option<&Dir<'_>> { pub fn next(&mut self) -> Option<&Dir<'_>> {
while let Some(idx) = self.idxs.next() { while self.pos < self.entries.len() {
let dir = &self.db.dirs()[idx]; let dir = &self.entries[self.pos];
self.pos += 1;
if !self.filter_by_keywords(&dir.path) { if !self.filter_by_keywords(&dir.path) {
continue; continue;
@ -34,20 +37,20 @@ impl<'a> Stream<'a> {
} }
if !self.filter_by_exclude(&dir.path) { if !self.filter_by_exclude(&dir.path) {
self.db.swap_remove(idx); // lazily remove from database
let _ = self.db.remove(&*dir.path);
continue; continue;
} }
// Exists queries are slow, this should always be checked last. // Exists queries are slow, this should always be checked last.
if !self.filter_by_exists(&dir.path) { if !self.filter_by_exists(&dir.path) {
if dir.last_accessed < self.options.ttl { if dir.last_accessed < self.options.ttl {
self.db.swap_remove(idx); let _ = self.db.remove(&*dir.path);
} }
continue; continue;
} }
let dir = &self.db.dirs()[idx]; return Some(&self.entries[self.pos - 1]);
return Some(dir);
} }
None None
@ -203,9 +206,10 @@ mod tests {
#[case(&["/foo/", "/bar"], "/foo/bar", false)] #[case(&["/foo/", "/bar"], "/foo/bar", false)]
#[case(&["/foo/", "/bar"], "/foo/baz/bar", true)] #[case(&["/foo/", "/bar"], "/foo/baz/bar", true)]
fn query(#[case] keywords: &[&str], #[case] path: &str, #[case] is_match: bool) { fn query(#[case] keywords: &[&str], #[case] path: &str, #[case] is_match: bool) {
let db = &mut Database::new(PathBuf::new(), Vec::new(), |_| Vec::new(), false); let data_dir = tempfile::tempdir().unwrap();
let mut db = Database::open_dir(data_dir.path()).unwrap();
let options = StreamOptions::new(0).with_keywords(keywords.iter()); let options = StreamOptions::new(0).with_keywords(keywords.iter());
let stream = Stream::new(db, options); let stream = Stream::new(&mut db, options);
assert_eq!(is_match, stream.filter_by_keywords(path)); assert_eq!(is_match, stream.filter_by_keywords(path));
} }
} }

96
src/util.rs
View File

@ -1,5 +1,4 @@
use std::ffi::OsStr; use std::ffi::OsStr;
use std::fs::{self, File, OpenOptions};
use std::io::{self, Read, Write}; use std::io::{self, Read, Write};
use std::path::{Component, Path, PathBuf}; use std::path::{Component, Path, PathBuf};
use std::process::{Child, Command, Stdio}; use std::process::{Child, Command, Stdio};
@ -150,101 +149,6 @@ impl FzfChild {
} }
} }
/// Similar to [`fs::write`], but atomic (best effort on Windows).
pub fn write(path: impl AsRef<Path>, contents: impl AsRef<[u8]>) -> Result<()> {
let path = path.as_ref();
let contents = contents.as_ref();
let dir = path.parent().unwrap();
// Create a tmpfile.
let (mut tmp_file, tmp_path) = tmpfile(dir)?;
let result = (|| {
// Write to the tmpfile.
_ = tmp_file.set_len(contents.len() as u64);
tmp_file
.write_all(contents)
.with_context(|| format!("could not write to file: {}", tmp_path.display()))?;
// Set the owner of the tmpfile (UNIX only).
#[cfg(unix)]
if let Ok(metadata) = path.metadata() {
use std::os::unix::fs::MetadataExt;
use nix::unistd::{self, Gid, Uid};
let uid = Uid::from_raw(metadata.uid());
let gid = Gid::from_raw(metadata.gid());
_ = unistd::fchown(&tmp_file, Some(uid), Some(gid));
}
// Close and rename the tmpfile.
// In some cases, errors from the last write() are reported only on close().
// Rust ignores errors from close(), since it occurs inside `Drop`. To
// catch these errors, we manually call `File::sync_all()` first.
tmp_file
.sync_all()
.with_context(|| format!("could not sync writes to file: {}", tmp_path.display()))?;
mem::drop(tmp_file);
rename(&tmp_path, path)
})();
// In case of an error, delete the tmpfile.
if result.is_err() {
_ = fs::remove_file(&tmp_path);
}
result
}
/// Atomically create a tmpfile in the given directory.
fn tmpfile(dir: impl AsRef<Path>) -> Result<(File, PathBuf)> {
const MAX_ATTEMPTS: usize = 5;
const TMP_NAME_LEN: usize = 16;
let dir = dir.as_ref();
let mut attempts = 0;
loop {
attempts += 1;
// Generate a random name for the tmpfile.
let mut name = String::with_capacity(TMP_NAME_LEN);
name.push_str("tmp_");
while name.len() < TMP_NAME_LEN {
name.push(fastrand::alphanumeric());
}
let path = dir.join(name);
// Atomically create the tmpfile.
match OpenOptions::new().write(true).create_new(true).open(&path) {
Ok(file) => break Ok((file, path)),
Err(e) if e.kind() == io::ErrorKind::AlreadyExists && attempts < MAX_ATTEMPTS => {}
Err(e) => {
break Err(e).with_context(|| format!("could not create file: {}", path.display()));
}
}
}
}
/// Similar to [`fs::rename`], but with retries on Windows.
fn rename(from: impl AsRef<Path>, to: impl AsRef<Path>) -> Result<()> {
let from = from.as_ref();
let to = to.as_ref();
const MAX_ATTEMPTS: usize = if cfg!(windows) { 5 } else { 1 };
let mut attempts = 0;
loop {
match fs::rename(from, to) {
Err(e) if e.kind() == io::ErrorKind::PermissionDenied && attempts < MAX_ATTEMPTS => {
attempts += 1
}
result => {
break result.with_context(|| {
format!("could not rename file: {} -> {}", from.display(), to.display())
});
}
}
}
}
pub fn canonicalize(path: impl AsRef<Path>) -> Result<PathBuf> { pub fn canonicalize(path: impl AsRef<Path>) -> Result<PathBuf> {
dunce::canonicalize(&path) dunce::canonicalize(&path)
.with_context(|| format!("could not resolve path: {}", path.as_ref().display())) .with_context(|| format!("could not resolve path: {}", path.as_ref().display()))