Compare commits

..
4 changed files with 13 additions and 23 deletions
+2 -3
View File
@@ -278,7 +278,7 @@ fn collect_inputs(inputs: &[String]) -> Discovery {
}
/// Returns false only when the global file budget stops traversal.
fn walk(dir: &Path, depth: usize, out: &mut Discovery, seen: &mut std::collections::HashSet<u64>) -> bool {
fn walk(dir: &Path, depth: usize, out: &mut Discovery, seen: &mut std::collections::HashSet<(u64, u64)>) -> bool {
// A directory is visited once by identity (dev+inode on unix): several links to the same
// directory are aliases, not exponential branches of the scan. Other platforms keep the
// depth cap only.
@@ -286,8 +286,7 @@ fn walk(dir: &Path, depth: usize, out: &mut Discovery, seen: &mut std::collectio
{
use std::os::unix::fs::MetadataExt as _;
if let Ok(m) = std::fs::metadata(dir)
&& let Some(id) = m.dev().checked_mul(1 << 60).and_then(|d| d.checked_add(m.ino()))
&& !seen.insert(id)
&& !seen.insert((m.dev(), m.ino()))
{
return true;
}
-2
View File
@@ -308,8 +308,6 @@ mod tests {
assert_eq!(ev("REGEXREPLACE", vec![t("2026-10-22"), t("(\\d+)-(\\d+)-(\\d+)"), t("$3/$2/$1")]), Value::text("22/10/2026"));
assert_eq!(ev("REGEXREPLACE", vec![t("abc"), t("x*"), t("-")]), Value::text("-a-b-c-"));
assert_eq!(ev("REGEXREPLACE", vec![t("a1b2"), t("[0-9]"), t("#"), n(2.0)]), Value::text("a1b#"));
assert_eq!(ev("REGEXREPLACE", vec![t("a1b2"), t("[0-9]"), t("#"), n(-2.0)]), Value::text("a#b2"));
assert_eq!(ev("REGEXREPLACE", vec![t("a1b2"), t("[0-9]"), t("#"), n(-3.0)]), Value::text("a1b2"));
// One match whose replacement repeats it 10,000 times: refused before it is built.
assert_eq!(ev("REGEXREPLACE", vec![t(&many), t("^.*$"), t(&"$0".repeat(10_000))]), Value::Error(CellError::Value));
assert_eq!(ev("REGEXREPLACE", vec![t(&many), t("^.*$"), t(&"$0".repeat(10_000)), n(1.0)]), Value::Error(CellError::Value));
+4 -15
View File
@@ -883,24 +883,13 @@ fn regexreplace(a: &[Arg], _c: &mut dyn Ctx) -> R<Value> {
out.push_str(s.get(last..).unwrap_or_default());
return check_len(out);
}
let caps: Vec<regex::Captures> = re.captures_iter(&s).collect();
let k = occ.unsigned_abs() as usize;
// Keeps only the wanted match's captures: the nth from the front, or a window of the last `k`.
let cap = if occ > 0 {
re.captures_iter(&s).nth(k - 1)
} else {
let mut last = std::collections::VecDeque::new();
for cap in re.captures_iter(&s) {
if last.len() == k {
last.pop_front();
}
last.push_back(cap);
}
if last.len() == k { last.pop_front() } else { None }
};
let Some(cap) = cap else { return text_val(s) };
let idx = if occ > 0 { k.checked_sub(1) } else { caps.len().checked_sub(k) };
let Some(cap) = idx.and_then(|i| caps.get(i)) else { return text_val(s) };
let Some(m) = cap.get(0) else { return text_val(s) };
out.push_str(s.get(..m.start()).unwrap_or_default());
expand_capped(&cap, &rep, &mut out, limit).ok_or(CellError::Value)?;
expand_capped(cap, &rep, &mut out, limit).ok_or(CellError::Value)?;
out.push_str(s.get(m.end()..).unwrap_or_default());
check_len(out)
}
+7 -3
View File
@@ -13,7 +13,10 @@ struct Counts {
proptests: usize,
}
fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<u64>, budget: &mut usize) {
/// Larger files (or devices behind a symlink) are not source files and are not read.
const MAX_SOURCE_BYTES: u64 = 16 << 20;
fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<(u64, u64)>, budget: &mut usize) {
if *budget == 0 {
// Hard stop for pathological trees: two directory symlinks pointing at each other
// would otherwise be walked exponentially.
@@ -22,9 +25,9 @@ fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<u64>, b
// Directory identities (dev+inode on unix) make a symlink to a visited directory a
// no-op, not another exponential branch of the scan.
#[cfg(unix)]
let id = std::fs::metadata(dir).ok().and_then(|m| m.dev().checked_mul(1 << 60).and_then(|d| d.checked_add(m.ino())));
let id = std::fs::metadata(dir).ok().map(|m| (m.dev(), m.ino()));
#[cfg(not(unix))]
let id: Option<u64> = None;
let id: Option<(u64, u64)> = None;
if let Some(id) = id
&& !seen.insert(id)
{
@@ -45,6 +48,7 @@ fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<u64>, b
walk(&p, c, seen, budget);
}
} else if name.ends_with(".rs")
&& p.metadata().is_ok_and(|m| m.is_file() && m.len() <= MAX_SOURCE_BYTES)
&& let Ok(s) = std::fs::read_to_string(&p)
{
c.files += 1;