Add typo-tolerant fuzzy matching
This commit is contained in:
parent
11d7dfa28d
commit
97f9d8a61d
|
|
@ -86,7 +86,7 @@ impl<'a> Stream<'a> {
|
||||||
let path = util::to_lowercase(path);
|
let path = util::to_lowercase(path);
|
||||||
let mut path = path.as_str();
|
let mut path = path.as_str();
|
||||||
|
|
||||||
let (idx, end) = match rfind_component_match(path, keywords_last) {
|
let (idx, _end) = match rfind_component_match(path, keywords_last) {
|
||||||
Some((idx, end)) => {
|
Some((idx, end)) => {
|
||||||
if path[end..].contains(path::is_separator) {
|
if path[end..].contains(path::is_separator) {
|
||||||
return false;
|
return false;
|
||||||
|
|
@ -181,22 +181,26 @@ fn rfind_component_match(path: &str, keyword: &str) -> Option<(usize, usize)> {
|
||||||
return None;
|
return None;
|
||||||
}
|
}
|
||||||
|
|
||||||
// Favor exact substring matches; fall back to fuzzy subsequence within a
|
if keyword.contains(path::is_separator) {
|
||||||
// single component. Keywords that contain a path separator are matched
|
return path.rfind(keyword).map(|idx| (idx, idx + keyword.len()));
|
||||||
// literally to preserve existing slash semantics.
|
}
|
||||||
if !keyword.contains(path::is_separator) {
|
|
||||||
if let Some(idx) = path.rfind(keyword) {
|
if let Some(idx) = path.rfind(keyword) {
|
||||||
return Some((idx, idx + keyword.len()));
|
|
||||||
}
|
|
||||||
} else if let Some(idx) = path.rfind(keyword) {
|
|
||||||
return Some((idx, idx + keyword.len()));
|
return Some((idx, idx + keyword.len()));
|
||||||
}
|
}
|
||||||
|
|
||||||
// Fuzzy: rightmost component where keyword is a subsequence.
|
let keyword_len = keyword.chars().count();
|
||||||
|
|
||||||
|
// Fuzzy: rightmost component where keyword is a subsequence, or edit
|
||||||
|
// distance 1 (equal-length typo) within a single component.
|
||||||
for (component_start, component) in rsplit_components_with_indices(path) {
|
for (component_start, component) in rsplit_components_with_indices(path) {
|
||||||
if let Some((start, end)) = subsequence_bounds(component, keyword) {
|
if let Some((start, end)) = subsequence_bounds(component, keyword) {
|
||||||
return Some((component_start + start, component_start + end));
|
return Some((component_start + start, component_start + end));
|
||||||
}
|
}
|
||||||
|
|
||||||
|
if keyword_len == component.chars().count() && edit_distance_leq1(component, keyword) {
|
||||||
|
return Some((component_start, component_start + component.len()));
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
None
|
None
|
||||||
|
|
@ -245,6 +249,71 @@ fn subsequence_bounds(haystack: &str, needle: &str) -> Option<(usize, usize)> {
|
||||||
None
|
None
|
||||||
}
|
}
|
||||||
|
|
||||||
|
fn edit_distance_leq1(a: &str, b: &str) -> bool {
|
||||||
|
if a == b {
|
||||||
|
return true;
|
||||||
|
}
|
||||||
|
|
||||||
|
let a_chars: Vec<_> = a.chars().collect();
|
||||||
|
let b_chars: Vec<_> = b.chars().collect();
|
||||||
|
let a_len = a_chars.len();
|
||||||
|
let b_len = b_chars.len();
|
||||||
|
|
||||||
|
if (a_len as isize - b_len as isize).abs() > 1 {
|
||||||
|
return false;
|
||||||
|
}
|
||||||
|
|
||||||
|
if a_len == b_len {
|
||||||
|
// Single substitution?
|
||||||
|
let mut diffs = 0;
|
||||||
|
for (ac, bc) in a_chars.iter().zip(b_chars.iter()) {
|
||||||
|
if ac != bc {
|
||||||
|
diffs += 1;
|
||||||
|
if diffs > 1 {
|
||||||
|
break;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
if diffs == 1 {
|
||||||
|
return true;
|
||||||
|
}
|
||||||
|
|
||||||
|
// Single adjacent transposition?
|
||||||
|
for i in 0..a_len - 1 {
|
||||||
|
if a_chars[i] != b_chars[i] {
|
||||||
|
return i + 1 < a_len
|
||||||
|
&& a_chars[i] == b_chars[i + 1]
|
||||||
|
&& a_chars[i + 1] == b_chars[i]
|
||||||
|
&& a_chars[i + 2..] == b_chars[i + 2..]
|
||||||
|
&& a_chars[..i] == b_chars[..i];
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
return false;
|
||||||
|
}
|
||||||
|
|
||||||
|
// Lengths differ by exactly 1: check single insertion/deletion.
|
||||||
|
let (short, long) = if a_len < b_len { (&a_chars, &b_chars) } else { (&b_chars, &a_chars) };
|
||||||
|
let mut i = 0;
|
||||||
|
let mut j = 0;
|
||||||
|
let mut edits = 0;
|
||||||
|
|
||||||
|
while i < short.len() && j < long.len() {
|
||||||
|
if short[i] == long[j] {
|
||||||
|
i += 1;
|
||||||
|
j += 1;
|
||||||
|
} else {
|
||||||
|
edits += 1;
|
||||||
|
if edits > 1 {
|
||||||
|
return false;
|
||||||
|
}
|
||||||
|
j += 1; // skip one char in longer string
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
true
|
||||||
|
}
|
||||||
|
|
||||||
#[cfg(test)]
|
#[cfg(test)]
|
||||||
mod tests {
|
mod tests {
|
||||||
use std::path::PathBuf;
|
use std::path::PathBuf;
|
||||||
|
|
@ -276,6 +345,8 @@ mod tests {
|
||||||
// Fuzzy subsequence within component
|
// Fuzzy subsequence within component
|
||||||
#[case(&["docs"], "/home/Documents", true)]
|
#[case(&["docs"], "/home/Documents", true)]
|
||||||
#[case(&["dcmts"], "/home/Documents", true)]
|
#[case(&["dcmts"], "/home/Documents", true)]
|
||||||
|
// Typo tolerance (edit distance 1)
|
||||||
|
#[case(&["doucments"], "/home/Documents", 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 db = &mut Database::new(PathBuf::new(), Vec::new(), |_| Vec::new(), false);
|
||||||
let options = StreamOptions::new(0).with_keywords(keywords.iter());
|
let options = StreamOptions::new(0).with_keywords(keywords.iter());
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue