230 lines
5.3 KiB
Go
230 lines
5.3 KiB
Go
package main
|
|
|
|
import (
|
|
"io/fs"
|
|
"log"
|
|
"net/http"
|
|
"path/filepath"
|
|
"sort"
|
|
"strings"
|
|
"unicode"
|
|
)
|
|
|
|
type searchResult struct {
|
|
Name string
|
|
URL string
|
|
Path string
|
|
Score int
|
|
}
|
|
|
|
type searchPageData struct {
|
|
Title string
|
|
Crumbs []crumb
|
|
EditMode bool
|
|
Query string
|
|
Results []searchResult
|
|
}
|
|
|
|
// handleSearch walks the wiki root and renders a search results page for the
|
|
// query in r.URL.Query().Get("q"). Only invoked when path is "/" and "q" is
|
|
// present.
|
|
func (h *handler) handleSearch(w http.ResponseWriter, r *http.Request) {
|
|
query := strings.TrimSpace(r.URL.Query().Get("q"))
|
|
results := searchWiki(h.root, query)
|
|
|
|
title := "Search"
|
|
if query != "" {
|
|
title = "Search: " + query
|
|
}
|
|
data := searchPageData{
|
|
Title: title,
|
|
Crumbs: []crumb{{Name: "search", URL: "/?q=" + query}},
|
|
Query: query,
|
|
Results: results,
|
|
}
|
|
w.Header().Set("Content-Type", "text/html; charset=utf-8")
|
|
if err := searchTmpl.ExecuteTemplate(w, "layout", data); err != nil {
|
|
log.Printf("search template error: %v", err)
|
|
}
|
|
}
|
|
|
|
// searchWiki walks root and scores each directory by how well the folder name
|
|
// matches the query. Page contents are not searched. Higher score = more
|
|
// relevant; exact matches rank first.
|
|
func searchWiki(root, query string) []searchResult {
|
|
if query == "" {
|
|
return nil
|
|
}
|
|
qLower := strings.ToLower(query)
|
|
qTokens := tokenize(qLower)
|
|
if len(qTokens) == 0 {
|
|
return nil
|
|
}
|
|
|
|
walkRoot := resolveWalkRoot(root)
|
|
var results []searchResult
|
|
_ = filepath.WalkDir(walkRoot, func(fsPath string, d fs.DirEntry, err error) error {
|
|
if err != nil {
|
|
return nil
|
|
}
|
|
if skip, walkErr := hiddenSkip(fsPath, walkRoot, d); skip {
|
|
return walkErr
|
|
}
|
|
if !d.IsDir() || fsPath == walkRoot {
|
|
return nil
|
|
}
|
|
name := d.Name()
|
|
score := scoreName(strings.ToLower(name), qLower, qTokens)
|
|
if score == 0 {
|
|
return nil
|
|
}
|
|
rel, relErr := filepath.Rel(walkRoot, fsPath)
|
|
if relErr != nil {
|
|
return nil
|
|
}
|
|
results = append(results, searchResult{
|
|
Name: name,
|
|
URL: "/" + filepath.ToSlash(rel) + "/",
|
|
Path: filepath.ToSlash(rel),
|
|
Score: score,
|
|
})
|
|
return nil
|
|
})
|
|
|
|
sort.SliceStable(results, func(i, j int) bool {
|
|
if results[i].Score != results[j].Score {
|
|
return results[i].Score > results[j].Score
|
|
}
|
|
di, dj := strings.Count(results[i].Path, "/"), strings.Count(results[j].Path, "/")
|
|
if di != dj {
|
|
return di < dj
|
|
}
|
|
return strings.ToLower(results[i].Name) < strings.ToLower(results[j].Name)
|
|
})
|
|
return results
|
|
}
|
|
|
|
// scoreName ranks how well nameLower matches the query. Whole-name exact
|
|
// match dominates; otherwise score is the sum of each token's best match
|
|
// against the words in the name. Position within the name does not matter —
|
|
// nesting depth is the tiebreaker, applied by the caller.
|
|
func scoreName(nameLower, qLower string, qTokens []string) int {
|
|
if nameLower == qLower {
|
|
return 1000
|
|
}
|
|
score := 0
|
|
nameWords := tokenize(nameLower)
|
|
for _, qt := range qTokens {
|
|
best := 0
|
|
for _, w := range nameWords {
|
|
switch {
|
|
case w == qt:
|
|
if best < 100 {
|
|
best = 100
|
|
}
|
|
case strings.HasPrefix(w, qt):
|
|
if best < 50 {
|
|
best = 50
|
|
}
|
|
case strings.Contains(w, qt):
|
|
if best < 20 {
|
|
best = 20
|
|
}
|
|
case levenshtein(w, qt) <= 2:
|
|
if best < 5 {
|
|
best = 5
|
|
}
|
|
}
|
|
}
|
|
score += best
|
|
}
|
|
return score
|
|
}
|
|
|
|
// resolveWalkRoot resolves symlinks so WalkDir descends into the real tree
|
|
// even when the configured wiki root is itself a symlink (as on the NAS).
|
|
func resolveWalkRoot(root string) string {
|
|
if r, err := filepath.EvalSymlinks(root); err == nil {
|
|
return r
|
|
}
|
|
return root
|
|
}
|
|
|
|
// hiddenSkip handles dotfile/dot-dir entries during a WalkDir. It returns
|
|
// (skipped, walkErr): skipped=true means the caller should `return walkErr`
|
|
// to either prune the subtree (hidden dir) or move past the entry (hidden
|
|
// file). When skipped=false the entry should be processed normally.
|
|
func hiddenSkip(fsPath, walkRoot string, d fs.DirEntry) (bool, error) {
|
|
if !strings.HasPrefix(d.Name(), ".") {
|
|
return false, nil
|
|
}
|
|
if d.IsDir() && fsPath != walkRoot {
|
|
return true, filepath.SkipDir
|
|
}
|
|
return true, nil
|
|
}
|
|
|
|
// tokenize splits s into lowercase word tokens, breaking on any rune that is
|
|
// not a letter or digit. Unicode-aware so umlauts etc. survive intact.
|
|
func tokenize(s string) []string {
|
|
var tokens []string
|
|
var b strings.Builder
|
|
for _, r := range s {
|
|
if unicode.IsLetter(r) || unicode.IsDigit(r) {
|
|
b.WriteRune(unicode.ToLower(r))
|
|
continue
|
|
}
|
|
if b.Len() > 0 {
|
|
tokens = append(tokens, b.String())
|
|
b.Reset()
|
|
}
|
|
}
|
|
if b.Len() > 0 {
|
|
tokens = append(tokens, b.String())
|
|
}
|
|
return tokens
|
|
}
|
|
|
|
// levenshtein returns the edit distance between a and b. Operates on runes so
|
|
// multi-byte characters count as one edit.
|
|
func levenshtein(a, b string) int {
|
|
ar, br := []rune(a), []rune(b)
|
|
if len(ar) == 0 {
|
|
return len(br)
|
|
}
|
|
if len(br) == 0 {
|
|
return len(ar)
|
|
}
|
|
prev := make([]int, len(br)+1)
|
|
curr := make([]int, len(br)+1)
|
|
for j := range prev {
|
|
prev[j] = j
|
|
}
|
|
for i := 1; i <= len(ar); i++ {
|
|
curr[0] = i
|
|
for j := 1; j <= len(br); j++ {
|
|
cost := 1
|
|
if ar[i-1] == br[j-1] {
|
|
cost = 0
|
|
}
|
|
del := prev[j] + 1
|
|
ins := curr[j-1] + 1
|
|
sub := prev[j-1] + cost
|
|
curr[j] = min3(del, ins, sub)
|
|
}
|
|
prev, curr = curr, prev
|
|
}
|
|
return prev[len(br)]
|
|
}
|
|
|
|
func min3(a, b, c int) int {
|
|
m := a
|
|
if b < m {
|
|
m = b
|
|
}
|
|
if c < m {
|
|
m = c
|
|
}
|
|
return m
|
|
}
|