photoprism/internal/entity/search/conditions.go
Michael Mayer 0cecb5287f Search: Match folder album_path case-insensitively #5724
Folder search lowercases the query term, but album_path is VARBINARY and
compared byte-exact (case-sensitive) on MySQL/MariaDB, so a lowercased
term no longer matched uppercase folder paths. A parent folder still
appeared via its case-insensitive title, while child folders - matchable
only by path - dropped out of the results.

Add a dialect-aware PathLike helper: MySQL converts album_path to a
case-insensitive collation for the LIKE, SQLite and other dialects fall
back to a plain LIKE (SQLite is already ASCII case-insensitive). Only the
folder search filter is affected; byte-exact album_path identity lookups
are unchanged, so emoji and case-distinct folders stay distinct.
2026-07-16 16:53:28 +00:00

362 lines
9.7 KiB
Go

package search
import (
"fmt"
"strings"
"github.com/photoprism/photoprism/pkg/clean"
"github.com/photoprism/photoprism/pkg/dsn"
"github.com/photoprism/photoprism/pkg/txt"
"github.com/jinzhu/inflection"
)
// PathLike returns a case-insensitive "col LIKE ?" condition for a VARBINARY path column such as
// album_path or photo_path. form.Unserialize lowercases the search term, but these columns are
// compared byte-exact (case-sensitive) on MySQL, so an uppercase path would otherwise never match a
// lowercased query. MySQL therefore needs an explicit case-insensitive collation; SQLite LIKE is
// already ASCII case-insensitive, and any other or unknown dialect falls back to a plain LIKE (its
// default semantics, never an error). The dialect is the GORM dialect name, e.g. s.Dialect().GetName().
func PathLike(dialect, col string) string {
if dialect == dsn.DriverMySQL {
return "CONVERT(" + col + " USING utf8mb4) COLLATE utf8mb4_general_ci LIKE ?"
}
return col + " LIKE ?"
}
// SqlParam sanitizes user input for use as a LIKE-clause bind value. The
// surrounding pre/post strings are concatenated verbatim so callers can add
// SQL wildcards (e.g. "%") without exposing the underlying value to string
// interpolation.
func SqlParam(s, pre, post string) string {
return pre + strings.Trim(clean.SqlClean(s), " |&*%") + post
}
// LikeAny builds OR-chained LIKE predicates for a text column. The input string
// may contain AND / OR separators; keywords trigger stemming and plural
// normalization while exact mode disables wildcard suffixes.
// The returned wheres and values are aligned 1:1; callers feed each pair into
// gorm.Expr(wheres[i], values[i]...).
func LikeAny(col, s string, keywords, exact bool) (wheres []string, values [][]any) {
if s == "" {
return wheres, values
}
s = txt.StripOr(clean.SearchQuery(s))
var wildcardThreshold int
if exact {
wildcardThreshold = -1
} else if keywords {
wildcardThreshold = 4
} else {
wildcardThreshold = 2
}
for _, k := range txt.UnTrimmedSplitWithEscape(s, txt.AndRune, txt.EscapeRune) {
var orWheres []string
var orValues []any
var words []string
if keywords {
words = txt.UniqueKeywords(k)
} else {
words = txt.UniqueWords(strings.Fields(k))
}
if len(words) == 0 {
continue
}
for _, w := range words {
if wildcardThreshold > 0 && len(w) >= wildcardThreshold {
orWheres = append(orWheres, fmt.Sprintf("%s LIKE ?", col))
orValues = append(orValues, SqlParam(w, "", "%"))
} else {
orWheres = append(orWheres, fmt.Sprintf("%s LIKE ?", col))
orValues = append(orValues, SqlParam(w, "", ""))
}
if !keywords || !txt.ContainsASCIILetters(w) {
continue
}
singular := inflection.Singular(w)
if singular != w {
orWheres = append(orWheres, fmt.Sprintf("%s LIKE ?", col))
orValues = append(orValues, SqlParam(singular, "", ""))
}
}
if len(orWheres) > 0 {
wheres = append(wheres, strings.Join(orWheres, " OR "))
values = append(values, orValues)
}
}
return wheres, values
}
// LikeAnyKeyword is a keyword-optimized wrapper around LikeAny.
func LikeAnyKeyword(col, s string) (wheres []string, values [][]any) {
return LikeAny(col, s, true, false)
}
// LikeAnyWord matches whole words and keeps wildcard thresholds tuned for
// free-form text search instead of keyword lists.
func LikeAnyWord(col, s string) (wheres []string, values [][]any) {
return LikeAny(col, s, false, false)
}
// LikeAll produces AND-chained LIKE predicates for every significant token in
// the search string. When exact is false, longer words receive a suffix
// wildcard to support prefix matches.
// The returned wheres and values are aligned 1:1; callers feed each pair into
// gorm.Expr(wheres[i], values[i]...).
func LikeAll(col, s string, keywords, exact bool) (wheres []string, values [][]any) {
if s == "" {
return wheres, values
}
var words []string
var wildcardThreshold int
if keywords {
words = txt.UniqueKeywords(s)
wildcardThreshold = 4
} else {
words = txt.UniqueWords(strings.Fields(s))
wildcardThreshold = 2
}
if len(words) == 0 {
return wheres, values
} else if exact {
wildcardThreshold = -1
}
for _, w := range words {
if wildcardThreshold > 0 && len(w) >= wildcardThreshold {
wheres = append(wheres, fmt.Sprintf("%s LIKE ?", col))
values = append(values, []any{SqlParam(w, "", "%")})
} else {
wheres = append(wheres, fmt.Sprintf("%s LIKE ?", col))
values = append(values, []any{SqlParam(w, "", "")})
}
}
return wheres, values
}
// LikeAllKeywords is LikeAll specialized for keyword search.
func LikeAllKeywords(col, s string) (wheres []string, values [][]any) {
return LikeAll(col, s, true, false)
}
// LikeAllWords is LikeAll specialized for general word search.
func LikeAllWords(col, s string) (wheres []string, values [][]any) {
return LikeAll(col, s, false, false)
}
// LikeAllNames splits a name query into AND-separated groups and generates
// prefix or substring matches against each provided column, keeping multi-word
// tokens intact so "John Doe" still matches full-name columns.
func LikeAllNames(cols Cols, s string) (wheres []string, values [][]any) {
if len(cols) == 0 || len(s) < 1 {
return wheres, values
}
for _, k := range txt.UnTrimmedSplitWithEscape(s, txt.AndRune, txt.EscapeRune) {
var orWheres []string
var orValues []any
for _, w := range txt.UnTrimmedSplitWithEscape(k, txt.OrRune, txt.EscapeRune) {
w = strings.TrimSpace(w)
if w == txt.EmptyString {
continue
}
for _, c := range cols {
if strings.Contains(w, txt.Space) {
orWheres = append(orWheres, fmt.Sprintf("%s LIKE ?", c))
orValues = append(orValues, SqlParam(w, "", "%"))
} else {
orWheres = append(orWheres, fmt.Sprintf("%s LIKE ?", c))
orValues = append(orValues, SqlParam(w, "%", "%"))
}
}
}
if len(orWheres) > 0 {
wheres = append(wheres, strings.Join(orWheres, " OR "))
values = append(values, orValues)
}
}
return wheres, values
}
// AnySlug converts human-friendly search terms into slugs and matches them
// against the provided slug column, including the singularized variant for
// plural words (e.g. "Cats" -> "cat").
func AnySlug(col, search, sep string) (where string, values []any) {
if search == "" {
return "", values
}
if sep == "" {
sep = " "
}
var wheres []string
var words []string
for w := range strings.SplitSeq(search, sep) {
w = strings.TrimSpace(w)
words = append(words, txt.Slug(w))
if !txt.ContainsASCIILetters(w) {
continue
}
singular := inflection.Singular(w)
if singular != w {
words = append(words, txt.Slug(singular))
}
}
if len(words) == 0 {
return "", values
}
for _, w := range words {
wheres = append(wheres, fmt.Sprintf("%s = ?", col))
values = append(values, SqlParam(w, "", ""))
}
return strings.Join(wheres, " OR "), values
}
// AnyInt filters user-specified integers through an allowed range and returns
// an OR-chained equality predicate for the values that remain. Named low/high
// to avoid shadowing the predeclared min/max identifiers added in Go 1.21.
func AnyInt(col, numbers, sep string, low, high int) (where string, values []any) {
if numbers == "" {
return "", values
}
if sep == "" {
sep = txt.Or
}
var matches []int
var wheres []string
for n := range strings.SplitSeq(numbers, sep) {
i := txt.Int(n)
if i == 0 || i < low || i > high {
continue
}
matches = append(matches, i)
}
if len(matches) == 0 {
return "", values
}
for _, n := range matches {
wheres = append(wheres, fmt.Sprintf("%s = ?", col))
values = append(values, n)
}
return strings.Join(wheres, " OR "), values
}
// OrLike prepares a parameterized OR/LIKE clause for a single column. Star (* )
// wildcards are mapped to SQL percent wildcards before returning the query and
// bind values.
func OrLike(col, s string) (where string, values []any) {
if txt.Empty(col) || txt.Empty(s) {
return "", []any{}
}
s = strings.ReplaceAll(s, "*", "%")
s = strings.ReplaceAll(s, "%%", "%")
terms := txt.UnTrimmedSplitWithEscape(s, txt.OrRune, txt.EscapeRune)
values = make([]any, len(terms))
if l := len(terms); l == 0 {
return "", []any{}
} else if l == 1 {
values[0] = terms[0]
} else {
for i := range terms {
values[i] = strings.TrimSpace(terms[i])
}
}
like := fmt.Sprintf("%s LIKE ?", col)
where = like + strings.Repeat(" OR "+like, len(terms)-1)
return where, values
}
// OrLikeCols behaves like OrLike but fans out the same search terms across
// multiple columns, preserving the order of values so callers can feed them to
// database/sql.
func OrLikeCols(cols []string, s string) (where string, values []any) {
if len(cols) == 0 || txt.Empty(s) {
return "", []any{}
}
s = strings.ReplaceAll(s, "*", "%")
s = strings.ReplaceAll(s, "%%", "%")
terms := txt.UnTrimmedSplitWithEscape(s, txt.OrRune, txt.EscapeRune)
if len(terms) == 0 {
return "", []any{}
}
values = make([]any, len(terms)*len(cols))
for j := range terms {
for i := range cols {
values[j+i] = strings.TrimSpace(terms[j])
}
}
wheres := make([]string, len(cols))
for i, col := range cols {
for j := range terms {
k := len(terms) * i
values[j+k] = terms[j]
}
like := fmt.Sprintf("%s LIKE ?", col)
wheres[i] = like + strings.Repeat(" OR "+like, len(terms)-1)
}
return strings.Join(wheres, " OR "), values
}
// SplitOr splits a search string on OR separators (|) while respecting escape
// sequences so literals like "\|" survive unchanged.
func SplitOr(s string) (values []string) {
return txt.TrimmedSplitWithEscape(s, txt.OrRune, txt.EscapeRune)
}
// SplitAnd splits a search string on AND separators (&) while honoring escape
// sequences.
func SplitAnd(s string) (values []string) {
return txt.TrimmedSplitWithEscape(s, txt.AndRune, txt.EscapeRune)
}