From 505fef5dfcb057a73d047371dc36161f0041e4e1 Mon Sep 17 00:00:00 2001
From: Claudio <claudio@hubris.network>
Date: Sun, 17 May 2026 23:25:43 +0200
Subject: [PATCH] fix(folders): root shows whole library + sidecar counts
 recurse
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Two paired fixes for the folder tree on the timeline:

* +page.svelte: drop the `Path === ''` post-filter for the root entry.
  PhotoPrism's indexer always nests photos under YYYY/MM, so "photos
  whose Path is empty" is always the empty set in practice — the root
  entry looked broken instead of "whole library". Treat `/` as the
  unscoped view and rely on subfolder selections (now wildcarded via
  filters.svelte.ts) for narrowing.

* sidecar/handlers_folders.go: the per-folder count fan-out used
  `q=path:<x>`, the same exact-match operator that just got fixed in
  the web filter. Result: every year-level folder reported count=0
  in the sidebar. Switch to `q=path:"<x>*"` so the count reflects
  the whole subtree (dedupe by UID still in place).
---
 sidecar/handlers_folders.go | 20 ++++++++++++++------
 web/src/routes/+page.svelte | 12 +++++++-----
 2 files changed, 21 insertions(+), 11 deletions(-)

diff --git a/sidecar/handlers_folders.go b/sidecar/handlers_folders.go
index cd68660..d5a7e09 100644
--- a/sidecar/handlers_folders.go
+++ b/sidecar/handlers_folders.go
@@ -231,12 +231,20 @@ func handleFolderCounts(pp *ppClient) gin.HandlerFunc {
 			go func() {
 				defer wg.Done()
 				defer func() { <-sem }()
-				// `path:` is non-recursive in PhotoPrism's q-DSL: matches
-				// direct children only. `merged=false` returns one row per
-				// File on disk, so HEIC + companion JPG count twice unless
-				// we dedupe by UID — which is what the old client-side
-				// code did, and what we keep doing here.
-				q := url.QueryEscape("path:" + path)
+				// `path:<x>` is an exact match in PhotoPrism's q-DSL —
+				// it matches only photos whose `photo_path` field equals
+				// <x>, not descendants. The indexer always nests photos
+				// under YYYY/MM, so an internal tree node like `2024` has
+				// zero direct children and reports a count of 0 unless we
+				// recurse. `path:<x>*` is the documented wildcard form and
+				// matches both `<x>` itself (no harm if empty) and every
+				// `<x>/...` descendant.
+				//
+				// `merged=false` still returns one row per File on disk,
+				// so HEIC + companion JPG count twice unless we dedupe by
+				// UID — which is what the old client-side code did, and
+				// what we keep doing here.
+				q := url.QueryEscape(`path:"` + path + `*"`)
 				resp, err := pp.call(c.Request.Context(), http.MethodGet,
 					"/api/v1/photos?count=1000&offset=0&merged=false&q="+q, token, nil)
 				if err != nil || !resp.OK {
diff --git a/web/src/routes/+page.svelte b/web/src/routes/+page.svelte
index 2431659..a50abc5 100644
--- a/web/src/routes/+page.svelte
+++ b/web/src/routes/+page.svelte
@@ -177,12 +177,14 @@
 	// scoped server-side via the q-DSL and must not be re-filtered here,
 	// or labels / search will silently drop subfolder photos when the
 	// store hasn't fully hydrated from the URL yet.
+	//
+	// Root (`/`) historically also restricted to `Path === ''` — i.e.
+	// photos sitting directly at the originals root with no subfolder.
+	// PhotoPrism's indexer however always nests photos under YYYY/MM, so
+	// that view was always empty in practice. Treat root as "the whole
+	// library" instead; subfolders still scope normally.
 	function applyFolderScope(list: PpPhoto[], f: typeof filters): PpPhoto[] {
-		if (f.folderPath !== '/') return list;
-		if (f.section !== 'all-photos') return list;
-		if (f.heapUid) return list;
-		if (f.search) return list;
-		return list.filter((p) => !p.Path);
+		return list;
 	}
 	const pageCount = $derived(photosQuery.data?.pages.length ?? 0);