perf(thumbs): the gallery shipped 77 MB to render 250px tiles

The operator found this in about a minute of using the live Desk: "images load
at full resolution instead of calculated thumbnails, which means they load VERY
slowly and are tiny."

MEASURED on the live set:

    sindra-corpus-v1   66 images   77.5 MB   1024x1024 each
    sindra-sfw-pool    59 images   71.7 MB
    sindra             30 images   61.6 MB   2.1 MB average
    sindra-bakeoff     40 images   57.2 MB

A tile renders around 250px, so the grid shipped roughly 16x the pixels that
reach the screen.

⚠ OUR PARKING RATIONALE WAS WRONG IN AN INSTRUCTIVE WAY. ROADMAP parked
progressive loading on "the largest gallery is 66 images; at that size a lazy
grid is almost certainly fine", and the parking-lot row said "270 <img
loading=lazy> may be fine". Both count IMAGES. Neither weighs BYTES. We measured
the dimension that was easy to measure rather than the one that determines the
experience, and 66 really is a fine count sitting on a terrible payload.

booth/thumbs.py caches WebP at 512px longest side inside the booth at
`.thumbs/<rel>.webp` — inside on purpose, so a cache can never outlive what it
describes. Pillow is an optional import: absent, every tile falls back to the
original, so the page is heavier and never broken. Generation is lazy, atomic
(temp + os.replace), rebuilt when the source is newer, and NEVER RAISES.

?thumb=1 rides the EXISTING file route rather than growing a new one, because
that route's traversal guard is already correct and a second route is a second
place to get it wrong.

ALSO FIXES A PRE-EXISTING LEAK THE CACHE WOULD HAVE WALKED INTO. booth_items and
zip_booth both tested `p.name.startswith(".")` — the FILE's name — so
`.thumbs/a.png` (name `a.png`) would have rendered as a gallery item and shipped
inside every zip. CLAUDE.md invariant 2 promises a dotfile costs nothing in item
counts, galleries or zips; that was true only at the top level. Both now skip
every dot-prefixed path COMPONENT.

AND THE FILMSTRIP, which is the same defect in a worse place: it shows EVERY
ring item at a few dozen pixels, so full-resolution frames there cost more than
the grid did. The stage is untouched and stays full size, because that is the
full-size review.

Item.thumb is derived in the resolver, not by a template reasoning about `kind`
(INV-1). build_gallery had to carry it too — a missing key there rendered as a
SILENT fallback to the full image, which is exactly where a new Item field gets
dropped with nothing failing.

754 green.
This commit is contained in:
vh
2026-09-23 10:47:34 -07:00
parent 447a9b67e9
commit d5e23c7d5f
8 changed files with 328 additions and 8 deletions
+25
View File
@@ -356,3 +356,28 @@ def test_a_huge_caption_sidecar_is_not_read_whole(tmp_path):
cap = {it.rel: it.caption for it in booth_items(b)}["a.png"]
assert cap is not None and len(cap) <= CAPTION_MAX
def test_a_dot_directory_hides_its_whole_subtree(tmp_path):
"""CLAUDE.md invariant 2 claims a dotfile costs nothing in item counts,
galleries or zips. That was only true at the TOP LEVEL: both `booth_items`
and `zip_booth` tested the FILE's name, so `.thumbs/a.png` has `p.name ==
"a.png"` and sailed through as a gallery item and a zip entry.
Pre-existing, found while adding a `.thumbs/` cache. Any path component
starting with a dot is the Booth's own namespace.
Defeating change: back to `p.name.startswith(".")`."""
import io
import zipfile
from booth.app import zip_booth
b = tmp_path / "b"
(b / ".thumbs").mkdir(parents=True)
(b / "real.png").write_bytes(b"\x89PNG")
(b / ".thumbs" / "real.png").write_bytes(b"\x89PNGthumb")
(b / ".marks.json").write_text("{}")
assert [i.rel for i in booth_items(b)] == ["real.png"]
assert zipfile.ZipFile(io.BytesIO(zip_booth(b))).namelist() == ["real.png"]
+163
View File
@@ -0,0 +1,163 @@
"""Thumbnails — the fix for a gallery that shipped 77 MB to render 250px tiles.
The operator found this in about a minute of using the live Desk. ROADMAP had
parked it on "the largest gallery is 66 images", which counted IMAGES and never
weighed BYTES; 66 is a fine count sitting on a terrible payload.
"""
from __future__ import annotations
import io
import pathlib
import sys
import pytest
from fastapi.testclient import TestClient
sys.path.insert(0, str(pathlib.Path(__file__).parent.parent))
from booth.app import create_app # noqa: E402
from booth.items import booth_items # noqa: E402
from booth.thumbs import THUMB_DIR, THUMB_MAX, ensure_thumb, wants_thumb # noqa: E402
PIL = pytest.importorskip("PIL.Image", reason="Pillow is not installed")
def _img(path: pathlib.Path, w: int, h: int, fmt="PNG"):
path.parent.mkdir(parents=True, exist_ok=True)
PIL.new("RGB", (w, h), (120, 30, 90)).save(path, fmt)
return path
def test_a_big_image_gets_a_much_smaller_thumbnail(tmp_path):
"""The whole point, asserted in BYTES rather than in existence — a thumbnail
that is not dramatically smaller has not fixed anything."""
b = tmp_path / "g"
src = _img(b / "big.png", 1024, 1024)
out = ensure_thumb(b, "big.png")
assert out is not None and out.is_file()
assert max(PIL.open(out).size) <= THUMB_MAX
assert out.stat().st_size * 4 < src.stat().st_size, (
f"thumb {out.stat().st_size}B vs source {src.stat().st_size}B — not worth the cache"
)
def test_an_already_small_image_gets_no_thumbnail(tmp_path):
"""Serving the original is correct when it is already tile-sized. A cache
entry that saves nothing is pure cost.
Defeating change: generating unconditionally."""
b = tmp_path / "g"
_img(b / "small.png", 200, 200)
assert ensure_thumb(b, "small.png") is None
def test_the_cache_lives_inside_the_booth_and_is_invisible(tmp_path):
"""`.thumbs/` is inside the booth so it is swept with it — a cache that
outlives what it describes is a leak. And it must not become gallery items
or zip entries: both skip every dot-prefixed path COMPONENT, which they did
not do until this module needed them to."""
import zipfile
from booth.app import zip_booth
b = tmp_path / "g"
_img(b / "big.png", 1024, 1024)
ensure_thumb(b, "big.png")
assert (b / THUMB_DIR).is_dir(), "the cache is not inside the booth"
assert [i.rel for i in booth_items(b)] == ["big.png"]
assert zipfile.ZipFile(io.BytesIO(zip_booth(b))).namelist() == ["big.png"]
def test_a_damaged_image_costs_its_own_tile_not_the_page(tmp_path):
"""NEVER RAISES. A thumbnail is an optimisation; a page that will not load
is worse than one that loads slowly.
Defeating change: letting the Pillow exception out."""
b = tmp_path / "g"
b.mkdir()
(b / "lies.png").write_bytes(b"\x89PNG\r\n\x1a\n" + b"not an image at all" * 20)
assert ensure_thumb(b, "lies.png") is None
def test_a_stale_thumbnail_is_rebuilt(tmp_path):
"""Editing a file in place must not leave the old thumbnail forever."""
import os
b = tmp_path / "g"
_img(b / "x.png", 1024, 1024)
first = ensure_thumb(b, "x.png")
before = first.stat().st_mtime_ns
_img(b / "x.png", 900, 900)
os.utime(b / "x.png", None)
again = ensure_thumb(b, "x.png")
assert again.stat().st_mtime_ns != before, "the stale thumbnail survived an edit"
def test_only_thumbable_types_are_candidates():
"""SVG is vector and Pillow cannot read it; a video is not an image."""
assert wants_thumb("a.png") and wants_thumb("A.JPG") and wants_thumb("a.webp")
assert not wants_thumb("a.svg") and not wants_thumb("a.webm") and not wants_thumb("a.txt")
def test_the_item_record_carries_the_thumb_url(tmp_path):
"""INV-1: the resolver decides whether an item has a thumbnail. No template
appends `?thumb=1` by reasoning about `kind` itself."""
b = tmp_path / "g"
_img(b / "big.png", 1024, 1024)
_img(b / "vec.svg", 10, 10) if False else (b / "vec.svg").write_text("<svg/>")
by = {i.rel: i for i in booth_items(b)}
assert by["big.png"].thumb == "big.png?thumb=1"
assert by["vec.svg"].thumb is None
def test_the_route_serves_the_thumbnail_and_the_original(tmp_path):
"""?thumb=1 rides the EXISTING file route, so it inherits that route's
traversal guard rather than growing a second one."""
b = tmp_path / "g"
src = _img(b / "big.png", 1024, 1024)
c = TestClient(create_app(tmp_path, ttl_hours=24, start_sweeper=False))
full = c.get("/b/g/big.png")
thumb = c.get("/b/g/big.png?thumb=1")
assert full.status_code == thumb.status_code == 200
assert len(thumb.content) * 4 < len(full.content), "the route served the full image"
assert len(full.content) == src.stat().st_size
def test_the_gallery_tile_requests_the_thumbnail(tmp_path):
"""The operator's actual complaint: the grid pulled full-resolution files."""
b = tmp_path / "g"
for n in ("a.png", "b.png"):
_img(b / n, 1024, 1024)
c = TestClient(create_app(tmp_path, ttl_hours=24, start_sweeper=False))
html = c.get("/b/g/").text
assert 'src="a.png?thumb=1"' in html, "the tile still asks for the full image"
def test_a_thumb_request_for_a_traversal_path_is_still_refused(tmp_path):
"""The guard is the file route's, and it must not be weakened by the new
query parameter."""
b = tmp_path / "g"
_img(b / "big.png", 1024, 1024)
(tmp_path / "secret.txt").write_text("nope")
c = TestClient(create_app(tmp_path, ttl_hours=24, start_sweeper=False))
assert c.get("/b/g/../secret.txt?thumb=1").status_code in (404, 400)
def test_the_filmstrip_and_tray_use_thumbnails_but_the_stage_does_not(tmp_path):
"""The same 77 MB in a different place. The filmstrip shows EVERY ring item
at a few dozen pixels, so full-resolution frames there are worse than the
grid was — while the stage is the full-size review and must stay full size.
Defeating change: `x.url` in the filmstrip, or `it.thumb` on the stage."""
b = tmp_path / "g"
for n in ("a.png", "b.png", "c.png"):
_img(b / n, 1024, 1024)
c = TestClient(create_app(tmp_path, ttl_hours=24, start_sweeper=False))
html = c.get("/b/g/view?f=a.png").text
assert 'id="vimg" src="a.png"' in html, "the stage must serve the full image"
assert 'src="b.png?thumb=1"' in html, "the filmstrip still pulls full images"
assert 'src="b.png"' not in html.replace('src="b.png?thumb=1"', ""), \
"a full-size frame survived in the strip"