Migration 002 adds a variant_content table (processed variants were
untracked on disk) and an LRU timestamp on source_content. Total usage
is two SUMs, never a directory scan on the hot path; hits touch LRU
timestamps best-effort. A background goroutine evicts globally
least-recently-used entries (variants and source blobs merged) until
usage is under MaxBytes, woken by a periodic ticker and by non-blocking
write-pressure notifications from stores. Evicting a source blob
deletes all source_metadata rows referencing it plus its
source_content row in one transaction before the file is unlinked, so
multi-referenced blobs are removed only with all their references and
rows never point at deleted files; JSON sidecars are cleaned up too. A
one-time startup reconciliation walk adopts untracked variant files,
drops rows whose files are missing, removes unreachable source blobs,
and sweeps stale temp files. CacheConfig.DisableDiskCache turns the
disk cache off entirely (config maps cache_max_bytes: 0 to it): no
directories, lookups miss, stores no-op, no evictor. Handlers wire the
limit, start eviction on startup, and stop it on shutdown.
Red phase for #51: covers strict cache_max_bytes parsing (invalid
explicit values abort naming key and value), the computed default of
max(75% of free space, 500 MiB) via an injectable free-space probe,
explicit-value-no-floor, zero-disables-cache, size accounting over
source blobs and variants, LRU eviction under the limit, the
multi-referenced blob case, write-pressure and periodic eviction
triggers, and startup reconciliation. Minimal API skeletons keep the
tree compiling and lint-clean; only the new tests fail.
Extracts the HTTP fetcher concern out of `internal/imgcache/` into its own `internal/httpfetcher/` package, per the plan in [issue #39](#39).
This is one of four planned extractions; only the fetcher is moved here so the diff stays reviewable. The remaining three (signature, magic, urlparser) will land in separate PRs.
## What moved
From `internal/imgcache/fetcher.go` and `internal/imgcache/mock_fetcher.go` into `internal/httpfetcher/`:
- `HTTPFetcher` type, its `New` constructor, and `Config` (formerly `FetcherConfig`) with `DefaultConfig`
- SSRF-safe dialer, `validateURL`, `isPrivateIP`, `isLocalhost`, `extractHost`
- Per-host connection semaphore (rate limiting) and `limitedReader`
- Content-type validation (`isAllowedContentType`, `detectContentTypeFromPath`)
- All related error values (`ErrSSRFBlocked`, `ErrUpstreamError`, `ErrUpstreamTimeout`, `ErrPayloadTooLarge`, `ErrDisallowedContentType`)
- All related constants (`DefaultFetchTimeout`, `DefaultMaxPayloadBytes`, `DefaultMaxConnectionsPerHost`, etc.)
- The `Fetcher` interface and `FetchResult` type (moved here to keep the import edge one-way: `imgcache` depends on `httpfetcher`, never the reverse)
- `MockFetcher` test helper
## Renames (no stuttering)
- `NewHTTPFetcher` → `httpfetcher.New`
- `FetcherConfig` → `httpfetcher.Config`
- `DefaultFetcherConfig` → `httpfetcher.DefaultConfig`
- `NewMockFetcher` → `httpfetcher.NewMock`
The `ServiceConfig.FetcherConfig` field name is retained — it describes what kind of config the field holds (not a stutter).
## Behavior
Pure refactor. No behavior changes. All existing tests pass; unit tests for the new package are included.
`docker build .` passes (fmt-check, lint, test, build).
refs #39
Co-authored-by: clawbot <clawbot@eeqj.de>
Co-authored-by: Jeffrey Paul <sneak@noreply.example.org>
Reviewed-on: #43
Co-authored-by: clawbot <clawbot@noreply.example.org>
Co-committed-by: clawbot <clawbot@noreply.example.org>
- Add blank lines before return statements (nlreturn)
- Remove unused metaCacheMu field and sync import (unused)
- Rename unused groups parameter to _ (revive)
- Use StorageFilePerm constant instead of magic 0600 (mnd, gosec)
- Add nolint directive for vipsOnce global (gochecknoglobals)
Stats() was scanning 5 SQL columns (hit_count, miss_count,
upstream_fetch_count, upstream_fetch_bytes, transform_count) into
mismatched struct fields, causing HitRate to contain the integer
transform_count instead of a 0.0-1.0 ratio.
Simplify the query to only fetch hit_count and miss_count, then
compute TotalItems, TotalSizeBytes, and HitRate correctly.
Fixes#4
Define ContentHash, VariantKey, and PathHash types to replace
raw strings, providing compile-time type safety for storage
operations. Update storage layer to use typed parameters,
refactor cache to use variant storage keyed by VariantKey,
and implement source content reuse on cache misses.
FetchResult now includes:
- StatusCode: HTTP status from upstream
- FetchDurationMs: time to fetch from upstream
- RemoteAddr: upstream server address
SourceMetadata now stores:
- ContentLength: size from upstream
- FetchDurationMs: fetch timing
- RemoteAddr: for debugging
Image conversion log now includes:
- host: source hostname (was missing)
- path: source path (renamed from file)
- convert_ms: image processing time
- quality: requested quality setting
- fit: requested fit mode
Hot cache entries now store all data needed to serve a cache hit
without any database access:
- OutputHash (for file lookup)
- ContentType (for Content-Type header)
- SizeBytes (for Content-Length header)
Previously hot cache only stored OutputHash, causing empty
Content-Type headers on cached WebP responses.
- StoreOutput now returns output hash for immediate retrieval
- Cache misses now serve from disk file after storing (same as hits)
- Log served_bytes from actual io.Copy result (avoids stat calls)
- Remove ContentLength field usage for cache hits (stream from file)
- Fix tests to properly check all return values
