Documentation
¶
Overview ¶
Package attachcache stores attachment bytes on local disk.
Attachment bytes were proxied from Jira on every view, which contradicts the premise of the tool: everything else answers from local disk, and an issue with three screenshots re-downloaded them on every open. Bytes for a given attachment id are immutable in Jira, so they cache indefinitely and the only bound needed is total size.
A cached attachment also survives having no credential at all, which is what lets the bundled demo snapshot show real images offline.
Index ¶
- Constants
- func Key(site, profile, issue, id string) string
- func TooLarge(err error) bool
- type Cache
- func (c *Cache) Dir() string
- func (c *Cache) Fill(id string, fetch func() (io.ReadCloser, Meta, error)) error
- func (c *Cache) Get(id string) (io.ReadSeekCloser, Meta, error)
- func (c *Cache) Has(id string) bool
- func (c *Cache) ImportFile(key, path, contentType, filename string) error
- func (c *Cache) ImportManifest(dir, site, profile string, issueForID func(id string) (issue string, ok bool)) (ImportStats, error)
- func (c *Cache) MissReason(key, id string) string
- func (c *Cache) Stats() (files int, bytes int64)
- type ImportStats
- type Meta
Constants ¶
const DefaultMaxBytes int64 = 512 << 20
DefaultMaxBytes bounds the cache. Attachments are usually screenshots, so a few hundred megabytes holds a working set of thousands.
Variables ¶
This section is empty.
Functions ¶
func Key ¶ added in v0.13.0
Key is the on-disk identity for an attachment. Site (and issue) are mixed in so a profile that changes sites cannot serve the previous site's bytes, and a request for the wrong issue cannot read a cached id.
An empty site keeps the legacy id-only form so snapshot imports for `gadak demo` / export-static (no site) stay reachable. Existing id-only files on a real site miss after this change (the safe invalidation).
Types ¶
type Cache ¶
type Cache struct {
// contains filtered or unexported fields
}
Cache is a content-addressed store under one directory. Safe for concurrent use; a miss for the same id collapses into a single fetch.
func (*Cache) Fill ¶
Fill is the single-flight write path: fetch runs at most once per id even if ten renders miss at the same moment. It returns after the bytes are on disk.
fetch must return the body, its content type, and its length (0 if unknown).
func (*Cache) Get ¶
Get returns an open reader for a cached attachment. The caller closes it. A miss returns os.ErrNotExist.
func (*Cache) ImportFile ¶
ImportFile seeds an entry from a local file under the given cache key. Callers that know site/profile/issue must pass Key(...); do not pass a raw id when the site is set — that is the snapshot-import miss that D9 closed.
func (*Cache) ImportManifest ¶ added in v0.13.0
func (c *Cache) ImportManifest(dir, site, profile string, issueForID func(id string) (issue string, ok bool)) (ImportStats, error)
ImportManifest seeds the cache from a fixture directory's manifest.json. Each file is stored under Key(site, profile, issue, id). issueForID must return the issue key that owns id; ids it rejects are skipped — they are never written under the raw id (that would reopen the D9 cross-site hole).
func (*Cache) MissReason ¶ added in v0.13.0
MissReason explains a cache miss so a log can tell a key-scope mismatch (legacy id-only file still on disk) from a true absence. It never serves the legacy entry.
type ImportStats ¶ added in v0.13.0
ImportStats is what a snapshot import reports: how many files landed, and which manifest ids were skipped (not in the mirror).
Source Files