scorphus / thumbor / 31427369955

89%

Build Type

push

github

Committed by scorphus

Commit Message

feat: separate result storage entries by accepted image formats

The new thumbor.auto_image_format module computes a cache-key
discriminator from the configured AUTO_* options and the formats the
request accepts. The file result storage names its namespace with it,
so two clients accepting different formats never share a cached result.

An AUTO_WEBP-only configuration keeps the legacy auto_webp and default
namespaces, and legacy entries are still migrated there. Configurations
with AUTO_AVIF, AUTO_JPG, AUTO_HEIF, AUTO_PNG or AUTO_PNG_TO_JPG get an
isolated namespace and no migration, because their legacy default
namespace can hold a different format for the same URL. Subclasses that
override is_auto_webp keep controlling the legacy namespace choice.

Result storage is skipped entirely, reads and writes, when one of those
formats is active and no trustworthy cache key can be computed before
the engine exists: when a handler overrides accepts_mime_type, and when
the request object lacks the parsed accepts_* attributes. Serving
without the cache is slower but never serves a wrong format.

Co-authored-by: Pablo Santiago Blum de Aguiar <scorphus@gmail.com>

Coverage Stats

63 of 64 new or added lines in 3 files covered. (98.44%)

57 existing lines in 2 files now uncovered.

4099 of 4597 relevant lines covered (89.17%)

4.46 hits per line