image_compressor 0.3.0

image_compressor: ^0.3.0 copied to clipboard

Three additions, all verified on-device (iOS + Android):

• ImageCompressor.probe(input) — read dimensions / byte size / format without decoding pixels. EXIF-oriented dims, consistent across platforms.
• SizePreset + toPreset — named target sizes (thumbnail / avatar / web / hd) for the common cases. Size-first, not quality presets.
• keepMetadata: on toSize / toQuality (+ the batch variants) — preserve the source EXIF/GPS/camera metadata (Android + iOS, JPEG). Default false (stripping is the privacy-safe default). The orientation tag is always reset (rotation is baked into pixels). Web can't preserve metadata and ignores the flag; on Android the copied tags may push the output a few KB over maxBytes.
• Troubleshooting section in the README.

Docs only.

• Honest, sharper positioning. Target-size isn't unique anymore, so the README no longer claims "the thing no other package does". The real, defensible moat is the combination: the one package that does target-size and runs on web and ships zero native binaries.
• Added an FAQ and a chart showing why a target size beats guessing a quality number.

Breaking: toSizeAll / toQualityAll now return List<BatchResult> (previously List<CompressedImage>).

• Batch is no longer all-or-nothing. One unreadable image used to throw and discard every other result. Each input now gets its own BatchResult — BatchSuccess(source, image) or BatchFailure(source, error) — in input order, so a single bad file can't sink the batch. Only a CancelToken cancellation still throws (it aborts the whole operation on purpose).
  • Migration: results → results.whereType<BatchSuccess>().map((r) => r.image) for just the compressed images; check whereType<BatchFailure>() for the rest.
• Simpler example: pick a photo and compress it to 500 KB — the 30-second story, instead of the old benchmark harness.
• Docs: inline dart examples on toSize, toQuality, and toSizeAll in the API reference; documented the remaining CompressedImage fields.
• Docs: flutter pub add in place of a pinned version in the README.
• Kept IDE files (*.iml, .idea/) out of the published archive.
• Added a ROADMAP.md (release cadence).

• Retracted an unfair performance claim. 0.1.1's README said we return a small file "instead of an uncontrolled multi-megabyte one" — that comparison had us downsampling a 27 MP photo while the other package was forced to full resolution. Given the same dimension cap it gets there too; the real advantage is that toSize finds the quality instead of making you guess it.
• Corrected the speed claim. The benchmark that reported us ~20% slower was unwarmed and always timed us first. Warmed, in both run orders, it's a dead heat (103 ms vs 102 ms on 6.75 MP) — and our number includes the EXIF orientation pass the other side skips. See BENCHMARK.md.
• Added the GitHub Sponsors funding link (missed the 0.1.1 cut).

• Docs: reworded the README's comparison to a name-free "us vs most other packages" table (no competitor names).
• Example: added a real compress → save-to-gallery flow (via gal).
• Added GitHub Sponsors funding link (pubspec funding:, README, FUNDING.yml).

Initial release.

• ImageCompressor.toSize(input, maxBytes:) — compress to a target file size. The quality search runs natively (decode once), so it stays fast on large images. Returns reachedTarget + usedQuality instead of throwing when a size can't be met.
• ImageCompressor.toQuality(input, quality:) — fixed-quality mode.
• toSizeAll / toQualityAll — batch with bounded concurrency, onProgress, and CancelToken support.
• Any input via ImageSource — bytes, file path, asset key, or an XFile straight from image_picker.
• CompressedImage.saveTo(path) extension (native only).
• Automatic EXIF orientation (autoOrient), decode-time downsampling to avoid OOM on large images, and consistent per-edge resize across platforms.
• Formats: JPEG/PNG everywhere, WebP on Android + web, HEIC on iOS.
• Platforms: Android, iOS, Web.
• Never returns null — hard failures throw a typed CompressError.