Skip to content

CLAUDE.md

Latest commit

 

History

History
54 lines (40 loc) · 2.34 KB

CLAUDE.md

File metadata and controls

54 lines (40 loc) · 2.34 KB
Raw

CLAUDE.md

Pure Go image processing library (github.com/anthonynsimon/bild) with a built-in CLI. Minimal dependencies — just cobra for CLI and golang.org/x/image for BMP support.

Build & Test

make build    # Build binary to dist/
make test     # Run all tests (60s timeout, verbose)
make fmt      # Format code (go fmt)
make race     # Run tests with race detector
make bench    # Run benchmarks

CI runs make build && make test across Go 1.21–1.26 on Ubuntu and macOS. Arch differences can cause issues in floating point values, but it's generally acceptable as long as it's effectively imperceptible to the human eye.

Project Structure

Each image processing domain is its own importable package:

  • adjust/ — brightness, contrast, gamma, hue, saturation
  • blend/ — 16 blending modes (multiply, overlay, screen, etc.)
  • blur/ — box and gaussian blur
  • channel/ — RGBA channel extraction
  • clone/ — image cloning and padding
  • convolution/ — kernel convolution operations
  • effect/ — grayscale, sepia, sharpen, sobel, invert, median, erode, dilate, emboss, edge detection
  • histogram/ — histogram analysis and visualization
  • imgio/ — image I/O (PNG, JPEG, BMP)
  • noise/ — noise generation (uniform, binary, gaussian, perlin)
  • paint/ — flood fill
  • segment/ — thresholding
  • transform/ — rotate, resize (7 resampling filters), crop, flip, translate, shear, zoom

Supporting packages: parallel/ (goroutine work splitting), fcolor/ (float64 RGBA), math/ (clamp, min/max), util/ (color conversions, comparison helpers), perlin/.

CLI

  • main.go → cmd.Execute()
  • cmd/root.go — registers all subcommands via init()
  • cmd/helpers.go — apply() single-image pipeline, apply2() two-image pipeline, parsers, encoder resolution
  • cmd/<category>.go — one file per command group, mirrors the library packages

Patterns

  • All processing functions take image.Image and return *image.RGBA
  • Pixel-level parallelism via parallel.Line() which splits rows across CPUs
  • Tests are table-driven with manual pixel data; comparison via util.RGBAImageEqual() / util.RGBAImageApproxEqual()

When Adding Features

  • New library functions go in the appropriate domain package
  • New CLI commands follow the pattern in existing cmd/ files
  • Update CHANGELOG.md and README.md