Julia Package Development
Use this skill when working with Julia packages to ensure proper development workflows, testing patterns, documentation standards, and performance best practices.
Development Workflow
Environment Management
Start Julia with project environment
julia --project=.
Activate project in REPL
using Pkg Pkg.activate(".")
Install dependencies
Pkg.instantiate()
Update dependencies (PREFERRED over direct Project.toml editing)
Pkg.update()
Add new dependency
Pkg.add("PackageName")
Add development dependency
Pkg.add("PackageName"; io=devnull) # Then manually move to extras/test deps
Check package status
Pkg.status()
IMPORTANT: Always use Pkg.update() to update packages. Never edit Project.toml directly for version updates.
Testing
Run all tests (from project root)
julia --project=. -e 'using Pkg; Pkg.test()'
Run tests with test environment
julia --project=test test/runtests.jl
Run tests skipping quality checks (if supported)
julia --project=test test/runtests.jl skip_quality
Test Organization:
-
Use TestItemRunner with @testitem syntax for modular testing
-
Organize tests by component: test/component/ , test/package/
-
Package-level tests in test/package/ for quality (Aqua, DocTest, formatting)
-
Use @testitem "description" begin ... end for individual test items
Example test structure:
using TestItemRunner
@testitem "Basic functionality" begin using MyPackage @test my_function(1) == 2 end
@testitem "Edge cases" begin using MyPackage @test_throws ArgumentError my_function(-1) end
Documentation
Build documentation locally
julia --project=docs docs/make.jl
Build docs skipping notebooks (faster)
julia --project=docs docs/make.jl --skip-notebooks
or via environment variable
SKIP_NOTEBOOKS=true julia --project=docs docs/make.jl
Start Pluto server for interactive notebooks
(check project-specific task or command)
Documentation Structure:
-
Use Documenter.jl for documentation
-
Auto-deployment to GitHub Pages via CI
-
Structure defined in docs/pages.jl or docs/make.jl
Code Quality
Run pre-commit hooks
pre-commit run --all-files
JuliaFormatter (typically configured in .JuliaFormatter.toml)
Usually handled by pre-commit hooks
Quality Checks:
-
Aqua.jl tests for package quality
-
JuliaFormatter.jl for code formatting
-
Pre-commit hooks for automated checks
Code Style Guidelines
SciML Coding Standards
Follow SciML (Scientific Machine Learning) coding standards:
-
Avoid type instability
-
Ensure efficient precompilation
-
Use appropriate type annotations for performance
-
Write type-stable code
Type Stability:
Good - type stable
function compute(x::Float64) result = 0.0 # Type is known for i in 1:10 result += x * i end return result end
Avoid - type unstable
function compute_bad(x) result = 0 # Type might change for i in 1:10 result = result + x * i # Type may vary end return result end
Formatting Rules
-
Max 80 characters per line
-
No trailing whitespace
-
No spurious blank lines
-
Use JuliaFormatter.jl for consistent formatting
Documentation Standards
Docstring Syntax
Use @doc with either raw strings or regular strings:
For simple docstrings without LaTeX or templates
@doc " Brief description of the function.
Arguments
x: Description of xy: Description of y
Returns
- Description of return value
Examples
julia> my_function(1, 2)
3
"
function my_function(x, y)
return x + y
end
For docstrings with LaTeX math
@doc raw"
Computes the mathematical function:
f(x) = \int_0^x t^2 dt
Use raw strings when including LaTeX to preserve backslashes.
"
function math_function(x)
# implementation
end
### DocStringExtensions Templates
**IMPORTANT**: Template expansion rules:
- Use `@doc "` (regular string) for templates (allows expansion)
- Use `@doc """` with escaped backslashes when combining templates with LaTeX
- **NEVER** use `@doc raw"` with templates (prevents expansion)
```julia
using DocStringExtensions
# Good - template will expand
@doc "
$(TYPEDSIGNATURES)
Brief description.
# Fields
$(TYPEDFIELDS)
"
struct MyType
"Field description"
field::Int
end
# Good - template + LaTeX with escaped backslashes
@doc """
\$(TYPEDSIGNATURES)
Computes: ``f(x) = \\int_0^x t^2 dt``
Note the escaped backslashes in LaTeX: \\int, not \int
"""
function combined_function(x)
# implementation
end
# Avoid - raw string prevents template expansion
@doc raw"
$(TYPEDSIGNATURES) # This will NOT expand!
"
Documentation Structure Best Practices
- Keep interface method docstrings concise (1-2 lines)
- Use "See also" sections for cross-references
- Avoid duplication between related functions (pdf/logpdf, cdf/logcdf)
- Include mathematical formulations in main type/constructor docstrings
- Provide minimal but sufficient examples using @example
blocks
Cross-referencing:
@doc "
Compute the cumulative distribution function.
See also: [`logcdf`](@ref)
"
function cdf(d::MyDist, x::Real)
# implementation
end
@doc "
Compute the log cumulative distribution function.
See also: [`cdf`](@ref)
"
function logcdf(d::MyDist, x::Real)
# implementation
end
Package Structure
Typical Julia package structure:
MyPackage.jl/
├── src/
│ ├── MyPackage.jl # Main module file with exports
│ ├── component1.jl # Component implementations
│ ├── component2.jl
│ ├── docstrings.jl # DocStringExtensions templates
│ └── utils/
├── test/
│ ├── runtests.jl # Main test file
│ ├── component1/ # Tests by component
│ ├── component2/
│ └── package/ # Quality tests (Aqua, formatting)
├── docs/
│ ├── make.jl # Documentation build script
│ ├── src/ # Documentation source
│ └── pages.jl # Page structure (optional)
├── Project.toml # Package dependencies
└── README.md
Performance Best Practices
Type Stability
# Check type stability with @code_warntype
@code_warntype my_function(args...)
# Look for red (Any) types - indicates type instability
Precompilation
# Ensure efficient precompilation
# Use PrecompileTools.jl for complex packages
using PrecompileTools
@compile_workload begin
# Representative workload for precompilation
my_function(example_args...)
end
Performance Patterns
- Use in-place operations when possible (!
suffix convention)
- Preallocate arrays for loops
- Use @simd
, @inbounds
when safe
- Consider StaticArrays.jl
for small fixed-size arrays
- Profile with @time
, @benchmark
(BenchmarkTools.jl)
Common Dependencies and Patterns
Distributions.jl Interface
When implementing distributions:
- Implement required methods: pdf
, logpdf
, cdf
, logcdf
, quantile
, rand
- Implement support methods: minimum
, maximum
, insupport
- Optionally implement: mean
, var
, std
(if analytically tractable)
- Vectorization handled automatically via broadcasting
- Consider specialized batch methods: pdf!
, logpdf!
, cdf!
Turing.jl and AD Compatibility
Ensure compatibility with automatic differentiation:
- ForwardDiff.jl
- ReverseDiff.jl
- Zygote.jl
- Enzyme.jl
Avoid non-differentiable operations in AD-sensitive code.
Common Julia Ecosystem Tools
- Pkg: Package management
- TestItemRunner: Modern testing framework
- Documenter.jl: Documentation generation
- DocStringExtensions: Documentation templates
- JuliaFormatter.jl: Code formatting
- Aqua.jl: Package quality testing
- BenchmarkTools.jl: Performance benchmarking
- PrecompileTools.jl: Precompilation optimization
When to Use This Skill
Activate this skill when:
- Developing Julia packages
- Writing Julia tests
- Documenting Julia functions
- Setting up Julia package infrastructure
- Working with Distributions.jl, Turing.jl, or SciML packages
- Optimising Julia code performance
This skill provides Julia-specific development patterns.
Project-specific architecture and domain knowledge should remain in project CLAUDE.md files.