pzsh

Performance-first shell framework with sub-10ms startup. Like oh-my-zsh, but 50-200x faster.

Core Invariant

No shell startup shall exceed 10ms. This is not a goal—it is a hard constraint enforced at compile time, test time, and runtime.

Performance

$ pzsh bench
Startup Benchmark (100 iterations)
────────────────────────────────
min:       0.030ms
max:       0.051ms
mean:      0.032ms
p99:       0.051ms
────────────────────────────────
Budget: 10ms ✓ (p99 < 10ms)

Installation

cargo install pzsh

Usage

# Initialize configuration
pzsh init --shell zsh

# Benchmark startup time
pzsh bench

# Lint for slow patterns
pzsh lint ~/.pzshrc

# Profile startup breakdown
pzsh profile

# Compile configuration
pzsh compile

Configuration

# ~/.pzshrc
[pzsh]
version = "0.1.0"
shell = "zsh"

[performance]
startup_budget_ms = 10
lazy_load = true

[aliases]
ll = "ls -la"
gs = "git status"

[env]
EDITOR = "vim"
GOROOT = "/usr/local/opt/go/libexec"  # Pre-resolved, no $(brew ...)

Forbidden Patterns

pzsh enforces O(1) startup by rejecting slow patterns:

# FORBIDDEN: subprocess calls at startup
export GOROOT="$(brew --prefix golang)/libexec"  # 50-100ms per call

# ALLOWED: pre-resolved paths
export GOROOT="/usr/local/opt/go/libexec"  # 0ms

# FORBIDDEN: oh-my-zsh, NVM, conda init
source $ZSH/oh-my-zsh.sh  # 500-2000ms

Architecture

• Parser: O(1) with LRU caching, 2ms budget
• Executor: O(1) hash lookups, 2ms budget
• Prompt: Async git status, 2ms budget
• Config: Pre-compiled TOML, no runtime parsing

Testing

cargo test      # 41 tests, all enforcing time budgets
cargo bench     # Criterion benchmarks

Built With

• bashrs - Rust↔Shell transpiler for safety
• aprender - ML-powered optimization

Toyota Way

Development follows the Toyota Production System:

• Stop the line on defects (Andon)
• Continuous improvement (Kaizen)
• Go and see for yourself (Genchi Genbutsu)

License

MIT