HARific - High-Performance HAR File Toolkit

Quick Start

Installation

git clone https://github.com/pb33f/harific.git
cd harific
make build

Usage

# View HAR file in interactive TUI
./bin/harific recording.har

# Show all commands
./bin/harific --help

Background

Driven by frustration with diagnosing customer problems from browser experiences, HARific was built to solve a real problem: being unable to see what the customer saw, in the way they saw it. Diagnosing performance problems or rendering issues without proper tools is really hard. HARific provides visual exploration of gigantic HAR files in the terminal, with plans for a replay server that will replay every response back to the browser, complete with breakpoints to pause the conversation anywhere.

Search Engine Architecture

flowchart TB
    User[Search Query] --> Compile[Compile Pattern Once]
    Compile --> Partition[Partition Work<br/>10K entries → 100 batches]

    Partition --> Queue[Work Queue Channel]
    Queue --> W1[Worker 1]
    Queue --> W2[Worker 2]
    Queue --> W8[Worker 8]

    W1 --> Phase1[Phase 1: Metadata<br/>URL, Method, Status<br/>NO DISK I/O]
    W2 --> Phase1
    W8 --> Phase1

    Phase1 --> |No Match| Phase2[Phase 2: Load Entry<br/>Read from Disk]
    Phase1 --> |Match| Result[Return Result]

    Phase2 --> Search[Search Headers,<br/>Body, Cookies]
    Search --> Result

    Result --> Channel[Results Channel]
    Channel --> User

    subgraph "Resource Pools"
        BP[Buffer Pool<br/>64KB × 8]
        FP[File Handle Pool<br/>8 handles]
    end

    W1 -.-> BP
    W2 -.-> BP
    W8 -.-> BP
    W1 -.-> FP
    W2 -.-> FP
    W8 -.-> FP

    style Phase1 fill:#e1ffe1
    style Phase2 fill:#ffe1e1
    style BP fill:#e1f5ff
    style FP fill:#e1f5ff

Loading

Performance

File Size	Entries	Time	Throughput	Time/Entry
700MB	9,720	1.90s	367.76 MB/s	195.84 μs
1GB	14,126	2.77s	369.87 MB/s	196.00 μs
2GB	28,262	5.57s	367.93 MB/s	196.96 μs
5GB	70,689	13.62s	375.80 MB/s	192.73 μs

• Consistent Performance: ~370 MB/s regardless of file size
• Linear Scaling: Processing time scales perfectly with file size
• Predictable: ~195 microseconds per entry consistently
• Memory Efficient: ~57MB for 1.3GB file