rfasta

rfasta is a production-ready toolkit for parsing, cleaning, writing, and sharding protein FASTA files.

It provides:

• A high-performance Rust library
• A Python package for scripting and notebooks
• A CLI for batch and pipeline workflows

Highlights

• FASTA parsing from files or in-memory readers
• Configurable cleanup policies for duplicate records and invalid residues
• Deterministic shard generation for parallel processing
• Python bindings via PyO3
• CLI designed for operational workflows
• Documentation for Rust, Python, and CLI usage

Installation

Python (recommended for end users)

pip install rfasta

Rust (library development)

Add to Cargo.toml:

[dependencies]
rfasta = "0.1"

Quick Start

CLI

Clean a FASTA file:

rfasta clean proteins.fasta \
  --non-unique-header \
  --duplicate-record remove \
  --invalid-sequence convert-remove \
  -o cleaned.fasta

Split a large FASTA file into one-pass round-robin shards:

rfasta split proteins.fasta --output-dir shards --chunks 8

Python

import rfasta

rows = rfasta.read_fasta("proteins.fasta", expect_unique_header=True, verbose=False)
rfasta.write_fasta(rows, "proteins.copy.fasta", line_length=60)

Rust

use std::io::Cursor;

use rfasta::clean::{clean_sequences, CleanOptions, DuplicateAction, InvalidSequenceAction};
use rfasta::parse::{parse_fasta_reader, ParseOptions};
use rfasta::write::{write_fasta_writer, WriteOptions};

let input = b">seq1\nacdx\n>seq2\nTTTT\n";
let records = parse_fasta_reader(Cursor::new(input), ParseOptions::default())?;
let cleaned = clean_sequences(
    records,
    &CleanOptions {
        invalid_sequence_action: InvalidSequenceAction::ConvertRemove,
        duplicate_record_action: DuplicateAction::Fail,
        ..CleanOptions::default()
    },
)?;
let mut output = Vec::new();
write_fasta_writer(&mut output, &cleaned, &WriteOptions::default())?;
# Ok::<(), rfasta::RfastaError>(())

Documentation and Guides

• User guides: docs/
• API reference entry point: docs/api-reference.md
• Release and PyPI publishing runbook: RELEASING.md

Build the guide site locally:

pip install -r docs/requirements.txt
mkdocs build --strict

Local Performance Check

For a repo-local benchmark driver that does not require extra benchmark crates, run:

cargo run --release --example benchmark_driver