word-tally

Tallies the number of times each word appears in one or more unicode input sources. Use word-tally as a command-line tool or WordTally via the Rust library interface.

Four I/O strategies are available:

• stream: Sequential single-threaded streaming with minimal memory usage
• parallel-stream (default): Parallel streaming with balanced performance and memory usage
• parallel-in-memory: Load entire input into memory for parallel processing
• parallel-mmap: Memory-mapped I/O for best performance with large files

All parallel modes use SIMD-accelerated chunk boundary detection. Memory mapping requires seekable file descriptors and won't work with stdin or pipes.

Usage

Usage: word-tally [OPTIONS] [PATHS]...

Arguments:
  [PATHS]...  File paths to use as input (use "-" for stdin) [default: -]

Options:
  -I, --io <STRATEGY>          I/O strategy [default: parallel-stream] [possible values: parallel-stream, stream, parallel-mmap, parallel-in-memory]
  -e, --encoding <ENCODING>    Word boundary detection encoding [default: unicode] [possible values: unicode, ascii]
  -c, --case <FORMAT>          Case normalization [default: original] [possible values: original, upper, lower]
  -s, --sort <ORDER>           Sort order [default: desc] [possible values: desc, asc, unsorted]
  -m, --min-chars <COUNT>      Exclude words containing fewer than min chars
  -n, --min-count <COUNT>      Exclude words appearing fewer than min times
  -w, --exclude-words <WORDS>  Exclude words from a comma-delimited list
  -i, --include <PATTERNS>     Include only words matching a regex pattern
  -x, --exclude <PATTERNS>     Exclude words matching a regex pattern
  -f, --format <FORMAT>        Output format [default: text] [possible values: text, json, csv]
  -d, --delimiter <VALUE>      Delimiter between keys and values [default: " "]
  -o, --output <PATH>          Write output to file rather than stdout
  -v, --verbose                Print verbose details
  -h, --help                   Print help (see more with '--help')
  -V, --version                Print version

Installation

cargo install word-tally

Examples

I/O strategies

Choose an I/O strategy based on your performance and memory requirements:

# Default: Parallel streaming - balanced performance and memory
echo "tally me" | word-tally
word-tally file.txt

# Sequential streaming - minimize memory usage
word-tally --io=stream large-file.txt

# Parallel in-memory - fastest for small inputs and stdin
word-tally --io=parallel-in-memory document.txt

# Parallel memory-mapped - fastest for large files
word-tally --io=parallel-mmap large-file.txt

Additional features:

# Process multiple files
word-tally file1.txt file2.txt file3.txt

# Mix stdin and files
cat header.txt | word-tally - body.txt footer.txt

Note: Memory mapping (parallel-mmap) requires seekable files and cannot be used with stdin or pipes.

Output formats

Text (default)

# Write to file instead of stdout
word-tally README.md --output=tally.txt

# Custom delimiter between word and count
word-tally README.md --delimiter=": " --output=tally.txt

# Pipe to other tools
word-tally README.md | head -n10

CSV

# Using a comma delimiter (unescaped without headers)
word-tally --delimiter="," --output="tally.csv" README.md

# Using proper CSV format (escaped with headers)
word-tally --format=csv --output="tally.csv" README.md

JSON

word-tally --format=json --output="tally.json" README.md

Visualization

Convert JSON output for visualization with d3-cloud:

word-tally --format=json README.md | jq 'map({text: .[0], value: .[1]})' > d3-cloud.json

Format and pipe the JSON output to the wordcloud_cli to produce an image:

word-tally --format=json README.md | jq -r 'map(.[0] + " ") | join(" ")' | wordcloud_cli --imagefile wordcloud.png

Case normalization

# Convert to lowercase
word-tally --case=lower file.txt

# Preserve original case
word-tally --case=original file.txt

# Convert all to uppercase
word-tally --case=upper file.txt

Sorting options

# Sort by frequency (descending, default)
word-tally --sort=desc file.txt

# Sort alphabetically (ascending)
word-tally --sort=asc file.txt

# No sorting (sorted by order seen)
word-tally --sort=unsorted file.txt

Filtering words

# Only include words that appear at least 10 times
word-tally --min-count=10 file.txt

# Exclude words with fewer than 5 characters
word-tally --min-chars=5 file.txt

# Exclude words by pattern
word-tally --exclude="^a.*" --exclude="^the$" file.txt

# Combining include and exclude patterns
word-tally --include="^w.*" --include=".*o$" --exclude="^who$" file.txt

# Exclude specific words
word-tally --exclude-words="the,a,an,and,or,but" file.txt

Verbose output

echo "fe fi fi fo fo fo" | word-tally --verbose
#>> source -
#>> total-words 6
#>> unique-words 3
#>> delimiter " "
#>> case original
#>> order desc
#>> processing parallel
#>> io streamed
#>> min-chars none
#>> min-count none
#>> exclude-words none
#>> exclude-patterns none
#>>
#>> fo 3
#>> fi 2
#>> fe 1

Environment variables

The following environment variables configure various aspects of the library:

I/O and processing strategy configuration:

• WORD_TALLY_IO - I/O strategy (default: parallel-stream, options: stream, parallel-stream, parallel-in-memory, parallel-mmap)

Memory allocation and performance:

• WORD_TALLY_UNIQUENESS_RATIO - Ratio of total words to unique words for capacity estimation. Higher values allocate less initial memory. Books tend to have a 10:1 ratio, but a more conservative 256:1 is used as default to reduce unnecessary memory overhead (default: 256)
• WORD_TALLY_WORDS_PER_KB - Estimated words per KB of text for capacity calculation (default: 128, max: 512)
• WORD_TALLY_STDIN_BUFFER_SIZE - Buffer size for stdin when size cannot be determined (default: 262144)
• WORD_TALLY_DEFAULT_CAPACITY - Default initial capacity when there is no size hint (default: 1024)
• WORD_TALLY_WORD_DENSITY - Multiplier for estimating unique words per chunk (default: 15)

Parallel processing configuration:

• WORD_TALLY_THREADS - Number of threads for parallel processing (default: all available cores)
• WORD_TALLY_CHUNK_SIZE - Size of chunks for parallel processing in bytes (default: 65536)

Exit codes

word-tally uses standard unix exit codes to indicate success or the types of failure:

• 0: Success
• 1: General error
• 64: Command line usage error
• 65: Data format error
• 66: Cannot open input
• 73: Cannot create output file
• 74: I/O error
• 77: Permission denied

Library usage

[dependencies]
word-tally = "0.26.0"

use std::fs::File;
use word_tally::{Io, Options, Processing, WordTally};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let file = File::open("document.txt")?;
    // Parallel streamed
    let word_tally = WordTally::new(&file, &Options::default())?;
    assert!(word_tally.count() > 0);

    // Parallel memory mapped
    let another_file = File::open("another-document.txt")?;
    let options = Options::default()
        .with_io(Io::ParallelMmap);
    let mmap_tally = WordTally::new(another_file, &options)?;
    assert!(mmap_tally.count() > 0);

    println!("Words: {} total, {} unique", mmap_tally.count(), mmap_tally.uniq_count());

    for (word, count) in mmap_tally.tally().iter().take(5) {
        println!("{}: {}", word, count);
    }

    Ok(())
}

The library supports case normalization, sorting, filtering and I/O and processing strategies.

Stability notice

Pre-release level stability: This is prerelease software. Expect breaking interface changes at MINOR version (0.x.0) bumps until a stable release.

Tests & benchmarks

Tests

Clone the repository.

git clone https://github.com/havenwood/word-tally
cd word-tally

Run all tests.

cargo test

Run specific test modules.

cargo test --test api_tests
cargo test --test filters_tests
cargo test --test io_tests

Run individual tests

cargo test --test filters_tests -- test_min_chars
cargo test --test io_tests -- test_memory_mapped

Benchmarks

Run all benchmarks.

cargo bench

Run specific benchmark groups

cargo bench --bench core
cargo bench --bench io
cargo bench --bench features

Run specific individual benchmarks

cargo bench --bench features -- case_sensitivity
cargo bench --bench core -- parallel_vs_sequential

Documentation

https://docs.rs/word-tally