Add trait Model

.gitignore

@@ -1 +1,2 @@
 /target
+Cargo.lock

Cargo.toml

@@ -4,3 +4,4 @@ version = "0.1.0"
 edition = "2021"
 
 [dependencies]
+num = "0.4.3"

examples/example.rs

@@ -0,0 +1,46 @@
+use sac::model::Model;
+use sac::modelA::ModelA;
+
+const DATA: &[u8] = b"
+I'd just like to interject for a moment. What you're refering to as Linux, is in fact, GNU/Linux, or as I've re
+aken to calling it, GNU plus Linux. Linux is not an operating system unto itself, but rather another free compo
+a fully functioning GNU system made useful by the GNU corelibs, shell utilities and vital system components com
+a full OS as defined by POSIX.
+
+Many computer users run a modified version of the GNU system every day, without realizing it. Through a peculia
+f events, the version of GNU which is widely used today is often called Linux, and many of its users are not aw
+ it is basically the GNU system, developed by the GNU Project.
+
+There really is a Linux, and these people are using it, but it is just a part of the system they use. Linux is 
+el: the program in the system that allocates the machine's resources to the other programs that you run. The ke
+an essential part of an operating system, but useless by itself; it can only function in the context of a compl
+ating system. Linux is normally used in combination with the GNU operating system: the whole system is basicall
+th Linux added, or GNU/Linux. All the so-called Linux distributions are really distributions of GNU/Linux!
+";
+
+fn main() {
+    type CodeValue = u32;
+    println!(
+        "Using model: ModelA<{}>",
+        std::any::type_name::<CodeValue>()
+    );
+    let model: ModelA<CodeValue> = ModelA::default();
+    model.print_metrics();
+    println!("");
+
+    let mut compressed = Vec::new();
+    println!("compressing...");
+    model.compress(&DATA[..], &mut compressed).unwrap();
+    println!("ModelA compressed to {} bytes", compressed.len());
+    println!(
+        "Compression Ratio: {}",
+        DATA.len() as f64 / compressed.len() as f64
+    );
+    println!("");
+
+    println!("decompressing...");
+    let mut decompressed = Vec::new();
+    let model: ModelA<CodeValue> = ModelA::default();
+    model.decompress(&compressed, &mut decompressed).unwrap();
+    println!("{}", String::from_utf8_lossy(&decompressed));
+}

src/bit_buffer.rs

@@ -0,0 +1,181 @@
+use std::io::{self, Bytes, Cursor, Read, Write};
+
+pub struct BitWriter<'a, W: ?Sized + Write> {
+    bits: u8,
+    nextbit: usize,
+    output: &'a mut W,
+}
+impl<'a, W: Write> From<&'a mut W> for BitWriter<'a, W> {
+    fn from(value: &'a mut W) -> Self {
+        BitWriter::new(value)
+    }
+}
+impl<'a, W: Write> BitWriter<'a, W> {
+    pub fn new(writer: &'a mut W) -> Self {
+        return BitWriter {
+            bits: 0,
+            nextbit: 7,
+            output: writer,
+        };
+        //writer.into()
+    }
+    pub fn write(&mut self, bit: bool) -> io::Result<()> {
+        if bit {
+            self.bits = 1 << self.nextbit | self.bits;
+        }
+        if self.nextbit == 0 {
+            self.output.write(&[self.bits])?;
+            self.bits = 0;
+            self.nextbit = 7;
+        } else {
+            self.nextbit -= 1;
+        }
+        Ok(())
+    }
+    pub fn flush(self) -> std::io::Result<()> {
+        if self.bits != 0 {
+            self.output.write(&[self.bits])?;
+        }
+        return self.output.flush();
+    }
+}
+
+pub struct BitReader<T> {
+    next: u8,
+    bits: u8,
+    repeat_bits: i32,
+    input: Bytes<T>,
+}
+
+impl<R: Read> From<R> for BitReader<R> {
+    fn from(value: R) -> Self {
+        BitReader::new(value.bytes())
+    }
+}
+impl<T: AsRef<[u8]>> From<T> for BitReader<Cursor<T>> {
+    fn from(value: T) -> Self {
+        let c = Cursor::new(value);
+        c.into()
+    }
+}
+
+impl<R: Read> BitReader<R> {
+    pub fn new(value: Bytes<R>) -> Self {
+        BitReader {
+            next: 0,
+            bits: 0,
+            repeat_bits: 0,
+            input: value,
+        }
+    }
+    pub fn get_bit(&mut self) -> io::Result<bool> {
+        if self.next == 0 {
+            let next = self.input.next().transpose()?;
+            if let Some(byte) = next {
+                self.bits = byte;
+            } else if self.repeat_bits <= 0 {
+                return Err(io::Error::other("No more bits"));
+            } else {
+                self.repeat_bits -= 8;
+            }
+            self.next = 1 << 7;
+        }
+        let bit = (self.bits & self.next) > 0;
+        self.next = self.next >> 1;
+        return Ok(bit);
+    }
+}
+impl<T> BitReader<T> {
+    pub fn with_repeat_bits(mut self, n_bits: u16) -> Self {
+        self.repeat_bits = n_bits as i32;
+        self
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::model::Metrics;
+    use crate::modelA::tests::COMPRESSED_BYTES;
+
+    struct InputBits<'a> {
+        input: &'a [u8],
+        current_byte: u32,
+        last_mask: u32,
+        code_value_bits: i32,
+    }
+
+    pub trait Poppable {
+        fn pop(&mut self) -> Option<u8>;
+    }
+    impl Poppable for &[u8] {
+        fn pop(&mut self) -> Option<u8> {
+            if self.len() == 0 {
+                return None;
+            }
+            let out = self[0];
+            *self = &self[1..];
+            return Some(out);
+        }
+    }
+    impl<'a> InputBits<'a> {
+        pub fn new<T: Metrics>(data: &'a [u8]) -> Self {
+            Self {
+                input: data,
+                current_byte: 0,
+                last_mask: 1,
+                code_value_bits: T::CODE_VALUE_BITS as i32,
+            }
+        }
+        fn get_bit(&mut self) -> Option<bool> {
+            if self.last_mask == 1 {
+                match self.input.pop() {
+                    None => {
+                        if self.code_value_bits <= 0 {
+                            return None;
+                            //panic!("IDK Man");
+                        } else {
+                            self.code_value_bits -= 8;
+                        }
+                    }
+                    Some(byte) => self.current_byte = byte as u32,
+                }
+                self.last_mask = 0x80;
+            } else {
+                self.last_mask >>= 1;
+            }
+            let bit = (self.current_byte & self.last_mask) != 0;
+            return Some(bit);
+        }
+    }
+    #[test]
+    fn bit_reader_test_i32() {
+        bit_reader_test_type::<i32>();
+    }
+    #[test]
+    fn bit_reader_test_u32() {
+        bit_reader_test_type::<u32>();
+    }
+    #[test]
+    fn bit_reader_test_i64() {
+        bit_reader_test_type::<i64>();
+    }
+    #[test]
+    fn bit_reader_test_u64() {
+        bit_reader_test_type::<u64>();
+    }
+    #[test]
+    fn bit_reader_test_i128() {
+        bit_reader_test_type::<i128>();
+    }
+    fn bit_reader_test_type<T: Metrics>() {
+        let mut br = BitReader::from(COMPRESSED_BYTES).with_repeat_bits(T::CODE_VALUE_BITS as u16);
+        let mut ib = InputBits::new::<T>(&COMPRESSED_BYTES);
+
+        while let Some(a) = ib.get_bit() {
+            let b = br.get_bit().unwrap();
+            assert_eq!(a, b);
+        }
+        let _ = br.get_bit().expect_err("Extra bits");
+    }
+}

src/lib.rs

@@ -0,0 +1,6 @@
+// https://marknelson.us/posts/2014/10/19/data-compression-with-arithmetic-coding.html
+pub mod model;
+#[allow(non_snake_case)]
+pub mod modelA;
+
+pub mod bit_buffer;

src/main.rs

@@ -1,99 +1,52 @@
-// https://marknelson.us/posts/2014/10/19/data-compression-with-arithmetic-coding.html
-use std::collections::HashMap;
+use std::{
+    env,
+    fs::File,
+    io::{self, BufReader, Read},
+    path::Path,
+};
 
-type Model = HashMap<u8, (f64, f64)>;
+use sac::model::Model;
+use sac::modelA::ModelA;
 
-fn get_symbol(model: &Model, d: f64) -> Option<u8> {
-    // Brute force
-    for (symbol, (start, end)) in model {
-        if d >= *start && d < *end {
-            return Some(*symbol);
+enum Mode {
+    Compress,
+    Decompress,
+}
+type CodeValue = u32;
+
+fn get_file<FilePath: AsRef<Path>>(filepath: FilePath) -> io::Result<Box<dyn Read>> {
+    let f = File::open(&filepath)?;
+    Ok(Box::new(BufReader::new(f)))
+}
+
+fn main() -> io::Result<()> {
+    let args: Vec<String> = env::args().collect();
+    let name = &args[0];
+
+    let (input, mode): (Box<dyn Read>, Mode) = match args.len() {
+        2 if args[1].to_lowercase() == "compress" => (Box::new(io::stdin()), Mode::Compress),
+        2 if args[1].to_lowercase() == "decompress" => (Box::new(io::stdin()), Mode::Decompress),
+
+        3 if args[1].to_lowercase() == "compress" => (get_file(&args[2])?, Mode::Compress),
+        3 if args[1].to_lowercase() == "decompress" => (get_file(&args[2])?, Mode::Decompress),
+
+        _ => {
+            eprintln!("Usage:");
+            eprintln!("{name} compress file > output # compress file and save to output");
+            eprintln!("{name} decompress file > output # decompress file and save to output");
+            return Err(io::Error::other(format!(
+                "Invalid command `{}`",
+                args.join(" ")
+            )));
         }
-    }
-    return None;
-}
+    };
 
-fn encode(data: &[u8], model: &Model) -> f64 {
-    let mut high: f64 = 1.0;
-    let mut low: f64 = 0.0;
-    for symbol in data {
-        let p = model.get(symbol).expect("Invalid/Unsupported data");
-        let range = high - low;
-        high = low + range * p.1;
-        low = low + range * p.0;
-    }
-    return low + (high - low) / 2.0;
-}
+    let model: ModelA<CodeValue> = ModelA::default();
+    let mut output = io::stdout().lock();
+    match mode {
+        Mode::Compress => model.compress(input, &mut output)?,
+        Mode::Decompress => model.decompress(input, &mut output)?,
+    };
 
-fn decode(message: f64, model: &Model) {
-    let mut high: f64 = 1.0;
-    let mut low: f64 = 0.0;
-    loop {
-        let range = high - low;
-        let d = (message - low) / range;
-        let c = match get_symbol(&model, d) {
-            Some(c) => c,
-            None => {
-                println!("");
-                eprintln!("Decode error: d={d}");
-                return;
-            }
-        };
-        if c == b'-' {
-            println!("");
-            return;
-        }
-        print!("{}", c as char);
-        let p = model.get(&c).expect("Decode error");
-        high = low + range * p.1;
-        low = low + range * p.0;
-    }
-}
-
-fn make_model(probabilities: &[(u8, f64)]) -> Model {
-    let mut model = HashMap::new();
-    let mut end: f64 = 0.0;
-    for (symbol, probability) in probabilities {
-        let start: f64 = end;
-        end = start + probability;
-        model.insert(*symbol, (start, end));
-        println!("{}: [{}, {})", *symbol as char, start, end);
-    }
-    return model;
-}
-const ENGLISH: &[(u8, f64)] = &[
-    (b'a', 0.08),
-    (b'b', 0.01),
-    (b'c', 0.02),
-    (b'd', 0.04),
-    (b'e', 0.12),
-    (b'f', 0.02),
-    (b'g', 0.02),
-    (b'h', 0.06),
-    (b'i', 0.07),
-    (b'j', 0.01),
-    (b'k', 0.01),
-    (b'l', 0.04),
-    (b'm', 0.02),
-    (b'n', 0.06),
-    (b'o', 0.07),
-    (b'p', 0.01),
-    (b'q', 0.01),
-    (b'r', 0.06),
-    (b's', 0.06),
-    (b't', 0.09),
-    (b'u', 0.02),
-    (b'v', 0.01),
-    (b'w', 0.02),
-    (b'x', 0.01),
-    (b'y', 0.02),
-    (b'z', 0.01),
-    (b' ', 0.01),
-    (b'-', 0.02),
-];
-fn main() {
-    let model: Model = make_model(ENGLISH);
-    let message = encode(b"hello world-", &model);
-    println!("{message}");
-    decode(message, &model);
+    Ok(())
 }

src/model.rs

@@ -0,0 +1,207 @@
+use num::{FromPrimitive, Integer};
+use std::{
+    io::{self, Read, Write},
+    ops::{BitAnd, Shl},
+    usize,
+};
+
+use crate::bit_buffer::{BitReader, BitWriter};
+
+trait Precision {
+    const PRECISION: usize;
+}
+macro_rules! unsignedImplDigits {
+    ($($type: ident),*) => { $(
+        impl Precision for $type {
+            const PRECISION: usize = (std::mem::size_of::<$type>() * 8);
+        }
+    )* };
+}
+macro_rules! signedImplDigits {
+    ($($type: ident),*) => { $(
+        impl Precision for $type {
+            const PRECISION: usize = (std::mem::size_of::<$type>() * 8) - 1;
+        }
+    )* };
+}
+unsignedImplDigits!(u32, u64);
+signedImplDigits!(i32, i64, i128);
+
+pub trait Metrics:
+    Integer + FromPrimitive + Copy + BitAnd<Output = Self> + Shl<Output = Self>
+{
+    const PRECISION: usize;
+
+    const FREQUENCY_BITS: usize = (Self::PRECISION / 2) - 1;
+    const CODE_VALUE_BITS: usize = Self::FREQUENCY_BITS + 2;
+    const MAX_CODE: usize = if Self::CODE_VALUE_BITS == 64 {
+        u64::MAX as usize
+    } else {
+        (1 << Self::CODE_VALUE_BITS) - 1
+    };
+    const MAX_FREQ: usize = (1 << Self::FREQUENCY_BITS) - 1;
+
+    const ONE_FOURTH: usize = 1 << (Self::CODE_VALUE_BITS - 2);
+    const ONE_HALF: usize = 2 * Self::ONE_FOURTH;
+    const THREE_FOURTHS: usize = 3 * Self::ONE_FOURTH;
+
+    fn print_metrics() {
+        println!("--------- Metrics ---------");
+        println!("      PRECISION: {}", Self::PRECISION);
+        println!(" FREQUENCY_BITS: {}", Self::FREQUENCY_BITS);
+        println!("CODE_VALUE_BITS: {}", Self::CODE_VALUE_BITS);
+        println!("       MAX_CODE: {}", Self::MAX_CODE);
+        println!("       MAX_FREQ: {}", Self::MAX_FREQ);
+        println!("     ONE_FOURTH: {}", Self::ONE_FOURTH);
+        println!("       ONE_HALF: {}", Self::ONE_HALF);
+        println!("  THREE_FOURTHS: {}", Self::THREE_FOURTHS);
+    }
+}
+impl<
+        T: Precision + Integer + FromPrimitive + Copy + BitAnd<Output = Self> + Shl<Output = Self>,
+    > Metrics for T
+{
+    const PRECISION: usize = T::PRECISION;
+}
+
+#[derive(Debug)]
+pub struct Prob<T> {
+    pub low: T,
+    pub high: T,
+    pub max_code: T,
+}
+
+pub trait Model<CodeWord: Metrics> {
+    fn get_probability(&mut self, c: i32) -> Prob<CodeWord>;
+    fn get_char(&mut self, scaled_value: CodeWord) -> Option<(i32, Prob<CodeWord>)>;
+    fn get_max_code(&self) -> CodeWord;
+
+    #[allow(non_snake_case)]
+    fn decompress<T: Read, O: Write, I: Into<BitReader<T>>>(
+        mut self,
+        input: I,
+        output: &mut O,
+    ) -> io::Result<()>
+    where
+        Self: Sized,
+    {
+        let ONE: CodeWord = CodeWord::one();
+        let ZERO: CodeWord = CodeWord::zero();
+        let ONE_HALF: CodeWord = CodeWord::from_usize(CodeWord::ONE_HALF).unwrap();
+        let ONE_FORTH: CodeWord = CodeWord::from_usize(CodeWord::ONE_FOURTH).unwrap();
+        let THREE_FOURTHS: CodeWord = CodeWord::from_usize(CodeWord::THREE_FOURTHS).unwrap();
+
+        let mut input: BitReader<T> = input
+            .into()
+            .with_repeat_bits(CodeWord::CODE_VALUE_BITS as u16);
+
+        let mut low: CodeWord = ZERO;
+        let mut high: CodeWord = CodeWord::from_usize(CodeWord::MAX_CODE).unwrap();
+        let mut value: CodeWord = ZERO;
+
+        for _ in 0..CodeWord::CODE_VALUE_BITS {
+            value = (value << CodeWord::one()) + if input.get_bit()? { ONE } else { ZERO };
+        }
+        loop {
+            let range: CodeWord = high - low + ONE;
+            let scaled_value = ((value - low + ONE) * self.get_max_code() - ONE) / range;
+            let (c, p) = self.get_char(scaled_value).unwrap();
+            if c > 255 || c < 0 {
+                break;
+            }
+            output.write(&[c as u8])?;
+            high = low + (range * p.high) / p.max_code - ONE;
+            low = low + (range * p.low) / p.max_code;
+            loop {
+                if high < ONE_HALF {
+                } else if low >= ONE_HALF {
+                    value = value - ONE_HALF;
+                    low = low - ONE_HALF;
+                    high = high - ONE_HALF
+                } else if low >= ONE_FORTH && high < THREE_FOURTHS {
+                    value = value - ONE_FORTH;
+                    low = low - ONE_FORTH;
+                    high = high - ONE_FORTH;
+                } else {
+                    break;
+                }
+                low = low << ONE;
+                high = (high << ONE) + ONE;
+                value = (value << ONE) + if input.get_bit()? { ONE } else { ZERO };
+            }
+        }
+        return Ok(());
+    }
+
+    #[allow(non_snake_case)]
+    fn compress<IN: Read, OUT: Write>(mut self, input: IN, output: &mut OUT) -> std::io::Result<()>
+    where
+        Self: Sized,
+    {
+        let ONE: CodeWord = CodeWord::one();
+        let ZERO: CodeWord = CodeWord::zero();
+        let MAX_CODE: CodeWord = CodeWord::from_usize(CodeWord::MAX_CODE).unwrap();
+        let ONE_HALF: CodeWord = CodeWord::from_usize(CodeWord::ONE_HALF).unwrap();
+        let ONE_FORTH: CodeWord = CodeWord::from_usize(CodeWord::ONE_FOURTH).unwrap();
+        let THREE_FOURTHS: CodeWord = CodeWord::from_usize(CodeWord::THREE_FOURTHS).unwrap();
+
+        let mut output: BitWriter<OUT> = output.into();
+
+        let mut pending_bits: i32 = 0;
+        let mut low: CodeWord = ZERO;
+        let mut high: CodeWord = MAX_CODE;
+
+        let mut iter = input
+            .bytes()
+            .map(|r| r.map(|b| b as i32))
+            .chain([Ok(256_i32)]);
+        while let Some(Ok(mut c)) = iter.next() {
+            if c > 255 || c < 0 {
+                c = 256;
+            }
+            let p = self.get_probability(c);
+            let range: CodeWord = high - low + ONE;
+            high = low + (range * p.high / p.max_code) - ONE;
+            low = low + (range * p.low / p.max_code);
+
+            loop {
+                if high < ONE_HALF {
+                    write_with_pending(false, &mut pending_bits, &mut output)?;
+                } else if low >= ONE_HALF {
+                    write_with_pending(true, &mut pending_bits, &mut output)?;
+                } else if low >= ONE_FORTH && high < THREE_FOURTHS {
+                    pending_bits += 1;
+                    low = low - ONE_FORTH;
+                    high = high - ONE_FORTH;
+                } else {
+                    break;
+                }
+                high = ((high << ONE) + ONE) & MAX_CODE;
+                low = (low << ONE) & MAX_CODE;
+            }
+            if c == 256 {
+                break;
+            }
+        }
+        pending_bits += 1;
+        if low < ONE_FORTH {
+            write_with_pending(false, &mut pending_bits, &mut output)?;
+        } else {
+            write_with_pending(true, &mut pending_bits, &mut output)?;
+        }
+
+        return output.flush();
+    }
+}
+fn write_with_pending<W: std::io::Write>(
+    bit: bool,
+    pending: &mut i32,
+    output: &mut BitWriter<W>,
+) -> std::io::Result<()> {
+    output.write(bit)?;
+    for _ in 0..*pending {
+        output.write(!bit)?;
+    }
+    *pending = 0;
+    Ok(())
+}

src/modelA.rs

@@ -0,0 +1,99 @@
+use crate::model::{Metrics, Model, Prob};
+
+pub struct ModelA<T> {
+    cumulative_frequency: [T; 258],
+    m_frozen: bool,
+}
+
+impl<T: Metrics> Default for ModelA<T> {
+    fn default() -> Self {
+        let m_frozen = false;
+        let mut cumulative_frequency = [T::zero(); 258];
+        for i in 0..258 {
+            cumulative_frequency[i] = T::from_usize(i).unwrap();
+        }
+        Self {
+            cumulative_frequency,
+            m_frozen,
+        }
+    }
+}
+
+impl<T: Metrics> ModelA<T> {
+    pub fn print_metrics(&self) {
+        T::print_metrics();
+    }
+    fn update(&mut self, c: i32) {
+        for i in (c as usize + 1)..258 {
+            self.cumulative_frequency[i] = self.cumulative_frequency[i] + T::one();
+        }
+        if self.cumulative_frequency[257] >= T::from_usize(T::MAX_FREQ).unwrap() {
+            self.m_frozen = true;
+        }
+    }
+}
+impl<T: Metrics> Model<T> for ModelA<T> {
+    fn get_probability(&mut self, c: i32) -> crate::model::Prob<T> {
+        let p = Prob {
+            low: self.cumulative_frequency[c as usize],
+            high: self.cumulative_frequency[c as usize + 1],
+            max_code: self.cumulative_frequency[257],
+        };
+        if !self.m_frozen {
+            self.update(c);
+        }
+        return p;
+    }
+
+    fn get_char(&mut self, scaled_value: T) -> Option<(i32, crate::model::Prob<T>)> {
+        for i in 0..258 {
+            if scaled_value < self.cumulative_frequency[i + 1] {
+                let p = Prob {
+                    low: self.cumulative_frequency[i],
+                    high: self.cumulative_frequency[i + 1],
+                    max_code: self.cumulative_frequency[257],
+                };
+                if !self.m_frozen {
+                    self.update(i as i32)
+                }
+                return Some((i as i32, p));
+            }
+        }
+        return None;
+    }
+
+    fn get_max_code(&self) -> T {
+        self.cumulative_frequency[257]
+    }
+}
+
+#[cfg(test)]
+pub mod tests {
+    use super::*;
+    pub const UNCOMPRESSED_BYTES: &[u8; 13] = b"hello world-\n";
+    /// Compressed bytes taken from output of the c++ version
+    pub const COMPRESSED_BYTES: [u8; 14] = [
+        0x67, 0xfc, 0x3e, 0x4a, 0x9d, 0x03, 0x7f, 0x35, 0xf1, 0x08, 0xd8, 0xa6, 0xbc, 0xda,
+    ];
+
+    #[test]
+    fn compression_test() {
+        let model: ModelA<i32> = ModelA::default();
+        let mut enc = Vec::new();
+        model.compress(&UNCOMPRESSED_BYTES[..], &mut enc).unwrap();
+        assert_eq!(COMPRESSED_BYTES.len(), enc.len());
+        for (a, b) in enc.iter().zip(COMPRESSED_BYTES.iter()) {
+            assert_eq!(a, b);
+        }
+    }
+    #[test]
+    fn decompression_test() {
+        let model: ModelA<i32> = ModelA::default();
+        let mut dec = Vec::new();
+        model.decompress(&COMPRESSED_BYTES, &mut dec).unwrap();
+        assert_eq!(UNCOMPRESSED_BYTES.len(), dec.len());
+        for (a, b) in dec.iter().zip(UNCOMPRESSED_BYTES.iter()) {
+            assert_eq!(a, b);
+        }
+    }
+}