update dependencies, create benches

The fixed line length solution is still just as slow, even using memmap

src/main/rust/Cargo.toml

@@ -6,18 +6,17 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-bstr = "1.9.1"
+bstr = "1.10.0"
 fast-float = "0.2.0"
 memchr = "2.7.4"
-memmap = "0.7.0"
-polars = { version = "0.36.2", features = ["csv", "lazy", "nightly", "streaming"]}
+memmap2 = "0.9.4"
 rayon = "1.10.0"
 rustc-hash = "2.0.0"
-libc = "0.2.155"
-smol = "2.0.0"
+libc = "0.2.158"
+smol = "2.0.1"
 
 [dev-dependencies]
-criterion = { version = "0.5", features = ["html_reports"] }
+criterion = { version = "0.5.1", features = ["html_reports"] }
 
 [features]
 json = []
@@ -36,7 +35,7 @@ name = "multi_threaded"
 harness = false
 
 [[bench]]
-name = "polars"
+name = "multi_threaded_smol"
 harness = false
 
 [[bench]]

src/main/rust/benches/libraries.rs

@@ -1,8 +1,9 @@
 use criterion::{Criterion, criterion_group, criterion_main};
+use onebrc::implementations::libraries::run;
 
 pub fn criterion_benchmark(c: &mut Criterion) {
-    c.bench_function("polars", |b| {b.iter(|| /*run_polars()*/ ())});
+    c.bench_function("libraries", |b| {b.iter(|| run())});
 }
 
 criterion_group!(benches, criterion_benchmark);
-criterion_main!(benches);
+criterion_main!(benches);

src/main/rust/benches/multi_threaded_smol.rs

@@ -0,0 +1,9 @@
+use criterion::{Criterion, criterion_group, criterion_main};
+use onebrc::implementations::multi_threaded_smol::run;
+
+pub fn criterion_benchmark(c: &mut Criterion) {
+    c.bench_function("multithreadedsmol", |b| {b.iter(|| run())});
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);

src/main/rust/src/bin/polars.rs

@@ -1,4 +0,0 @@
-
-fn main() {
-    // let _ = run_polars();
-}

src/main/rust/src/implementations.rs

@@ -4,7 +4,6 @@ pub mod multi_threaded;
 pub mod multi_threaded_smol;
 pub mod multi_threaded_structured;
 pub mod phcs;
-pub mod polars;
 pub mod reference_impl;
 pub mod single_thread;
 pub mod smol;

src/main/rust/src/implementations/libraries.rs

@@ -1,67 +1,57 @@
-use std::collections::HashMap;
-use std::io::{BufRead, Seek, SeekFrom};
-use std::sync::mpsc;
-use std::time::Instant;
-use std::{fs::File, io::BufReader, thread};
-use memmap::MmapOptions;
 use crate::models::station_measurements::StationMeasurements;
 use crate::utils::parse;
-use crate::utils::parse::hashstr;
+use memmap2::MmapOptions;
+use rustc_hash::{FxBuildHasher, FxHashMap as HashMap};
+use std::sync::mpsc;
+use std::time::Instant;
+use std::{fs::File, thread};
 
 const DEFAULT_HASHMAP_LENGTH: usize = 10000;
 
 pub fn run() {
-    const FILE_PATH: &str = "../../../measurements.txt";
     let now = Instant::now();
+    const FILE_PATH: &str = "../../../measurements.txt";
+    let file = File::open(FILE_PATH).expect("File measurements.txt not found");
+    let mmap = unsafe { MmapOptions::new().map(&file).unwrap() };
+    let file_length = mmap.len();
+    let hasher = FxBuildHasher::default();
+    let mut stations: HashMap<String, StationMeasurements> =
+        HashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
+    let (tx, rx) = mpsc::channel();
+    let cores = thread::available_parallelism().unwrap().into();
+    let chunk_length = file_length / cores;
+    let mut bounds = Vec::with_capacity(cores + 1);
+    let mut start = 0;
+    for _ in 0..cores {
+        let end = (start + chunk_length).min(mmap.len());
+        let next_new_line = match memchr::memchr(b'\n', &mmap[end..]) {
+            Some(v) => v,
+            None => {
+                assert_eq!(end, mmap.len());
+                0
+            }
+        };
+        let end = end + next_new_line;
+        bounds.push((start, end));
+        start = end + 1;
+    }
     thread::scope(|s| {
-        let mut stations: HashMap<usize, (String, StationMeasurements)> =
-            HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
-        let (tx, rx) = mpsc::channel();
-        let cores = thread::available_parallelism().unwrap().into();
-        let file = File::open(FILE_PATH).expect("File measurements.txt not found");
-        let mmap = unsafe { MmapOptions::new().map(&file).unwrap() };
-        let file_length = mmap.len();
-        let chunk_length = file_length / cores;
-        let mut bounds = Vec::with_capacity(cores + 1);
-        bounds.push(0);
-        for i in 1..cores {
-            let mut reader = BufReader::new(&file);
-            let mut byte_start = chunk_length * i;
-            reader
-                .seek(SeekFrom::Start(byte_start as u64))
-                .expect("could not seek");
-            let mut line = Vec::with_capacity(108);
-            let line_len = reader
-                .read_until(b'\n', &mut line)
-                .expect("could not read bytes");
-            byte_start += line_len;
-            bounds.push(byte_start);
-        }
-        bounds.push(file_length);
         for i in 0..cores {
             let tx = tx.clone();
-            let mut currposition = *bounds.get(i).unwrap();
-            let end = *bounds.get(i + 1).unwrap();
+            let (start, end) = *bounds.get(i).unwrap();
+            let mmap_slice = &mmap[start..end];
             s.spawn(move || {
-                let file = File::open(FILE_PATH).expect("File measurements.txt not found");
-                let mut reader = BufReader::new(&file);
-                reader.seek(SeekFrom::Start(currposition as u64)).unwrap();
-                let mut t_stations: HashMap<usize, (String, StationMeasurements)> =
-                    HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
-                let mut line = Vec::with_capacity(108);
-                loop {
-                    let line_len = reader
-                        .read_until(b'\n', &mut line)
-                        .expect("could not read bytes");
-                    if line_len == 0 {
+                let mut t_stations: HashMap<String, StationMeasurements> =
+                    HashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
+                for line in mmap_slice.split(|&byte| { byte == b'\n' }) {
+                    if line.len() == 0 {
                         break;
                     }
                     let (station, temp) = line.rsplit_once(|&byte| byte == b';').unwrap();
-                    let hash = hashstr(station);
                     let station = unsafe { String::from_utf8_unchecked(Vec::from(station)) };
-                    let temp = parse::temp(temp.split_last().unwrap().1);
-                    let measurements_option = t_stations.get_mut(&hash);
-                    if let Some((_, measurements)) = measurements_option {
+                    let temp = parse::temp(temp);
+                    let measurements_option = t_stations.get_mut(&station);
+                    if let Some(measurements) = measurements_option {
                         measurements.update(temp);
                     } else {
                         let measurements = StationMeasurements {
@@ -70,31 +60,26 @@ pub fn run() {
                             count: 1,
                             sum: temp,
                         };
-                        t_stations.insert(hash, (station, measurements));
+                        t_stations.insert(station, measurements);
                     }
-                    currposition += line_len;
-                    if currposition >= end {
-                        break;
-                    }
-                    line.clear();
                 }
                 let _ = tx.send(t_stations);
             });
         }
         drop(tx);
         while let Ok(t_stations) = rx.recv() {
-            for (&hash, (station, measurements)) in t_stations.iter() {
-                let joined_measurements_options = stations.get_mut(&hash);
-                if let Some((_, joined_measurements)) = joined_measurements_options {
+            for (station, measurements) in t_stations.iter() {
+                let joined_measurements_options = stations.get_mut(station);
+                if let Some(joined_measurements) = joined_measurements_options {
                     joined_measurements.merge(measurements);
                 } else {
-                    stations.insert(hash, (station.to_owned(), *measurements));
+                    stations.insert(station.to_owned(), *measurements);
                 }
             }
         }
         let mut stations: Vec<String> = stations
             .iter()
-            .map(|(_, (station, measurements))| {
+            .map(|(station, measurements)| {
                 let measurements = measurements.to_string();
                 #[cfg(feature = "json")]
                 {

src/main/rust/src/implementations/multi_threaded_smol.rs

@@ -1,10 +1,9 @@
 use smol::fs::File;
 use smol::io::{AsyncBufReadExt, AsyncSeekExt, BufReader, SeekFrom};
+use rustc_hash::{FxHashMap as HashMap, FxBuildHasher};
 
 use crate::models::station_measurements::StationMeasurements;
 use crate::utils::parse;
-use crate::utils::parse::hashstr;
-use std::collections::HashMap;
 use std::sync::mpsc;
 use std::thread;
 use std::time::Instant;
@@ -15,8 +14,9 @@ pub fn run() {
     const FILE_PATH: &str = "../../../measurements.txt";
     let now = Instant::now();
     thread::scope(|s| {
-        let mut stations: HashMap<usize, (String, StationMeasurements)> =
-            HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
+        let hasher = FxBuildHasher::default();
+        let mut stations: HashMap<String, StationMeasurements> =
+            HashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
         let (tx, rx) = mpsc::channel();
         let cores = thread::available_parallelism().unwrap().into();
         let bounds = smol::block_on(async {
@@ -57,8 +57,8 @@ pub fn run() {
                         .expect("File measurements.txt not found");
                     let mut reader = BufReader::new(&mut file);
                     reader.seek(SeekFrom::Start(currposition)).await.unwrap();
-                    let mut t_stations: HashMap<usize, (String, StationMeasurements)> =
-                        HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
+                    let mut t_stations: HashMap<String, StationMeasurements> =
+                        HashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
                     let mut line = Vec::with_capacity(108);
                     loop {
                         let line_len = reader
@@ -69,11 +69,10 @@ pub fn run() {
                             break;
                         }
                         let (station, temp) = line.rsplit_once(|&byte| byte == b';').unwrap();
-                        let hash = hashstr(station);
                         let station = unsafe { String::from_utf8_unchecked(Vec::from(station)) };
                         let temp = parse::temp(temp.split_last().unwrap().1);
-                        let measurements_option = t_stations.get_mut(&hash);
-                        if let Some((_, measurements)) = measurements_option {
+                        let measurements_option = t_stations.get_mut(&station);
+                        if let Some(measurements) = measurements_option {
                             measurements.update(temp);
                         } else {
                             let measurements = StationMeasurements {
@@ -82,7 +81,7 @@ pub fn run() {
                                 count: 1,
                                 sum: temp,
                             };
-                            t_stations.insert(hash, (station, measurements));
+                            t_stations.insert(station, measurements);
                         }
                         currposition += line_len as u64;
                         if currposition >= end {
@@ -96,18 +95,18 @@ pub fn run() {
         }
         drop(tx);
         while let Ok(t_stations) = rx.recv() {
-            for (&hash, (station, measurements)) in t_stations.iter() {
-                let joined_measurements_options = stations.get_mut(&hash);
-                if let Some((_, joined_measurements)) = joined_measurements_options {
+            for (station, measurements) in t_stations.iter() {
+                let joined_measurements_options = stations.get_mut(station);
+                if let Some(joined_measurements) = joined_measurements_options {
                     joined_measurements.merge(measurements);
                 } else {
-                    stations.insert(hash, (station.to_owned(), *measurements));
+                    stations.insert(station.to_owned(), *measurements);
                 }
             }
         }
         let mut stations: Vec<String> = stations
             .iter()
-            .map(|(_, (station, measurements))| {
+            .map(|(station, measurements)| {
                 let measurements = measurements.to_string();
                 #[cfg(feature = "json")]
                 {

src/main/rust/src/implementations/multi_threaded_structured.rs

@@ -1,12 +1,11 @@
-use std::collections::HashMap;
-use std::io::{Read, Seek, SeekFrom};
 use std::sync::mpsc;
 use std::time::Instant;
-use std::{fs::File, io::BufReader, thread};
+use std::{fs::File, thread};
 use std::ffi::CStr;
+use memmap2::MmapOptions;
+use rustc_hash::{FxBuildHasher, FxHashMap};
 use crate::models::station_measurements::StationMeasurements;
 use crate::utils::parse;
-use crate::utils::parse::hashstr;
 
 const DEFAULT_HASHMAP_LENGTH: usize = 10000;
 
@@ -14,55 +13,54 @@ pub fn run() {
     print!("\x1b[J");
     const FILE_PATH: &str = "structured_measurements.txt";
     let now = Instant::now();
+    let file = File::open(FILE_PATH).expect("File structured_measurements.txt not found");
+    let mmap = unsafe { MmapOptions::new().map(&file).unwrap() };
+    let file_length = mmap.len();
+    let hasher = FxBuildHasher::default();
+    let mut stations: FxHashMap<String, StationMeasurements> =
+        FxHashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
+    let (tx, rx) = mpsc::channel();
+    let cores = thread::available_parallelism().unwrap().into();
+    let chunk_length = file_length / cores;
+    let mut bounds = Vec::with_capacity(cores + 1);
+    let mut start = 0;
+    for _ in 0..cores {
+        let end = (start + chunk_length).min(mmap.len());
+        let next_new_line = match memchr::memchr(b'\n', &mmap[end..]) {
+            Some(v) => v,
+            None => {
+                assert_eq!(end, mmap.len());
+                0
+            }
+        };
+        let end = end + next_new_line;
+        bounds.push((start, end));
+        start = end + 1;
+    }
     thread::scope(|s| {
-        let mut stations: HashMap<usize, (String, StationMeasurements)> =
-            HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
-        let (tx, rx) = mpsc::channel();
-        let cores = thread::available_parallelism().unwrap().into();
-        let file = File::open(FILE_PATH).expect("File structured_measurements.txt not found");
-        let mut reader = BufReader::new(&file);
-        let file_length = reader.seek(SeekFrom::End(0)).unwrap();
-        let chunk_length = file_length as usize / cores;
-        let mut bounds = Vec::with_capacity(cores + 1);
-        bounds.push(0);
         for i in 0..cores {
             let tx = tx.clone();
-            let mut currposition = (i * chunk_length) as u64;
-            let end = ((i + 1) * chunk_length) as u64;
+            let (start, end) = *bounds.get(i).unwrap();
+            let mmap_slice = &mmap[start..end];
             s.spawn(move || {
-                let file = File::open(FILE_PATH).expect("File measurements.txt not found");
-                let mut reader = BufReader::new(&file);
-                reader.seek(SeekFrom::Start(currposition)).unwrap();
-                let mut t_stations: HashMap<usize, (String, StationMeasurements)> =
-                    HashMap::with_capacity(DEFAULT_HASHMAP_LENGTH);
-                let mut line = [0u8; 107];
-                let mut line_num = 0;
-                loop {
+                let mut t_stations: FxHashMap<String, StationMeasurements> =
+                    FxHashMap::with_capacity_and_hasher(DEFAULT_HASHMAP_LENGTH, hasher);
+                let lines = mmap_slice.chunks_exact(107);
+                for (line_num, line) in lines.enumerate() {
                     if line_num % 100000 == 0 {
                         print!("\x1b[{i};0Hlines: {line_num}");
                     }
-                    line_num += 1;
-                    let read_res = reader
-                        .read_exact(&mut line);
-                    match read_res {
-                        Ok(_) => (),
-                        Err(e) => match e.kind() {
-                            std::io::ErrorKind::UnexpectedEof => break,
-                            _ => panic!("Could not read")
-                        },
-                    };
                     let (station, temp) = unsafe { line.split_at_unchecked(100) };
-                    let hash = hashstr(station);
                     let station = {
                         if station[station.len() - 1] == 0u8 {
                             unsafe { std::str::from_utf8_unchecked(CStr::from_bytes_until_nul(station).unwrap().to_bytes()) }
                         } else {
                             unsafe { std::str::from_utf8_unchecked(station) }
                         }
-                    };
+                    }.to_owned();
                     let temp = parse::temp_new(&temp[1..6]);
-                    let measurements_option = t_stations.get_mut(&hash);
-                    if let Some((_, measurements)) = measurements_option {
+                    let measurements_option = t_stations.get_mut(&station);
+                    if let Some(measurements) = measurements_option {
                         measurements.update(temp);
                     } else {
                         let measurements = StationMeasurements {
@@ -71,11 +69,7 @@ pub fn run() {
                             count: 1,
                             sum: temp,
                         };
-                        t_stations.insert(hash, (station.to_string(), measurements));
-                    }
-                    currposition += 107;
-                    if currposition >= end {
-                        break;
+                        t_stations.insert(station, measurements);
                     }
                 }
                 let _ = tx.send(t_stations);
@@ -83,18 +77,18 @@ pub fn run() {
         }
         drop(tx);
         while let Ok(t_stations) = rx.recv() {
-            for (&hash, (station, measurements)) in t_stations.iter() {
-                let joined_measurements_options = stations.get_mut(&hash);
-                if let Some((_, joined_measurements)) = joined_measurements_options {
+            for (station, measurements) in t_stations.iter() {
+                let joined_measurements_options = stations.get_mut(station);
+                if let Some(joined_measurements) = joined_measurements_options {
                     joined_measurements.merge(measurements);
                 } else {
-                    stations.insert(hash, (station.to_owned(), *measurements));
+                    stations.insert(station.to_owned(), *measurements);
                 }
             }
         }
         let mut stations: Vec<String> = stations
             .iter()
-            .map(|(_, (station, measurements))| {
+            .map(|(station, measurements)| {
                 let measurements = measurements.to_string();
                 #[cfg(feature = "json")]
                 {

src/main/rust/src/implementations/polars.rs

@@ -1,31 +0,0 @@
-use polars::prelude::*;
-use std::time::Instant;
-use std::vec;
-
-pub fn run_polars() -> Result<DataFrame, PolarsError> {
-    let now = Instant::now();
-
-    let f1: Field = Field::new("station", DataType::String);
-    let f2: Field = Field::new("measure", DataType::Float64);
-    let sc: Schema = Schema::from_iter(vec![f1, f2]);
-
-    let q = LazyCsvReader::new("../../../measurements.txt")
-        .has_header(false)
-        .with_schema(Some(Arc::new(sc)))
-        .with_separator(b';')
-        .finish()?
-        .group_by(vec![col("station")])
-        .agg(vec![
-            col("measure").alias("min").min(),
-            col("measure").alias("mean").mean(),
-            col("measure").alias("max").max(),
-        ])
-        .sort("station", Default::default())
-        .with_streaming(true);
-
-    let df = q.collect()?;
-
-    println!("Time={} μs", now.elapsed().as_micros());
-
-    Ok(df)
-}

src/main/rust/src/implementations/reference_impl.rs

@@ -1,5 +1,5 @@
 use bstr::{BStr, ByteSlice};
-use memmap::MmapOptions;
+use memmap2::MmapOptions;
 use rayon::prelude::*;
 use rustc_hash::FxHashMap as HashMap;
 use std::time::Instant;