mod heaptrack; mod nextjs; mod turbopack; use std::{ any::Any, env, fs::File, io::{self, BufReader, Read, Seek, SeekFrom}, path::PathBuf, sync::Arc, thread::{self, JoinHandle}, time::{Duration, Instant}, }; use anyhow::Result; use flate2::bufread::GzDecoder; use crate::{ reader::{heaptrack::HeaptrackFormat, nextjs::NextJsFormat, turbopack::TurbopackFormat}, store_container::StoreContainer, }; const MIN_INITIAL_REPORT_SIZE: u64 = 100 * 1024 * 1024; trait TraceFormat { type Reused: Default; fn read(&mut self, buffer: &[u8], reuse: &mut Self::Reused) -> Result; fn stats(&self) -> String { String::new() } } type ErasedReused = Box; struct ErasedTraceFormat(Box); trait ObjectSafeTraceFormat { fn create_reused(&self) -> ErasedReused; fn read(&mut self, buffer: &[u8], reuse: &mut ErasedReused) -> Result; fn stats(&self) -> String; } impl ObjectSafeTraceFormat for T where T::Reused: 'static, { fn create_reused(&self) -> ErasedReused { Box::new(T::Reused::default()) } fn read(&mut self, buffer: &[u8], reuse: &mut ErasedReused) -> Result { let reuse = reuse.downcast_mut().expect("Type of reuse is invalid"); TraceFormat::read(self, buffer, reuse) } fn stats(&self) -> String { TraceFormat::stats(self) } } impl ObjectSafeTraceFormat for ErasedTraceFormat { fn create_reused(&self) -> ErasedReused { self.0.create_reused() } fn read(&mut self, buffer: &[u8], reuse: &mut ErasedReused) -> Result { self.0.read(buffer, reuse) } fn stats(&self) -> String { self.0.stats() } } #[derive(Default)] enum TraceFile { Raw(File), Zstd(zstd::Decoder<'static, BufReader>), Gz(GzDecoder>), #[default] Unloaded, } impl TraceFile { fn read(&mut self, buffer: &mut [u8]) -> io::Result { match self { Self::Raw(file) => file.read(buffer), Self::Zstd(decoder) => decoder.read(buffer), Self::Gz(decoder) => decoder.read(buffer), Self::Unloaded => unreachable!(), } } fn stream_position(&mut self) -> io::Result { match self { Self::Raw(file) => file.stream_position(), Self::Zstd(decoder) => decoder.get_mut().stream_position(), Self::Gz(decoder) => decoder.get_mut().stream_position(), Self::Unloaded => unreachable!(), } } fn seek(&mut self, pos: SeekFrom) -> io::Result { match self { Self::Raw(file) => file.seek(pos), Self::Zstd(decoder) => decoder.get_mut().seek(pos), Self::Gz(decoder) => decoder.get_mut().seek(pos), Self::Unloaded => unreachable!(), } } fn size(&mut self) -> io::Result { match self { Self::Raw(file) => file.metadata().map(|m| m.len()), Self::Zstd(decoder) => decoder.get_mut().get_ref().metadata().map(|m| m.len()), Self::Gz(decoder) => decoder.get_mut().get_ref().metadata().map(|m| m.len()), Self::Unloaded => unreachable!(), } } } pub struct TraceReader { store: Arc, path: PathBuf, } impl TraceReader { pub fn spawn(store: Arc, path: PathBuf) -> JoinHandle<()> { let mut reader = Self { store, path }; std::thread::spawn(move || reader.run()) } pub fn run(&mut self) { let mut file_warning_printed = false; loop { let read_success = self.try_read(); if !file_warning_printed && !read_success { println!("Unable to read trace file at {:?}, waiting...", self.path); file_warning_printed = true; } thread::sleep(Duration::from_millis(500)); } } fn trace_file_from_file(&self, file: File) -> io::Result { let path = &self.path.to_string_lossy(); Ok(if path.ends_with(".zst") { TraceFile::Zstd(zstd::Decoder::new(file)?) } else if path.ends_with(".gz") { TraceFile::Gz(GzDecoder::new(BufReader::new(file))) } else { TraceFile::Raw(file) }) } fn try_read(&mut self) -> bool { let Ok(mut file) = File::open(&self.path) else { return false; }; println!("Trace file opened"); let stop_at = env::var("STOP_AT") .unwrap_or_default() .parse() .map_or(u64::MAX, |v: u64| v * 1024 * 1024); if stop_at != u64::MAX { println!("Will stop reading file at {} MB", stop_at / 1024 / 1024) } { let mut store = self.store.write(); store.reset(); } let mut format: Option<(ErasedTraceFormat, ErasedReused)> = None; let mut current_read = 0; let mut initial_read = file .seek(SeekFrom::End(0)) .ok() .map(|total| (total, Instant::now())); if file.seek(SeekFrom::Start(0)).is_err() { return false; } let mut file = match self.trace_file_from_file(file) { Ok(f) => f, Err(err) => { println!("Error creating zstd decoder: {err}"); return false; } }; let mut buffer = Vec::new(); let mut index = 0; let mut chunk = vec![0; 64 * 1024 * 1024]; loop { match file.read(&mut chunk) { Ok(bytes_read) => { if bytes_read == 0 { if let Some(value) = self.wait_for_more_data( &mut file, &mut initial_read, format.as_ref().map(|(f, _)| f), ) { return value; } } else { // If we have partially consumed some data, and we are at buffer capacity, // remove the consumed data to make more space. if index > 0 && buffer.len() + bytes_read > buffer.capacity() { buffer.splice(..index, std::iter::empty()); index = 0; } buffer.extend_from_slice(&chunk[..bytes_read]); if format.is_none() && buffer.len() >= 8 { let erased_format = if buffer.starts_with(b"TRACEv0") { index = 7; ErasedTraceFormat(Box::new(TurbopackFormat::new( self.store.clone(), ))) } else if buffer.starts_with(b"[{\"name\"") { ErasedTraceFormat(Box::new(NextJsFormat::new(self.store.clone()))) } else if buffer.starts_with(b"v ") { ErasedTraceFormat(Box::new(HeaptrackFormat::new( self.store.clone(), ))) } else { // Fallback to the format without magic bytes // TODO Remove this after a while and show an error instead ErasedTraceFormat(Box::new(TurbopackFormat::new( self.store.clone(), ))) }; let reuse = erased_format.create_reused(); format = Some((erased_format, reuse)); } if let Some((format, reuse)) = &mut format { match format.read(&buffer[index..], reuse) { Ok(bytes_read) => { index += bytes_read; } Err(err) => { println!("Trace file error: {err}"); return true; } } if self.store.want_to_read() { thread::yield_now(); } let prev_read = current_read; current_read += bytes_read as u64; if let Some((total, start)) = &mut initial_read { let old_mbs = prev_read / (97 * 1024 * 1024); let new_mbs = current_read / (97 * 1024 * 1024); if old_mbs != new_mbs { let pos = file.stream_position().unwrap_or(current_read); if pos > *total { *total = file.size().unwrap_or(pos); } *total = (*total).max(pos); let percentage = pos * 100 / *total; let read = pos / (1024 * 1024); let uncompressed = current_read / (1024 * 1024); let total = *total / (1024 * 1024); let stats = format.stats(); print!( "{}% read ({}/{} MB, {} MB/s)", percentage, read, total, read * 1000 / (start.elapsed().as_millis() + 1) as u64 ); if uncompressed != read { print!(" ({uncompressed} MB uncompressed)"); } if stats.is_empty() { println!(); } else { println!(" - {stats}"); } } } if current_read >= stop_at { println!( "Stopped reading file as requested by STOP_AT env var. \ Waiting for new file..." ); self.wait_for_new_file(&mut file); return true; } } } } Err(err) => { if err.kind() == io::ErrorKind::UnexpectedEof { if let Some(value) = self.wait_for_more_data( &mut file, &mut initial_read, format.as_ref().map(|(f, _)| f), ) { return value; } } else { // Error reading file, maybe it was removed println!("Error reading trace file: {err:?}"); return true; } } } } } fn wait_for_more_data( &mut self, file: &mut TraceFile, initial_read: &mut Option<(u64, Instant)>, format: Option<&ErasedTraceFormat>, ) -> Option { let Ok(pos) = file.stream_position() else { return Some(true); }; if let Some((total, start)) = initial_read.take() { if let Some(format) = format { let stats = format.stats(); println!("{stats}"); } if total > MIN_INITIAL_REPORT_SIZE { println!( "Initial read completed ({} MB, {}s)", total / (1024 * 1024), (start.elapsed().as_millis() / 100) as f32 / 10.0 ); } } loop { // No more data to read, sleep for a while to wait for more data thread::sleep(Duration::from_millis(100)); let Ok(mut real_file) = File::open(&self.path) else { return Some(true); }; let Ok(end) = real_file.seek(SeekFrom::End(0)) else { return Some(true); }; if end < pos { // new file return Some(true); } else if end != pos { // file has more data return None; } } } fn wait_for_new_file(&self, file: &mut TraceFile) { let Ok(pos) = file.stream_position() else { return; }; loop { thread::sleep(Duration::from_millis(1000)); let Ok(end) = file.seek(SeekFrom::End(0)) else { return; }; if end < pos { return; } } } }