1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
//! Functions for parallel processing of record sets and records. //! //! Sequences are read and processed in batches (`RecordSet`) because sending //! data across channels has a performance impact. The process works as follows: //! //! * Sequence parsing is done in a background thread //! * Record sets are sent to worker threads, where expensive operations take //! place (e.g. sequence analysis). //! * The results are sent to the main thread along with the record sets. //! * The record sets are recycled by sending them back to the background //! reader. //! //! # Per-record processsing //! //! The easiest to use are the functions, which operate directly on sequence //! records without having to deal with record sets: //! //! * [`read_process_fasta_records`](read_process_fasta_records) //! * [`read_process_fastq_records`](read_process_fastq_records) //! * [`read_process_fastx_records`](read_process_fastx_records) //! //! They are specific for the given sequence format, but it is possible to //! generate functions for other types using the //! [`parallel_record_impl`](parallel_record_impl) macro. //! //! ## Example //! //! This example filters sequences by the occurrence of a pattern: //! //! ```no_run //! use seq_io::prelude::*; //! use seq_io::fastq::{Reader,Record}; //! use seq_io::parallel::read_process_fastq_records; //! use std::fs::File; //! use std::io::BufWriter; //! //! let reader = Reader::from_path("seqs.fastq").unwrap(); //! let mut writer = BufWriter::new(File::create("filtered.fastq").unwrap()); //! //! read_process_fastq_records(reader, 4, 2, //! |record, found| { // runs in worker //! *found = record.seq().windows(3).position(|s| s == b"AAA").is_some(); //! }, //! |record, found| { // runs in main thread //! if *found { //! record.write(&mut writer).unwrap(); //! } //! // Some(value) will stop the reader, and the value will be returned. //! // In the case of never stopping, we need to give the compiler a hint about the //! // type parameter, thus the special 'turbofish' notation is needed. //! None::<()> //! }).unwrap(); //! ``` //! //! # Record set processing //! //! It is still possible to directly work with record sets using the following //! generic functions: //! //! * [`read_process_recordsets`](read_process_recordsets) //! * [`read_process_recordsets_init`](read_process_recordsets_init) //! //! ## Example //! //! This example searches for the first occurrence of a sequence pattern and //! then stops the parser. //! //! ```no_run //! use seq_io::prelude::*; //! use seq_io::fastq; //! use seq_io::parallel::read_process_recordsets; //! //! let reader = fastq::Reader::from_path("seqs.fastq").unwrap(); //! //! read_process_recordsets(reader, 4, 2, //! |record_set, position| { //! // This function does the heavy work. //! // The code is not necessarily very efficient, just for demonstration. //! for (i, record) in record_set.into_iter().enumerate() { //! if let Some(pos) = record.seq().windows(3).position(|s| s == b"AAA") { //! *position = Some((i, pos)); //! } //! } //! *position = None; //! }, |mut record_sets| { //! // This function runs in the main thread. It provides a streaming iterator over //! // record sets and the corresponding return values from the worker function //! // (not necessarily in the same order as in the file) //! while let Some(result) = record_sets.next() { //! let (record_set, position) = result?; //! if let Some(&(i, pos)) = position.as_ref() { //! let record = record_set.into_iter().nth(i).unwrap(); //! println!("Found AAA in record {} at position {}", record.id().unwrap(), pos); //! return Ok(()); //! } //! } //! // Here, we need to give the compiler a type hint about the returned //! // result, since it is not smart enough to infer it. //! // In real-world programs, this may be less of an issue because the //! // returned result type is often known. //! Ok::<_, fastq::Error>(()) //! } //! ).expect("FASTQ reading error"); //! ``` use crate::{fasta, fastq, fastx}; use std::sync::mpsc; /// A simple trait required to be implemented for readers fed into the /// functions in this module. pub trait RecordSetReader { type RecordSet: Send; type Err: Send; fn fill_data(&mut self, record: &mut Self::RecordSet) -> Result<bool, Self::Err>; } /// This function reads record sets and processes them in parallel threads. /// /// * It takes a [`RecordSetReader`](RecordSetReader), which reads data into /// record sets in a background thread. /// * These are then sent to `n_workers` worker threads, where the heavy work /// is done in the `work` closure. /// * Once ready, the record sets and work results are sent to the main thread /// and provided to the `func` closure. The won't necessarily arrive in the /// same order as they were read. pub fn read_process_recordsets<R, W, F, O, Out>( reader: R, n_workers: u32, queue_len: usize, work: W, func: F, ) -> Out where R: RecordSetReader + Send, R::RecordSet: Default + Send, O: Default + Send, W: Send + Sync + Fn(&mut R::RecordSet, &mut O), F: FnOnce(ParallelDataSets<R::RecordSet, R::Err, O>) -> Out, { read_process_recordsets_init(|| Ok::<_, ()>(reader), n_workers, queue_len, work, func).unwrap() } /// Like [`read_process_recordsets`](read_process_recordsets), but additionally /// allows initiating the reader in the background thread using a closure /// (`reader_init`). /// This is useful for readers, which don't implement `Send`. /// The `reader_init` closure has to return a result. Errors are returned from /// the main function witout being mixed with reading errors. This may lead to /// nested `Result` being returned if the `func` closure returns `Result`. pub fn read_process_recordsets_init<R, Ri, Ei, W, F, O, Out>( reader_init: Ri, n_workers: u32, queue_len: usize, work: W, func: F, ) -> Result<Out, Ei> where R: RecordSetReader, Ri: Send + FnOnce() -> Result<R, Ei>, R::RecordSet: Default + Send, O: Default + Send, W: Send + Sync + Fn(&mut R::RecordSet, &mut O), F: FnOnce(ParallelDataSets<R::RecordSet, R::Err, O>) -> Out, Ei: Send, { let (done_send, done_recv) = mpsc::sync_channel(queue_len); let (empty_send, empty_recv) = mpsc::sync_channel(queue_len); crossbeam::scope(|scope| { let handle = scope.spawn::<_, Result<(), Ei>>(move |_| { let mut reader = reader_init()?; let mut pool = scoped_threadpool::Pool::new(n_workers); pool.scoped(|pool_scope| { let work = &work; loop { // recycle an old RecordSet sent back after use let (mut data, mut out) = if let Ok(r) = empty_recv.recv() { r } else { // ParallelDataSets dropped -> stop return; }; let done_send = done_send.clone(); match reader.fill_data(&mut data) { Ok(has_data) => { if !has_data { break; } // expensive work carried out by func() pool_scope.execute(move || { work(&mut data, &mut out); done_send.send(Some(Ok((data, out)))).ok(); }); } Err(e) => { done_send.send(Some(Err(e))).ok(); break; } } } pool_scope.join_all(); done_send.send(None).ok(); }); Ok(()) }); for _ in 0..queue_len { if empty_send .send((R::RecordSet::default(), O::default())) .is_err() { break; } } let dsets: ParallelDataSets<R::RecordSet, R::Err, O> = ParallelDataSets { empty_send, done_recv, current_recordset: (R::RecordSet::default(), O::default()), }; let out = func(dsets); handle.join().unwrap()?; Ok(out) }) .unwrap() } pub struct ParallelDataSets<D, E, O = ()> where D: Send, E: Send, O: Send, { empty_send: mpsc::SyncSender<(D, O)>, done_recv: mpsc::Receiver<Option<Result<(D, O), E>>>, current_recordset: (D, O), } impl<D, E, O> ParallelDataSets<D, E, O> where D: Send, E: Send, O: Send, { /// Returns a tuple of the next processed record set and processing results, /// if present. pub fn next(&mut self) -> Option<Result<(&mut D, &mut O), E>> { self.done_recv.recv().unwrap().map(move |result| { match result { Ok(d) => { let prev_rset = std::mem::replace(&mut self.current_recordset, d); self.empty_send.send(prev_rset).ok(); // error: channel closed is not a problem, happens after calling stop() Ok((&mut self.current_recordset.0, &mut self.current_recordset.1)) } Err(e) => Err(e), } }) } } /// Allows generating functions equivalent to the `read_process_xy_records` /// functions in this crate for your own types. This is rather a workaround /// because the generic approach ([`read_process_records_init`](read_process_records_init)) /// does currently not work. /// /// * `$format`: String specifying the name of the sequence format /// (for genrated documentation) /// * `$name`: name of the generated function /// * `$name_init`: name of another generated function, which takes a closure /// initializing the readers in the background thread. /// * `$RecordSet`: record set type (see [`RecordSetReader::RecordSet`](RecordSetReader::RecordSet)). /// In addition to the trait requirements, `&$RecordSet` needs to implement /// `IntoIterator<Item=$Record>`. /// * `$Record`: record type returned by the record set iterator. /// * `$Error`: reading error type ([`RecordSetReader::Err`](RecordSetReader::Err)) #[macro_export] macro_rules! parallel_record_impl { ($format:expr, $name:ident, $name_init:ident, $RecordSet:ty, $Record:ty, $Error:ty) => { _parallel_record_impl!( $format, $name, $name_init, $RecordSet, $Record, $Error, concat!("[`", stringify!($name), "`](", stringify!($name), ")") ); }; } macro_rules! _parallel_record_impl { ($format:expr, $name:ident, $name_init:ident, $RecordSet:ty, $Record:ty, $Error:ty, $name_link:expr) => { /// This function wraps [`read_process_recordsets`](read_process_recordsets), /// hiding the complexity related to record sets and allowing it to /// directly work on #[doc = $format] /// sequence records. /// /// Apart from this, the process is similar: /// /// * The records are read (as part of record sets) in a background /// thread. /// * Then they are sent to `n_workers` worker threads. Work is done /// in the `work` closure supplied to this function. /// * Once ready, records an results are sent to the main thread, /// where they are supplied to the `func` closure. The order of the /// records may be different. pub fn $name<R, S, W, F, O, Out>( reader: R, n_workers: u32, queue_len: usize, work: W, func: F, ) -> Result<Option<Out>, R::Err> where R: RecordSetReader<RecordSet = $RecordSet, Err = $Error> + Send, O: Default + Send, S: crate::core::PositionStore, W: Send + Sync + Fn($Record, &mut O), F: FnMut($Record, &mut O) -> Option<Out>, { $name_init::<_, _, _, _, _, _, _, $Error>( || Ok(reader), n_workers, queue_len, work, func ) } /// Like #[doc = $name_link] ///, but instead of a [`RecordSetReader`](RecordSetReader), it takes a /// closure (`reader_init`) returning an `RecordSetReader` instance. /// This allows using readers that don't implement `Send`. /// `reader_init` should return a result. The error type needs to /// implement `From<RecordSetReader::Err>` /// pub fn $name_init<R, Ri, S, W, F, O, Out, E>( reader_init: Ri, n_workers: u32, queue_len: usize, work: W, mut func: F, ) -> Result<Option<Out>, E> where R: RecordSetReader<RecordSet = $RecordSet, Err = $Error>, Ri: Send + FnOnce() -> Result<R, E>, O: Default + Send, S: crate::core::PositionStore, W: Send + Sync + Fn($Record, &mut O), F: FnMut($Record, &mut O) -> Option<Out>, E: Send + From<R::Err>, { read_process_recordsets_init( reader_init, n_workers, queue_len, |rset: &mut $RecordSet, output: &mut Vec<O>| { let mut record_iter = rset.into_iter(); for (out, record) in output.iter_mut().zip(&mut record_iter) { work(record, out); } for record in record_iter { output.push(O::default()); work(record, output.last_mut().unwrap()); } }, |mut records| { while let Some(result) = records.next() { let (rset, out) = result?; for (record, o) in rset.into_iter().zip(out.iter_mut()) { if let Some(out) = func(record, o) { return Ok(Some(out)); } } } Ok(None) }, ).and_then(From::from) } }; } parallel_record_impl!( "FASTA", read_process_fasta_records, read_process_fasta_records_init, fasta::RecordSet<S>, fasta::RefRecord<S>, fasta::Error ); parallel_record_impl!( "FASTQ", read_process_fastq_records, read_process_fastq_records_init, fastq::RecordSet<S>, fastq::RefRecord<S>, fastq::Error ); parallel_record_impl!( "FASTX", read_process_fastx_records, read_process_fastx_records_init, fastx::RecordSet<S>, fastx::RefRecord<S>, fastx::Error ); /// Using this function currently does not work due to a /// [compiler bug](https://github.com/rust-lang/rust/issues/62529). /// /// [`read_process_fasta_records`](read_process_fasta_records), /// [`read_process_fastq_records`](read_process_fastq_records) and /// [`read_process_fastx_records`](read_process_fastx_records) /// provide the same functionality for now /// (implemented using [`parallel_record_impl`](parallel_record_impl) macro). pub fn read_process_records_init<R, Ri, W, F, O, Out, E>( reader_init: Ri, n_workers: u32, queue_len: usize, work: W, mut func: F, ) -> Result<Option<Out>, E> where R: RecordSetReader, Ri: Send + FnOnce() -> Result<R, E>, R::RecordSet: Default + Send, for<'a> &'a R::RecordSet: IntoIterator + Send, O: Default + Send, W: Send + Sync + Fn(<&R::RecordSet as IntoIterator>::Item, &mut O), F: FnMut(<&R::RecordSet as IntoIterator>::Item, &mut O) -> Option<Out>, E: From<<R as RecordSetReader>::Err> + Send, { read_process_recordsets_init( reader_init, n_workers, queue_len, |rset, out: &mut Vec<O>| { let mut record_iter = rset.into_iter(); for mut d in (&mut record_iter).zip(out.iter_mut()) { work(d.0, &mut d.1); } for record in record_iter { out.push(O::default()); work(record, out.last_mut().unwrap()); } }, |mut records| { while let Some(result) = records.next() { let (rset, out) = result?; for (record, o) in rset.into_iter().zip(out.iter_mut()) { if let Some(out) = func(record, o) { return Ok(Some(out)); } } } Ok(None) }, ) .and_then(From::from) } macro_rules! impl_parallel_reader { ($($l:lifetime)?; $SeqReader:ty, $RecordSet:ty, $Error:ty, $read_fn:ident) => { impl<$($l,)? R, P, S> RecordSetReader for $SeqReader where R: std::io::Read, P: crate::policy::BufPolicy + Send, S: crate::core::PositionStore { type RecordSet = $RecordSet; type Err = $Error; fn fill_data(&mut self, rset: &mut $RecordSet) -> Result<bool, $Error> { self.$read_fn(rset) } } } } impl_parallel_reader!(; fasta::Reader<R, P, S>, fasta::RecordSet<S>, fasta::Error, read_record_set); impl_parallel_reader!(; fasta::single_line::Reader<R, P, S>, fasta::RecordSet<S>, fasta::Error, read_record_set); impl_parallel_reader!(; fastq::Reader<R, P, S>, fastq::RecordSet<S>, fastq::Error, read_record_set); impl_parallel_reader!(; fastq::multiline::Reader<R, P, S>, fastq::RecordSet<S>, fastq::Error, read_record_set); impl_parallel_reader!(; fastx::Reader<R, P, S>, fastx::RecordSet<S>, fastx::Error, read_record_set); impl_parallel_reader!(; fastx::multiline_qual::Reader<R, P, S>, fastx::RecordSet<S>, fastx::Error, read_record_set); impl_parallel_reader!('a ; &'a mut (dyn fastx::dynamic::FastxReader<R, P, S> + Send), fastx::RecordSet<S>, fastx::Error, read_record_set_fastx); impl_parallel_reader!('a ; Box<dyn fastx::dynamic::FastxReader<R, P, S> + Send + 'a>, fastx::RecordSet<S>, fastx::Error, read_record_set_fastx);