add subcommand grep

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "fakit"
-version = "0.3.2"
+version = "0.3.3"
 edition = "2021"
 authors = ["sharkLoc <mmtinfo@163.com>"]
 rust-version = "1.65.0"
@@ -27,3 +27,5 @@ regex = "1.9.5"
 rust-htslib = "0.40.2"
 xz2 = "0.1.7"
 
+[profile.release]
+strip = true

README.md

@@ -19,7 +19,7 @@ cargo install fakit
 ```bash
 Fakit: A simple program for fasta file manipulation
 
-Version: 0.3.2
+Version: 0.3.3
 
 Authors: sharkLoc <mmtinfo@163.com>
 Source code: https://github.com/sharkLoc/fakit.git
@@ -47,6 +47,7 @@ Commands:
   rename   rename sequence id in fasta file
   reverse  get a reverse-complement of fasta file [aliases: rev]
   window   stat dna fasta gc content by sliding windows [aliases: slide]
+  grep     grep fasta sequences by name/seq
   seq      convert all bases to lower/upper case, filter by length
   sort     sort fasta file by name/seq/gc/length
   search   search subsequences/motifs from fasta file

src/cli.rs

@@ -4,7 +4,7 @@ use clap::{Parser,value_parser};
 #[command(
     name = "Fakit",
     author = "sharkLoc",
-    version = "0.3.2",
+    version = "0.3.3",
     about = "A simple program for fasta file manipulation",
     long_about = None,
     next_line_help = false,
@@ -173,6 +173,27 @@ pub enum Subcli {
         #[arg(short = 'o', long = "out",verbatim_doc_comment, value_name = "str")]
         output: Option<String>,
     }, 
+    /// grep fasta sequences by name/seq
+    grep {
+        /// input fasta file, or read from stdin
+        input: Option<String>,
+        /// specify regex pattern/motif, e.g., -p "ATC{2,}" or -p ATCCG, search multiple pattern/motif, -p "ATCCG|GCTAA"
+        /// when searching by sequence name, the sequence prefix ">" is not included in the header.
+        #[arg(short = 'p', long = "pattern",verbatim_doc_comment, value_name = "str")]
+        pat: String,
+        /// grep sequences by full name
+        #[arg(short = 'n', long = "by-name", help_heading = Some("FLAGS"))]
+        name: bool,
+        /// grep sequences by sequence
+        #[arg(short = 's', long = "by-seq", help_heading = Some("FLAGS"))]
+        seq: bool,
+        /// grep sequences by sequence
+        #[arg(short = 'i', long = "ignore-case", help_heading = Some("FLAGS"))]
+        ignore: bool,
+        /// output search result file name, or write to stdout, file name ending in .gz/.bz2/.xz will be compressed automatically
+        #[arg(short = 'o', long = "out", value_name = "str" )]
+        output: Option<String>,
+    },
     /// convert all bases to lower/upper case, filter by length
     seq {
         /// input fasta file, or read from stdin
@@ -225,11 +246,11 @@ pub enum Subcli {
     },
     /// search subsequences/motifs from fasta file
     search {
-        /// input fasta[.gz] file, or read from stdin
+        /// input fasta file, or read from stdin
         input: Option<String>,
         /// specify uppercase pattern/motif, e.g., -p "ATC{2,}" or -p ATCCG
         ///search multiple pattern/motif, -p "ATCCG|GCTAA"
-        #[arg(short = 'p', long = "pattern",verbatim_doc_comment)]
+        #[arg(short = 'p', long = "pattern",verbatim_doc_comment, value_name = "str")]
         pat: String,
         /// if specified, show header in result
         #[arg(short = 'H', long = "header", help_heading = Some("FLAGS"))]

src/grep.rs

@@ -0,0 +1,85 @@
+use anyhow::Error;
+use regex::RegexBuilder;
+use bio::io::fasta;
+use std::time::Instant;
+use log::*;
+use crate::utils::*;
+use crate::wrap::*;
+
+
+pub fn grep_fasta(
+    file: &Option<&str>,
+    out: &Option<&str>,
+    pat: &str,
+    case: bool,
+    by_id: bool,
+    by_seq: bool,
+    line_width: usize,
+    compression_level: u32,
+) -> Result<(),Error> {    
+    let start = Instant::now();
+    if let Some(file) = file {
+        info!("reading from file: {}",file);
+    } else {
+        info!("reading from stdin");
+    }
+
+    let mut n = 0usize;
+    if by_id {
+        n += 1;
+    }
+    if by_seq {
+        n += 1;
+    }
+    match n {
+        1 => { () }
+        0 => {
+            error!("please specifiy one of the flags: -n or -s");
+            std::process::exit(1);
+        }
+        _ => {
+            error!("only one of the flags -n (--by-name) or -s (--by-seq), is allowed");
+            std::process::exit(1);
+        }
+    }
+    info!("regex pattern is: {}",pat);
+    let mut n = 0usize;
+
+    let re = RegexBuilder::new(pat)
+        .case_insensitive(case)
+        .unicode(true)
+        .build()?;
+
+    let mut fo = file_writer(out, compression_level).map(fasta::Writer::new)?;
+
+    if by_seq {
+        let fp = file_reader(file).map(fasta::Reader::new)?;
+        for rec in fp.records().flatten(){
+            let seq_str = std::str::from_utf8(rec.seq())?;
+            if let Some(_) = re.captures(seq_str) {
+                n += 1;
+                let seq_new = wrap_fasta(rec.seq(), line_width)?;
+                fo.write(rec.id(), rec.desc(), seq_new.as_slice())?;  
+            }
+        }
+    }
+    if by_id {
+        let fp = file_reader(file).map(fasta::Reader::new)?;
+        for rec in fp.records().flatten(){
+            let name = if let Some(desc) = rec.desc() {
+                format!("{} {}",rec.id(), desc)
+            } else {
+                rec.id().to_owned()
+            };
+            if let Some(_) = re.captures(&name) {
+                n += 1;
+                let seq_new = wrap_fasta(rec.seq(), line_width)?;
+                fo.write(rec.id(), rec.desc(), seq_new.as_slice())?;  
+            }
+        }
+    }
+    fo.flush()?;
+    info!("total match sequences number: {}",n);
+    info!("time elapsed is: {:?}",start.elapsed());
+    Ok(())
+}

src/main.rs

@@ -24,6 +24,8 @@ mod reverse;
 use reverse::*;
 mod search;
 use search::*;
+mod grep;
+use grep::*;
 mod seq;
 use seq::*;
 mod shuffle;
@@ -189,6 +191,21 @@ fn main() -> Result<(), Error> {
                 }
             }
         } 
+        Subcli::grep { input, pat, name, seq, ignore, output } => {
+            if let Some(input) = input {
+                if let Some(out) = output {
+                    grep_fasta(&Some(&input), &Some(&out), &pat, ignore, name, seq, args.width, args.compression_level)?;
+                } else {
+                    grep_fasta(&Some(&input), &None, &pat, ignore, name, seq, args.width, args.compression_level)?;
+                }
+            } else {
+                if let Some(out) = output {
+                    grep_fasta(&None, &Some(&out), &pat, ignore, name, seq, args.width, args.compression_level)?;
+                } else {
+                    grep_fasta(&None, &None, &pat, ignore, name, seq, args.width, args.compression_level)?;
+                }
+            }
+        }
         Subcli::seq { input, lower, upper, min, max, gc_min, gc_max, out } => {
             if let Some(input) = input {
                 if let Some(out) = out {