Mercurial > repos > petr-novak > repeat_annotation_pipeline3
diff get_contigs_from_re_archive.py @ 1:814cba36e435 draft
Uploaded
author | mvdbeek |
---|---|
date | Mon, 21 Feb 2022 10:21:39 +0000 |
parents | ea6a3059a6af |
children |
line wrap: on
line diff
--- a/get_contigs_from_re_archive.py Mon Oct 18 11:01:20 2021 +0000 +++ b/get_contigs_from_re_archive.py Mon Feb 21 10:21:39 2022 +0000 @@ -8,6 +8,7 @@ import zipfile import tempfile import textwrap +import os def parse_args(): '''Argument parsin''' @@ -143,7 +144,10 @@ ids = [] s = [] for i in fobj: - ii = i.decode('utf-8') + if isinstance(i, str): + ii = i + else: + ii = i.decode('utf-8') if ii[0] == ">": ids.append(ii) s.append("") @@ -151,6 +155,21 @@ s[-1] = s[-1] + ii.strip() return ids, s + +def extract_contigs_from_re_directory(re_dir, aln_output): + with open(aln_output, 'w') as fout: + for subdir, dirs, files in os.walk(re_dir): + for fn in files: + fn_full = subdir + os.sep + fn + if re.match('^.+seqclust.+[.]aln$', fn_full): + print(fn_full) + with open(fn_full) as aln: + for l in aln: + fout.write(l) + return aln_output + + + def extract_tarean_contigs_from_re_archive(archive): with zipfile.ZipFile(archive, 'r') as zip_object: flist = zip_object.infolist() @@ -168,10 +187,21 @@ return ids_all, seqs_all -def extract_contigs_from_re_directory(dir, aln_output): - # TODO - pass - +def extract_tarean_contigs_from_re_directory(re_dir): + seqs_all = [] + ids_all = [] + for subdir, dirs, files in os.walk(re_dir): + for fn in files: + fn_full = subdir + os.sep + fn + if re.match("^.+seqclust.+dir_CL[0-9]+[/]tarean_contigs.fasta", fn_full): + print (fn_full) + with open(fn_full) as fobj: + ids, seqs = read_tarean_fasta(fobj) + # wrap sequences + seqs = ["\n".join(textwrap.wrap(s + s, 80)) for s in seqs] + seqs_all += seqs + ids_all += ids + return ids_all, seqs_all def filter_contigs(consensus, coverage, min_coverage=5): x = "".join([ @@ -184,11 +214,19 @@ def main(): args = parse_args() - # extract aln from archive - ids, seqs = extract_tarean_contigs_from_re_archive(args.re_file) - aln_file = extract_contigs_from_re_archive( - args.re_file, - tempfile.NamedTemporaryFile().name) + if os.path.isdir(args.re_file): + # extract from directory + ids, seqs = extract_tarean_contigs_from_re_directory(args.re_file) + aln_file = extract_contigs_from_re_directory( + args.re_file, + tempfile.NamedTemporaryFile().name) + else: + # extract aln from archive + ids, seqs = extract_tarean_contigs_from_re_archive(args.re_file) + aln_file = extract_contigs_from_re_archive( + args.re_file, + tempfile.NamedTemporaryFile().name) + with open(aln_file, 'r') as f1, open(args.fasta, 'w') as ffasta: while True: contig_name, seq_start = get_header(f1)