def check_format(self):
"""
检测文件是否满足要求,发生错误时应该触发FileError异常
:return: bool
"""
self.is_gz = self.is_gzip
if self.is_gz:
with gzip.open(self.prop['path'], 'rb') as f:
line1 = f.next()
line = f.next()
line = f.next()
line = f.next()
line5 = f.next()
if not (re.search(r'^@', line1) and re.search(r'^@', line5)):
raise FileError("非压缩后的fastq格式文件")
myline1 = re.split('_', line1)
myline2 = re.split('_', line5)
if len(myline1) > 1 and len(myline2) > 1:
self.has_sample_info = True
else:
with open(self.prop['path'], 'r') as r:
line = r.next()
if not re.search(r'^@', line):
raise FileError("fastq文件格式错误")
myline1 = re.split('_', line)
line = r.next()
line = r.next()
line = r.next()
line = r.next()
if not re.search(r'^@', line):
raise FileError("fastq文件格式错误")
myline2 = re.split('_', line)
if len(myline1) > 1 and len(myline2) > 1:
self.has_sample_info = True
return True
def check(self):
"""
检测文件是否满足要求,发生错误时应该触发FileError异常
:return:
"""
if super(HdrsFile, self).check():
if not self.check_line():
raise FileError("文件格式错误")
if self._has_fa:
if not self.check_consistence():
raise FileError("hdrs文件（{}）和他的fasta文件（{}）内容上不一致".format(self.path, self.fasta.path))
return True
def check_fa_index_coherence(self):
index_file_set = set(os.listdir(self.ref_seq_fa_index))
for fa_name in os.listdir(self.ref_single_seq_fa):
m_fa = re.match(r'^(\S+\.fa)$', fa_name)
if m_fa:
seq_name = m_fa.group(1)
ebwt1 = '{}.1.ebwt'.format(seq_name)
ebwt2 = '{}.2.ebwt'.format(seq_name)
ebwt3 = '{}.3.ebwt'.format(seq_name)
ebwt4 = '{}.4.ebwt'.format(seq_name)
rev_ebwt1 = '{}.rev.1.ebwt'.format(seq_name)
rev_ebwt2 = '{}.rev.2.ebwt'.format(seq_name)
condition = (ebwt1 in index_file_set) and (
ebwt2 in index_file_set
) and (ebwt3
in index_file_set) and (ebwt4 in index_file_set) and (
rev_ebwt1 in index_file_set) and (rev_ebwt2
in index_file_set)
if not condition:
raise FileError(
'file {} in the fasta dir {} does not have legal(enough) index files'
)
else:
raise RuntimeError(
'the file name {} in path {} is an illegal single seq fasta file name!'
.format(fa_name, self.ref_single_seq_fa))
self.step.add_steps('mapsplice_map')
self.step.mapsplice_map.start()
self.step.update()
def is_gzip(self):
"""
依据文件后缀名检测是是gz个是压缩文件
"""
if re.search('\.tar\.gz$', self.prop['path']):
raise FileError("不支持tar.gz格式文件")
if re.search('\.gz$', self.prop['path']) or re.search(
'\.gzip$', self.prop['path']):
return True
else:
return False
def _prepare(self, work_path):
"""
为获取序列的信息做准备
生成临时文件夹，当输入的文件是gz格式时，解压到tmp里
"""
self.unzipfile = self.prop['path']
basename = os.path.basename(self.prop['path'])
self.fastaname = os.path.join(work_path, basename + ".fasta")
if self.is_gz:
basename = re.search(r'(.+)\.gz', basename).group(1)
self.unzipfile = os.path.join(work_path, basename)
try:
subprocess.check_call('gunzip -c ' + self.prop['path'] + "> " +
self.unzipfile)
except subprocess.CalledProcessError:
raise FileError("非标准格式的gz文件！")
def check_line(self):
"""
检查每行是否满足hdrs的语法要求
:return:
"""
if super(HdrsFile, self).check():
try:
with open(self.path) as fr:
line = fr.readline().strip()
m = re.match(r">(\S+)\s+/len=(\d+)\s+/nonNlen=(\d+)\s+/org=(\S+)", line)
if not m:
raise FileError("hdrs文件（{}）的这一行：{}不满足要求".format(self.path, line))
except Exception as e:
raise Exception("打开hdrs文件运行出错: {}".format(e))
return True
def check_line_format(self):
'''
:return: hdrs文件里每行注释的id集合
'''
hdrs_abs_path = ""
co_fa_path = ""
id_set = set()
try:
with open(hdrs_abs_path) as hdrs:
line = hdrs.readline()
m = re.match(r'>(\S+)\s+/len=\d+\s+/nonNlen=\d+\s+/org=\S+',line.strip())
id = m.group(1)
id_set.add(id)
except Exception as e:
raise FileError("{}的hdrs文件：{}不符合要求".format(co_fa_path,hdrs_abs_path))
return id_set
def write_hdrs(self):
'''
生成fasta文件的.hdrs文件: 每一行的格式为">seq_id len=seq_length /nonNlen=non_N_seq_length /org=org_name"
供asprofile使用
:param hdrs_file_path:
:return: hdrs_file_path
author: linfang.jin
date:2017.01.15
'''
try:
fw = open(self.path, 'w')
org_name = os.path.basename(self.fasta.path).strip().split(".")[0]
for seq in SeqIO.parse(self.fasta.path, "fasta"):
seq_id = seq.id
seq_len = len(seq.seq)
nonNlen = seq_len - seq.seq.count("N") - seq.seq.count("n")
newline = ">{} len={} /nonNlen={} /org={}\n".format(str(seq_id), str(seq_len), str(nonNlen), org_name)
fw.write(newline)
fw.close()
except Exception as e:
raise FileError("{}:生成{}的hdrs文件({})过程中出错".format(e, self.fasta.path, self.path))
def set_fasta(self, fasta_obj):
if not isinstance(fasta_obj, FastaFile):
raise FileError("传入的文件不是FastaFile对象")
self._properties['co_fasta_file'] = fasta_obj
self._has_fa =True
return self