def realfile_(self,hadoop,outdir,tmpdir):
m = re.search(r'\.(gz|bz2)',self.file)
if m:
if m.group(1) == 'gz':
hadoop.syscall('gunzip -c %s > %s' % (self.file,os.path.join(tmpdir,'ttable')))
else:
hadoop.syscall('bunzip2 -c %s > %s' % (self.file,os.path.join(tmpdir,'ttable')))
return os.path.join(tmpdir,'ttable')
else:
return self.file
def run(self,hadoop):
bname = []
output = self.output
if len(output) > 0:
bname = [os.path.basename(o) for o in output]
# bname and output frequently used together
opbn = zip(output, bname)
if len(output) > 0 and all([os.path.exists(o) for o in output]):
for op, bn in opbn:
hadoop.put(op, bn)
return
else:
hadoop.syscall(self.execute)
if len(output) > 0:
for op, bn in opbn:
if self.stage == 'training':
hadoop.getmerge(bn, op)
else:
hadoop.get(bn, op)
def run(self,hadoop,outdir,tmpdir):
bname = []
output_suffix = [o + self.suffix for o in self.output]
if len(self.output) > 0:
bname = [os.path.basename(o) for o in self.output]
# bname+suffix and output_suffix frequently used together
osbn = zip(output_suffix, [b + self.suffix for b in bname])
if len(self.output) > 0 and all([os.path.exists(o) for o in output_suffix]):
for os, bn in osbn:
hadoop.put(os, bn)
return
else:
file = self.realfile_(hadoop,outdir,tmpdir)
execute = PTemplate(self.execute).safe_substitute({'file':file,'suffix':self.suffix})
hadoop.syscall(execute)
if len(self.output) > 0:
for os, bn in osbn:
if self.stage == 'training':
hadoop.getmerge(bn, os)
else:
hadoop.get(bn, os)
self.cleanfile_(file,hadoop,outdir,tmpdir)
def cleanfile_(self,file,hadoop,outdir,tmpdir):
if file != self.file:
hadoop.syscall('rm -rf %s' % os.path.join(tmpdir,'ttable'))