Search code examples
pythoniteratorseek

Python: rewinding one line in file when iterating with f.next()


Python's f.tell doesn't work as I expected when you iterate over a file with f.next():

>>> f=open(".bash_profile", "r")
>>> f.tell()
0
>>> f.next()
"alias rm='rm -i'\n"
>>> f.tell()
397
>>> f.next()
"alias cp='cp -i'\n"
>>> f.tell()
397
>>> f.next()
"alias mv='mv -i'\n"
>>> f.tell()
397

Looks like it gives you the position of the buffer rather than the position of what you just got with next().

I've previously used the seek/tell trick to rewind one line when iterating over a file with readline(). Is there a way to rewind one line when using next()?


Solution

  • No. I would make an adapter that largely forwarded all calls, but kept a copy of the last line when you did next and then let you call a different method to make that line pop out again.

    I would actually make the adapter be an adapter that could wrap any iterable instead of a wrapper for file because that sounds like it would be frequently useful in other contexts.

    Alex's suggestion of using the itertools.tee adapter also works, but I think writing your own iterator adapter to handle this case in general would be cleaner.

    Here is an example:

    class rewindable_iterator(object):
        not_started = object()
    
        def __init__(self, iterator):
            self._iter = iter(iterator)
            self._use_save = False
            self._save = self.not_started
    
        def __iter__(self):
            return self
    
        def next(self):
            if self._use_save:
                self._use_save = False
            else:
                self._save = self._iter.next()
            return self._save
    
        def backup(self):
            if self._use_save:
                raise RuntimeError("Tried to backup more than one step.")
            elif self._save is self.not_started:
                raise RuntimeError("Can't backup past the beginning.")
            self._use_save = True
    
    
    fiter = rewindable_iterator(file('file.txt', 'r'))
    for line in fiter:
        result = process_line(line)
        if result is DoOver:
            fiter.backup()
    

    This wouldn't be too hard to extend into something that allowed you to backup by more than just one value.