On Nov 17, 5:52 am, Tim Chase <[EMAIL PROTECTED]> wrote:
> >> Files are fixed format no field delimiters, fields are position and
> >> length records are terminated by newline.
>
> > Assuming no COMPUTATIONAL fields, it should be easy enough to split each
> > line up into fixed-length pieces, e.g. assuming a simple example
>
> > 01 Sample-Record.
> > 02 Field-1 pic XXX.
> > 02 Field-2 pic XXXX.
> > 02 Field-3 pic XXXXX.
>
> > then a Python sequence that read one line's worth of fields might be
>
> > line = infile.read()
> > (field_1, field_2, field_3) = (line[0:3], line[3:7], line[7:12])
>
> A recent posting on the list offered an elegant solution to this
> with a function something like
>
> def splitter(s, *lens):
> offset = 0
> pieces = []
> for length in lens:
> pieces.append(s[offset:offset+length])
> offset += length
> return pieces
>
> which could then be used to simplify that to
>
> (f1, f1, f3) = splitter(line, 3, 4, 5)
>
> It may not be quite so significant with just 3 items, but the OP
> mentioned having a large number of items in each record. One
> could even use something like a mapping for this. Something like:
>
> field_lens = [
> ("field1", 3),
> ("field2", 4),
> ("field3", 5),
> # stacks more
> ]
> fields = dict(zip(
> (name for name, _ in field_lens),
> splitter(line, *[length for _, length in field_lens])
> ))
> something = "Whatever %s you want" % fields["field2"]
>
> If you like this method, you can even make a more targeted
> splitter() function and add some function mappings like
>
> field_lens = [ # somewhat similar to your Cobol masks
> ("field1", 3, str),
> ("field2", 4, int),
> ("field3", 5, float),
> ("field4", 17, lambda s: s.strip().upper() ),
> # stacks more
> ]
>
> def splitter(s, field_lens):
> pieces = {}
> offset = 0
> for name, length, fn in field_lens:
> pieces[name] = fn(s[offset:offset+length])
> offset += length
> return pieces
>
> bits = splitter(line), field_lens)
> # do addition of int+float rather than string concat
> print bits["field2"] + bits["field3"]
>
> -tkc
Thank for your reply. I will have to study your code and
see if I can inprove mine.
Len
--
http://mail.python.org/mailman/listinfo/python-list
