Schwartzian Transform in Python?

Christian Tismer tismer at
Thu Apr 27 18:18:30 CEST 2000

Art Haas wrote:
> The python documentation doesn't make
> clear to me how to do things like "sort a list of lists by the second
> field in each sublist", which is what the Schwartzian transform is
> based on.

I think to remember an FAQ entry.
Here an (older) version of mine.
Not that today I would code that without
(left as an exercise to the reader)

You would use the below by supplying a metric like

sort(list, lambda x:x[1])


# this is taken from the FAQ and modified slightly:
4.51. I want to do a complicated sort: can you do a Schwartzian
     Transform in Python?

     Yes, and in Python you only have to write it once: 

      def st(List, Metric):
          def pairing(element, M = Metric):
                return (M(element), element)
          paired = map(pairing, List)
          return map(stripit, paired)

      def stripit(pair):
          return pair[1]

# now, this is my version:

def sort(List, Metric=None, stable=1) :
	"""sort a list by a given metric function. The sort defaults
	to be stable, keeping the order of elements which cannot be
	distinguished by the metric"""

	# transpose has problems with small lists, but in that
	# case we don't need it:
	if len(List) < 2: return List[:]
	shaker = [List]
	where = 0
	if Metric :
		# sorter is a list of function results
		shaker.insert(0, map(Metric, List))
		where = -1 # we are at the end
	if stable :
		# always stabilize behind the key
		shaker.insert(1, range(len(List)))
		# and we are still either at front or end
	shaker = transpose(shaker)
	return list(transpose(shaker)[where])

def transpose(x) :
	"""transpose a sequence of lists into a list of tuples. This
	is very fast, but doesn't work for len(x)==1"""
	return apply(map, (None,)+tuple(x))

# Special problem with fast transpose:
# Due to map, transpose of a list with a single element
# returns a list, shape is lost.

# Solution:

def transpose(x) :
	if len(x) == 1 : 
		return map(lambda y : (y,), x[0])
	return apply(map, (None,)+tuple(x))

# General problems with transpose (not solvable here):
# A list of empty tuples returns an empty list (shape lost)


ciao - chris

Christian Tismer             :^)   <mailto:tismer at>
Applied Biometrics GmbH      :     Have a break! Take a ride on Python's
Kaunstr. 26                  :    *Starship*
14163 Berlin                 :     PGP key ->
PGP Fingerprint       E182 71C7 1A9D 66E9 9D15  D3CC D4D7 93E2 1FAE F6DF
     where do you want to jump today?

More information about the Python-list mailing list