[Python-Dev] xreadlines : readlines :: xrange : range

[Skip Montanaro]

    Paul> I suggested this at one point but it was pointed out that there is
    Paul> probably a lot of code that works with the resulting list *as a
    Paul> list*

How about this idea?  What if readlines() was allowed to return a lazy
evaluator if a sizehint > 0 was given?  I only saw one example outside of
test cases in the current CVS tree where readlines(sizehint) was used
(Tools/idle/GrepDialog.py), and it used it as expected:

    while 1:
      block = f.readlines(sizehint)
      if not block:
        break
      for line in block:
        more stuff

My suspicion is that most uses of sizehint will be like this.  It hasn't
been around all that long in Python-years (since 1.5a2), so there's probably
not tons of code to break (I agree the semantics would change), and the
majority of code that uses it probably looks like the above, which is almost
safe (if it returned "" instead of an empty evaluator when nothing was left
to read it would be safe).  The advantage would be that the above could
become the more obvious

    for line in f.readlines(sizehint):
      more stuff

and the change to file reading code that is "too slow" becomes much simpler.
(Of course, xreadlines() has that advantage as well.)

I scanned my own code quickly.  I found about 10 uses with sizehint and 300
without.

I presume we are talking about 2.1 here.  In any case, it seems to me that
in Py3k readlines should be lazy.

Skip

P.S.  Why did FileInput class never grow a readlines method?