Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!asuvax!cs.utexas.edu!uunet!ksr!bill
From: bill@ksr.com (Bill Mann)
Newsgroups: comp.lang.perl
Subject: All-uppercase text to mixed case
Message-ID: <598@ksr.UUCP>
Date: 19 Jan 90 16:01:18 GMT
Sender: bill@ksr.UUCP
Organization: Kendall Square Research Corp.
Lines: 53

A few weeks ago Carl Hommel posted a request for a perl script to convert
all-uppercase messages to mixed case.  I editted his message slightly,
*converted it to uppercase*, and ran it through the perl script below.
Here's the result:

  A user asked for help in converting files from uppercase to lowercase, 
  and back again.  Of course, there are simple one-liners using tr that
  will do that.  However, i foolishly said that it would be easy to write
  a program to convert an all uppercase message, which often are sent by our
  novice users, to a more normal format.

  The algorithm i thought of is:
      o uppercase any alpha two spaces after a "."
      o uppercase any alpha at the beginning of a line, when there
        was a "." at the end of the previous line.

  I could brute force this, but that would be work, not fun.  I have twiddled
  with the following features of perl, but got nothing to work:
      o setting $/ = "\177" and reading in the whole file at once
      o using   if (/\.  (.)/) {  And $1.

  I really want an elegant regexp, combined with the tr command.  Any ideas?

  Carl hommel
  carlton@apollo.hp.com

Here's my perl script:

#!/usr/bin/perl

# This program copies its input to STDOUT, converting uppercase characters
# to lowercase according to a simple rule.

# RULE: a sentence begins with an alphanumeric and ends at the end of
# the input file or at the first terminator (period, question mark, or 
# exclamation point) which is followed by white space; all but the
# first character of each sentence is converted to lower case.

    $/ = "\177";        # do not split the input into lines
    $_ = <>;            # read the entire input

    s/(\w)(([^.?!]+|[.?!]+\S)+)/($a=$2)=~tr|A-Z|a-z|,$1.$a/eg;

    print;              # print the results


Notes:
  . The ($a=$2) is necessary, as modifying $2 directly does not work.
  . The description of $* in the perl manual seems to indicate it should 
    be set to 1, but the script works either way for this example.

Bill Mann
bill@ksr.com (ksr!bill@uunet.uu.net)