1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
|
# Copyright (C) 2006 Tobi Vollebregt
# This sed script can be used to convert double precision floating point
# constants in C/C++ code to single precision floating point constants.
# Run it with: sed -r --file=double_to_single_precision.sed -i foo.cpp bar.cpp
# Known quirks:
# * 1.0l is incorrectly replaced by 1.0fl.
# * Any number with a dot behind it is replaced, so also numbered lists in
# comments, copyright statements and numbers in strings (special measures
# have been taken against conversion of number on the beginning or end of
# a string though, because this was a common case in Spring) risk
# conversion to single precision.
# * Number on the beginning or end of a line aren't converted. This is no
# problem for C/C++ though as there normally are no numbers there.
# The command is put here twice because:
# * sed doesn't match the same regex multiple times, and
# * this regex looks at the char before and after the number too.
# Hence the 30.0 in (16.0/30.0) wouldn't be converted properly because the '/'
# would have been included in the first match already and it can't be matched
# a second time.
s/([^A-Z_0-9"])(((([0-9]*\.[0-9]+)|([0-9]+\.[0-9]*))([E][-+]?[0-9]+)?)|([0-9]+[E][-+]?[0-9]+))([^EF0-9"])/\1\2f\9/gi
s/([^A-Z_0-9"])(((([0-9]*\.[0-9]+)|([0-9]+\.[0-9]*))([E][-+]?[0-9]+)?)|([0-9]+[E][-+]?[0-9]+))([^EF0-9"])/\1\2f\9/gi
|
