28 lines
1,017 B
Text
28 lines
1,017 B
Text
|
#!/bin/sh
|
||
|
|
usage()
|
||
|
|
{
|
||
|
|
printf 'usage: %s [file.lisp]\n' $0
|
||
|
|
exit 1
|
||
|
|
}
|
||
|
|
## The first program finds certain definitions and inserts a new blank
|
||
|
|
## line *before* the definition. Such action makes function
|
||
|
|
## definitions separated by two blank lines in some cases. We then
|
||
|
|
## remove the excess with the second program. Notice we need the -E
|
||
|
|
## option because we're using the | metacharacter that is only
|
||
|
|
## supported by popular sed programs with the -E option. This
|
||
|
|
## violates POSIX sed, but keep in mind that we only run this when
|
||
|
|
## releasing the package. This is a building tool, not part of the
|
||
|
|
## service.
|
||
|
|
sed -E '/^\(defun |\(defmacro /{
|
||
|
|
i\
|
||
|
|
|
||
|
|
}' $* | sed '/^[ \t]*$/{
|
||
|
|
N
|
||
|
|
/^[ \t]*\n$/D
|
||
|
|
}'
|
||
|
|
## We first find a blank line. Then we say N to expand the pattern
|
||
|
|
## space to include the next line. Then we delete the *first* blank
|
||
|
|
## line and not the second---that's what the D command does. This
|
||
|
|
## strategy is explained by Dale Dougherty and Arnold Robbins in ``sed
|
||
|
|
## & awk'' second edition, pages 112--114.
|