wyle@solaris.UUCP (Mitchell Wyle) (07/28/88)
Both Chris (the wiz) Torek <chris@mimsy.umd.edu>, and
Dave (the guru) Elliot (dce@mips.com) suggested using
IFS and case rather than sed. Using these constructs
saves 2 forks and an exec each time!! This savings is
about 50% of the time used by the script.
My heart-felt thanks to both Chris Torek and Dave Elliot.
The net is a great place to grock unix.
Here is the code now. Anyone care to make it even
faster? >> grin << perl? awk? ;-)
#!/bin/sh
#
# parse a range of numbers in the form:
#
# <range> :== <number>
# | <number>,range
# | <number>-<number>
# | <number>-
#
# <number>:== <digit><number>
#
# <digit> :== 0|1|2|3|4|5|6|7|8|9
#
# to create a list of numbers to send such as 1,3,8-11 -> 1 3 8 9 10 11
#
# input range is in variable $RA ($1 for testing)
RA=$1
R=""
echo in: $RA
SIFS="$IFS"
IFS=","
for i in $RA ; do
R=$R" "$i
done
RA=$R
IFS=$SIFS
R=""
for tok in $RA ; do
case $tok in
*-*)
SIFS="$IFS"
IFS='-'
set $tok
IFS="$SIFS"
i=$1
max=${2:-99}
while test $i -le $max ; do
R=$R" "$i
i=`expr $i + 1`
done ;;
*)
R=$R" "$tok ;;
esac
RA=$R
done
echo out: $RA
--
-Mitchell F. Wyle wyle@ethz.uucp
Institut fuer Informatik wyle%ifi.ethz.ch@relay.cs.net
ETH Zentrum
8092 Zuerich, Switzerland +41 1 256-5237morrell@hpsal2.HP.COM (Michael Morrell) (08/02/88)
/ hpsal2:comp.unix.wizards / wyle@solaris.UUCP (Mitchell Wyle) / 12:03 am Jul 28, 1988 /
Here is the code now. Anyone care to make it even
faster? >> grin << perl? awk? ;-)
[ code deleted ]
--
-Mitchell F. Wyle wyle@ethz.uucp
Institut fuer Informatik wyle%ifi.ethz.ch@relay.cs.net
ETH Zentrum
8092 Zuerich, Switzerland +41 1 256-5237
----------
Well, this is an awk version of this which I wrote some time back.
Note it also handles something of the form "-8" as a shorthand for "1-8".
This seemed to run significantly faster on my machine that any of the
posted shell scripts.
Michael Morrell
-------------------
#! /bin/sh
RA=$1
echo in: $RA
RA=`echo $RA | awk -F, -f parse.awk`
echo out: $RA
-------------------
where parse.awk contains
-------------------
{ for (i = 1; i <= NF; i++) {
j = index($i,"-");
if (j == 0)
printf "%d ",$i
else {
if (j == 1)
kmin = 1;
else
kmin = substr($i,1,j-1) + 0;
if (j == length($i))
kmax = 99;
else
kmax = substr($i,j+1) + 0;
for (k = kmin; k <= kmax; k++)
printf "%d ",k
}
}
}
-------------------leo@philmds.UUCP (Leo de Wit) (08/02/88)
In article <470@solaris.UUCP> wyle@solaris.UUCP (Mitchell Wyle) writes: |Both Chris (the wiz) Torek <chris@mimsy.umd.edu>, and |Dave (the guru) Elliot (dce@mips.com) suggested using |IFS and case rather than sed. Using these constructs |saves 2 forks and an exec each time!! This savings is |about 50% of the time used by the script. I don't see how you get to a number of 50%; it depends entirely on the parameters fed to the script. With long ranges the saving is only a fraction of the code. |My heart-felt thanks to both Chris Torek and Dave Elliot. |The net is a great place to grock unix. | |Here is the code now. Anyone care to make it even |faster? >> grin << perl? awk? ;-) C. But I don't think that was the original intention. If you plan to use only sh and 'the tools', you can however speed up the range generation considerably: | while test $i -le $max ; do | R=$R" "$i | i=`expr $i + 1` | done ;; First you can avoid the test command: do something like: if test $i -le $max then while : do R=$R" "$i i=`expr $i + 1` case $i in $max) break;; esac done fi Since case and : are builtin, this avoids a fork and exec for each number. Note the first test IS needed, because the program would loop forever (?) if $i was greater than $max already (can it be?). Second you could count with ten numbers 'at-a-time'; i and max will contain i/10 and max/10 respectively - the start and end have to be dealt with separately because they don't have to start with a number ending on 0 or end with a number ending on 9. The main part becomes: while : do R=$R" "${i}0" "${i}1" "${i}2" "${i}3" "${i}4" "${i}5" "${i}6" "${i}7" "${i}8" "${i}9 i=`expr $i + 1` case $i in $max) break;; esac done The details of the rest that is needed to make it run OK I leave up to you. Success! Leo.
wu@spot.Colorado.EDU (WU SHI-KUEI) (08/03/88)
In article <470@solaris.UUCP> wyle@solaris.UUCP (Mitchell Wyle) writes: >... Here is the code now. Anyone care to make it even >faster? >> grin << perl? awk? ;-) #! /bin/sh # # parse a range of numbers in the form: # # <range> :== <number> # | <number>,range # | <number>-<number> # | <number>- # # <number>:== <digit><number> # # <digit> :== 0|1|2|3|4|5|6|7|8|9 # # to create a list of numbers to send such as 1,3,8-11 -> 1 3 8 9 10 11 # # input range is in variable $RA ($1 for testing) Using a severe test like 1,2,3-8,11-15,34,35,127-199 for $1, the following awk program runs about 20 times faster on a 3B2/400 than Chris Torek's solution. ---------------------- cut here -------------------------------- echo "$1" | awk -F',' ' { for (i = 1; i <= NF; ++i) { if ((p = index($i, "-")) == 0) printf"%s ", $i else { start = substr($i, 0, p - 1) end = substr($i, p + 1) for (j = int(start); j <= int(end); ++j) printf"%s ", j } } printf"\n" }' ------------------------------------------------------------------ Just a guest on this machine. In real life Carl Brandauer daemon associates 1760 Sunset Blvd Boulder, CO 80302 303-442-1731 uunet!nbires!bdaemon!carl
fst@mcgp1.UUCP (Skip Tavakkolian) (08/03/88)
In article <470@solaris.UUCP>, wyle@solaris.UUCP (Mitchell Wyle) writes: > Here is the code now. Anyone care to make it even > faster? >> grin << perl? awk? ;-) > #!/bin/sh > # to create a list of numbers to send such as 1,3,8-11 -> 1 3 8 9 10 11 > RA=$1 > R="" > echo in: $RA > SIFS="$IFS" > IFS="," > for i in $RA ; do > R=$R" "$i > done > RA=$R > IFS=$SIFS > R="" > for tok in $RA ; do > case $tok in > *-*) > SIFS="$IFS" > IFS='-' > set $tok > IFS="$SIFS" > i=$1 > max=${2:-99} > while test $i -le $max ; do > R=$R" "$i > i=`expr $i + 1` > done ;; > *) > R=$R" "$tok ;; > esac > RA=$R > done > echo out: $RA > -- > -Mitchell F. Wyle wyle@ethz.uucp Here is an AWK script that does what you want and some time measurements as run on a 3b2-600 SYSV 3.1.1 using time(1) (I am using the old awk). awk ' END { # no error checking done here arg = "'$1'" if (length(arg) == 0) exit print "in: ", arg printf "out: " NF = split(arg, args, ",") for (i = 1 ; i <= NF ; i++) { if (split(args[i], range, "-") < 2) printf "%d ", args[i] else { for (j = range[1]; j <= range[2]; j++) printf "%d ", j } } print }' </dev/null Here are the time results: # SHELL version time range.sh 1,3,8-11,22,24-32 in: 1,3,8-11,22,24-32 out: 1 3 8 9 10 11 22 24 25 26 27 28 29 30 31 32 real 1.5 user 0.4 sys 0.8 # AWK version time range.awk 1,3,8-11,22,24-32 in: 1,3,8-11,22,24-32 out: 1 3 8 9 10 11 22 24 25 26 27 28 29 30 31 32 real 0.4 user 0.1 sys 0.1 Sincerely -- Fariborz ``Skip'' Tavakkolian UUCP ...!uw-beaver!tikal!mcgp1!fst UNIX is a registered trademark of AT&T