fred@maccs.dcss.mcmaster.ca (Fred Whiteside) (08/30/90)
i have a simple little shell question. I am attempting
to write a script that will perform various manipulations on news
articles. my problem seems to be with understanding variable
substitution in sh. a minimal example is the enclosed sh script. what
it should do is set the variable files to be the names of those
articles in comp.std.c whose name (article number) is greater than or
equal to the numeric value of the scripts first argument. this does
not behave as expected. i have tried quite a unmber of variations on
this theme (mostly centering around multiple $'s and interesting
quoting of commands, but i tend to get all of the files or none of them
depending on (for some reason) the number of $'s preceding the zots
variable name (also on the number in front of the 1, $1 or $$1, etc.
but i expected that ...) this is running under sunos4.1 if that is
interesting. i have tried to read the manual, but with little
success. anyone interested in pointing out my clear stupidity
publicly? i *would* appreciate it ....
many thanks ...
#!/bin/sh
cd /usr/spool/news/comp/std/c
zots=$1
echo "Should ignore files with name < $zots"
files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}`
echo $files
--
Fred Whiteside Chedoke-McMaster Hospitals, Hamilton, Ontario, Canada
whitesid@{mcmaster.ca,darwin.cmh.mcmaster.ca,130.113.2.18}
fred@maccs.DCSS.McMaster.CA ...!uunet!utai!utgpu!maccs!fredjtp@ngs.fi (Jukka Partanen) (08/30/90)
In article <26DC6447.15922@maccs.dcss.mcmaster.ca> fred@maccs.dcss.mcmaster.ca (Fred Whiteside) writes: >files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` files=`ls -rt | grep -v '.*[a-zA-Z].*' | awk '$1 >= '$zots' {print}` -- A mountain is something you don't wanna fuck with.
jtp@ngs.fi (Jukka Partanen) (08/30/90)
In article <JTP.90Aug30141736@ngs.fi> jtp@ngs.fi (Jukka Partanen) writes: Argh, forgot one ' >files=`ls -rt | grep -v '.*[a-zA-Z].*' | awk '$1 >= '$zots' {print}` files=`ls -rt | grep -v '.*[a-zA-Z].*' | awk '$1 >= '$zots' {print}'` -- A mountain is something you don't wanna fuck with.
jeff@quark.WV.TEK.COM (Jeff Beadles) (08/30/90)
fred@maccs.dcss.mcmaster.ca (Fred Whiteside) writes: :> :> i have a simple little shell question. I am attempting :>to write a script that will perform various manipulations on news :>articles. my problem seems to be with understanding variable :>substitution in sh. a minimal example is the enclosed sh script. what :>it should do is set the variable files to be the names of those :>articles in comp.std.c whose name (article number) is greater than or :>equal to the numeric value of the scripts first argument. this does :>not behave as expected. i have tried quite a unmber of variations on :>this theme (mostly centering around multiple $'s and interesting :>quoting of commands, but i tend to get all of the files or none of them :>depending on (for some reason) the number of $'s preceding the zots :>variable name (also on the number in front of the 1, $1 or $$1, etc. :>but i expected that ...) this is running under sunos4.1 if that is :>interesting. i have tried to read the manual, but with little :>success. anyone interested in pointing out my clear stupidity :>publicly? i *would* appreciate it .... :> :> many thanks ... :> :>#!/bin/sh :>cd /usr/spool/news/comp/std/c :>zots=$1 :>echo "Should ignore files with name < $zots" :>files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` I see a quoting problem here ^^^^^^^^^^^^^^^^^^^^^ :>echo $files :> :>-- :>Fred Whiteside Chedoke-McMaster Hospitals, Hamilton, Ontario, Canada :>whitesid@{mcmaster.ca,darwin.cmh.mcmaster.ca,130.113.2.18} :>fred@maccs.DCSS.McMaster.CA ...!uunet!utai!utgpu!maccs!fred >files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` This line is the problem, replace it with: files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk "\$1 >= $zots {print}"` Or, to make it a little more simpler... (You don't need .*'s in the grep) files=`ls -rt * |grep -v '[a-zA-Z]' | awk "\$1 >= $zots {print}"` To explain: The only real change was to change thw quoting around the awk script. $1 was getting expanded to the shell's $1, not the 1st awk field. By using " as a quoting character, the shell is able to expand $zot to whatever number it represents. The \$1 is to prevent the shell from trying to expand $1 to the shell's $1, rather than the 1st awk field. After the shell is finished with variable substitution, awk "\$1 >= $zots {print}" is transformed, and awk see's: (if zot = 10) awk "$1 >= 10 {print}" Hope this helps! A decent book on shell programming and quoting is the Howard Sam's book on shell programming. -Jeff -- Jeff Beadles jeff@quark.WV.TEK.COM Utek Engineering, Tektronix Inc. +1 503 685 2568 SPEEA - Just say no.
scott@tab00.larc.nasa.gov (Scott Yelich) (08/30/90)
>:>files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` >I see a quoting problem here ^^^^^^^^^^^^^^^^^^^^^ >files=`ls -rt * |grep -v '[a-zA-Z]' | awk "\$1 >= $zots {print}"` >To explain: > The only real change was to change thw quoting around the awk script. >$1 was getting expanded to the shell's $1, not the 1st awk field. > By using " as a quoting character, the shell is able to expand $zot >to whatever number it represents. The \$1 is to prevent the shell >from trying to expand $1 to the shell's $1, rather than the 1st awk >field. I still see a quoting problem here... At least, when I try this-- it doesn't work. I have only been able to get awk to work properly if I use single quotes and not double quotes. As a side question, does ANYONE have any bourne shell routines which do math... reasonably effeciently? (For numbers > 1000?) -- Signature follows. [Skip now] ----------------------------------------------------------------------------- Scott D. Yelich scott@[xanth.]cs.odu.edu [128.82.8.1] After he pushed me off the cliff, he asked me, as I fell, ``Why'd you jump?'' Administrator of: Game-Design requests to <game-design-request@cs.odu.edu> ODU/UNIX/BSD/X/C/ROOT/XANTH/CS/VSVN/ -----------------------------------------------------------------------------
brister@decwrl.dec.com (James Brister) (08/30/90)
On 30 Aug 90 00:56:38 GMT, fred@maccs.dcss.mcmaster.ca (Fred Whiteside) said: > #!/bin/sh > cd /usr/spool/news/comp/std/c > zots=$1 > echo "Should ignore files with name < $zots" > files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` > echo $files I changed the next to last line to this: files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk '$1 >= '$zots' {print}'` Awk expects its program to be one argument. You've given it four. You DON'T want the $1 to be expanded, but you DO want the $zots to be expanded, hence the funny quoting. James -- James Brister brister@decwrl.dec.com DEC Western Software Lab., Palo Alto, California. .....!decwrl!brister
jeff@quark.WV.TEK.COM (Jeff Beadles) (08/31/90)
scott@tab00.larc.nasa.gov (Scott Yelich) writes: >As a side question, does ANYONE have any bourne shell routines which do >math... reasonably effeciently? (For numbers > 1000?) Well, I use expr. For example: a=123456 b=123 c=`expr $a - $b` d=`expr $a \* $b` echo "c = $c" echo "d = $d" Note, that the '*' MUST be quoted, as the shell will expand it. "Shell qouting -- Know it, live it, be it. -Jeff" :-) -Jeff -- Jeff Beadles jeff@quark.WV.TEK.COM Utek Engineering, Tektronix Inc. +1 503 685 2568 SPEEA - Just say no.
mvadh@cbnews.att.com (andrew.d.hay) (08/31/90)
In article <26DC6447.15922@maccs.dcss.mcmaster.ca>, fred@maccs.dcss.mcmaster.ca (Fred Whiteside) writes:
[]
" what
" it should do is set the variable files to be the names of those
" articles in comp.std.c whose name (article number) is greater than or
" equal to the numeric value of the scripts first argument.
[]
"
" #!/bin/sh
" cd /usr/spool/news/comp/std/c
" zots=$1
" echo "Should ignore files with name < $zots"
" files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}`
" echo $files
how about:
#!/bin/sh
cd /usr/spool/news/comp/std/c
echo "Should ignore files with name < $1 (formerly $zots)"
for TMP in `echo * | grep -v '[a-zA-Z]'`; do if [ "$TMP" -ge $1 ]; then echo "$TMP"; fi; done
or, if you -need- to create a variable:
files=`ls -rt | grep -v '[a-zA-Z]' | while read TMP; do if [ "$TMP" -ge $1 ]; then echo "$TMP"; fi; done`
echo $files
these should also be much faster, as it relies more on shell builtins.
i'm not familiar with awk (shame on me!), but i don't see why you need
it for this...
--
Andrew Hay +------------------------------------------------------+
Ragged Individualist | You just have _N_O idea! It's the difference |
AT&T-BL Ward Hill MA | between _S_H_O_O_T_I_N_G a bullet and _T_H_R_O_W_I_N_G it! |
a.d.hay@att.com +------------------------------------------------------+mercer@npdiss1.StPaul.NCR.COM (Dan Mercer) (09/01/90)
In article <8465@orca.wv.tek.com> jeff@quark.WV.TEK.COM (Jeff Beadles) writes: :scott@tab00.larc.nasa.gov (Scott Yelich) writes: : :>As a side question, does ANYONE have any bourne shell routines which do :>math... reasonably effeciently? (For numbers > 1000?) : :Well, I use expr. For example: : :a=123456 :b=123 : :c=`expr $a - $b` : :d=`expr $a \* $b` : :echo "c = $c" :echo "d = $d" : :Note, that the '*' MUST be quoted, as the shell will expand it. : :"Shell qouting -- Know it, live it, be it. -Jeff" :-) : : -Jeff :-- :Jeff Beadles jeff@quark.WV.TEK.COM :Utek Engineering, Tektronix Inc. +1 503 685 2568 : SPEEA - Just say no. If your system has named pipes - fifos - and bc then you may find this useful: # # these instructions illustrate how to set up a background # engine in a shell script - a simple application is shown only # for illustration purposes - it numbers lines of standard input # first allocate fifos ITMP=/usr/tmp/i.$LOGNAME OTMP=/usr/tmp/o.$LOGNAME /bin/rm -f $ITMP $OTMP 2>/dev/null /etc/mknod $ITMP p # input fifo for engine /etc/mknod $OTMP p # output fifo for engine # start background engine and open $ITMP as file descriptor 3 # and $OTMP as file descriptor 4 bc <$ITMP >$OTMP & exec 3>$ITMP exec 4<$OTMP # send command to background engine # if background engine is bc, counters can be maintained # echo "x = 1" # echo "x = x + 1" or echo "++x" #the second version returns the new value of x echo "x = 0" >&3 while line=`line` do echo "x++" >&3 # read response read ans <&4 echo "$ans: $line" done < engine # terminate background engine and remove fifos echo "quit" >&3 /bin/rm -f $ITMP $OTMP exit 0 -- Dan Mercer Reply-To: mercer@npdiss1.StPaul.NCR.COM (Dan Mercer) "MAN - the only one word oxymoron in the English Language"
smithln@honeydew.cs.rochester.edu (Neil Smithline) (09/01/90)
In article <26DC6447.15922@maccs.dcss.mcmaster.ca>, fred@maccs (Fred Whiteside) writes: > >... > >#!/bin/sh >cd /usr/spool/news/comp/std/c >zots=$1 >echo "Should ignore files with name < $zots" >files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` >echo $files > The best way to figure this out is to go into the shell and start typing. To find the problem, run /bin/sh and go into the correct directory. Once there try the command ls -rt * and see if this produces the correct output - if it does see what happens if you add the grep. If this works try the awk. You'll see that this causes awk to bail out. Even if you reduce the awk statement to awk $1 '>=' 1 {print} awk still gives you an error message. From this you can conclude that this has nothing to do with the rest of the line and resort to the old awk man page. If you quote the arguments, all works fine. That is: awk '$1 >= 1 {$print}' works. Now to add the shell variable $zots to this just insert it *unqoted* into the line zots=1 awk '$1 >= '$zots' {print}' Once you have this just replace the awk statement in your program with the above line and all works properly. In general, I find that complex shell quoting is best solved with an inside-out approach, get the inside to work and then add the next level of quotes to that - of course, as you get better you need to do less and less of this - Neil -- ======================================================================== Neil Smithline ARPA: smithln@cs.rochester.edu UUCP: ..!rutgers!rochester!smithln
roger@kymijoki.wrs.com (Roger Rohrbach) (09/01/90)
fred@maccs.dcss.mcmaster.ca (Fred Whiteside) writes: >cd /usr/spool/news/comp/std/c >zots=$1 >echo "Should ignore files with name < $zots" >files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` >echo $files There are two problems: first, awk expects its first argument to be a an awk program; you are giving the program as 4 separate arguments ($1, '>=', $zots, {print}), so your program currently will cause awk to bail out. Second, you are not escaping the "$1" in your awk program, so the shell is substituting the first parameter to your script there (i.e., awk is seeing the number you gave as an argument, not "$1"). If you change the last part of your pipeline to awk '$1 >= '$zots' {print}' it will work, because the "$1" is quoted for awk's sake, and all white space is quoted to prevent the shell from passing 4 strings, not 1, to awk as arguments. Note that we step out of the quotes long enough to let the shell substitute for "$zots". ADVANCED DISCUSSION: One might think you could use double quotes around the whole awk program, and just escape the "$1" but not "$zots", thus: awk "\$1 >= $zots {print}" and indeed this would work if it were not for the fact that the entire pipeline is enclosed in backquotes. A peculiarity of backquoting is that it forces an extra level of evaluation, so "\$1" is no safer than "$1". You must use awk "\\$1 >= $zots {print}" to get this to work. Another subtlety: if your program were not doing a numeric comparison, you'd have to arrange for "$zots" to be presented to awk as a string. My first example would become something like awk '$1 == "'$zots'" {print}' and the second, awk "\\$1 == \"$zots\" {print}" Roger Rohrbach sun!wrs!roger roger@wrs.com - Eddie sez: ----------------------------------------------- (c) 1986, 1990 -. | {o >o | | \ -) "I was an infinitely hot and dense dot." |
wrp@PRC.Unisys.COM (William R. Pringle) (09/01/90)
In article <1209@wrs.wrs.com> roger@kymijoki.wrs.com (Roger Rohrbach) writes: >fred@maccs.dcss.mcmaster.ca (Fred Whiteside) writes: > >>cd /usr/spool/news/comp/std/c >>zots=$1 >>echo "Should ignore files with name < $zots" >>files=`ls -rt * |grep -v '.*[a-zA-Z].*' | awk $1 '>=' $zots {print}` >>echo $files > > (Followed by several variations on how to arrange quotes so that a shell variable can get into an AWK program.) I'm surprised no one mentioned the simple way: awk '$1 >= MAX{print}' MAX=$zots - The MAX=value sets the awk variable MAX to 'value', the dash is needed to indicate using stdin. You could eliminate the grep by letting awk do it for you: awk '/^[0-9]*/{if ($1 >= MAX) print}' MAX=$zots - No that we have probably confused the original poster beyond hope ... Good luck! Bill Pringle The nice thing about Unix is that there are so many right answers!
mday@iconsys (Matt Day) (09/02/90)
In article <SCOTT.90Aug30124930@tab00.larc.nasa.gov> scott@tab00.larc.nasa.gov (Scott Yelich) writes: >As a side question, does ANYONE have any bourne shell routines which do >math... reasonably effeciently? (For numbers > 1000?) For the Bourne shell, the most efficient way to do math is to use the "expr" program. It doesn't support floating point, but it doesn't have any trouble with numbers larger than 1000. If you have a shell script that needs to do a lot of math, I suggest you use the C shell built in math operators; it's much faster (or perl, of course). -- - Matthew T. Day, Sanyo/ICON, mday@iconsys.icon.com || uunet!iconsys!mday
george@hls0.hls.oz (George Turczynski) (09/05/90)
In article <SCOTT.90Aug30124930@tab00.larc.nasa.gov> scott@tab00.larc.nasa.gov (Scott Yelich) writes: >As a side question, does ANYONE have any bourne shell routines which do >math... reasonably effeciently? (For numbers > 1000?) I don't know whose UNIX you have but you may have a utility called "dc", for "desk calulator". If you have it, read the man page. It will handle numbers > 1000 AND numbers with decimal fractions. for example: |>---------Cut here----------<| #! /bin/sh num="10000.0" while [ $num != "10010.0" ]; do num=`echo "$num 0.2 + p" | dc` echo $num done |>---------Cut here----------<| Just another idea... -- | George P. J. Turczynski. |---------------------------------------------------- | Computer Systems Engineer. | ACSnet: george@highland.oz | I can't speak for the | | Highland Logic Pty. Ltd. | Phone: +61 48 683490 | company, I can barely | | Suite 1, 348-354 Argyle St | Fax: +61 48 683474 | speak for myself... | | Moss Vale. NSW. Australia. 2577 |----------------------------------------------------
scott@tab29.larc.nasa.gov (Scott Yelich) (09/05/90)
>In article <SCOTT.90Aug30124930@tab00.larc.nasa.gov> scott@tab00.larc.nasa.gov (Scott Yelich) writes: >>As a side question, does ANYONE have any bourne shell routines which do ^^^^^^ ^^^^^ ^^^^^^^^ |||||| ||||| |||||||| >>math... reasonably effeciently? (For numbers > 1000?) >I don't know whose UNIX you have but you may have a utility called >"dc", for "desk calulator". If you have it, read the man page. It >will handle numbers > 1000 AND numbers with decimal fractions. > for example: I'D RATHER USE PERL! (Thas'a joke, son... I could'a had'a V8! or at least a V5r4!) Sometimes people program for the ART of it (Oh no! Did I say that?) for instance, which do you prefer? Which do you consider better form? 1) if [ "$1" = "$2" ]; then echo "Yeas!" else echo "No-way!" fi - or - 2) test "$1" = "$2" && echo "Yeas!" || echo "No-way!" What about? 3) if [ "$1" ]; then echo "Yup!" fi 4) eval ${1:+'echo "Yup!"'} I prefer 2 and 4.... Now, does anyone have and BOURNE SHELL (ie: /bin/sh) routines to do math? what about getting information like ``ls -Flagsi filename'' without calling ls? Please: no perl, exec, c, pascal, fortran, csh, ksh, etc. /bin/sh -- Signature follows. [Skip now] ----------------------------------------------------------------------------- Scott D. Yelich scott@[xanth.]cs.odu.edu [128.82.8.1] After he pushed me off the cliff, he asked me, as I fell, ``Why'd you jump?'' Administrator of: Game-Design requests to <game-design-request@cs.odu.edu> ODU/UNIX/BSD/X/C/ROOT/XANTH/CS/VSVN/ -----------------------------------------------------------------------------
de5@de5.CTD.ORNL.GOV (Dave Sill) (09/06/90)
In article <SCOTT.90Sep5124415@tab29.larc.nasa.gov>, scott@tab29.larc.nasa.gov (Scott Yelich) writes: > >2) test "$1" = "$2" && echo "Yeas!" || echo "No-way!" > >4) eval ${1:+'echo "Yup!"'} > >I prefer 2 and 4.... I hope I never have to maintain any of your scripts. -- Dave Sill (de5@ornl.gov) These are my opinions. Martin Marietta Energy Systems Workstation Support
fuchs@it.uka.de (Harald Fuchs) (09/06/90)
scott@tab29.larc.nasa.gov (Scott Yelich) writes: >2) test "$1" = "$2" && echo "Yeas!" || echo "No-way!" >I prefer 2 and 4.... I don't. The third command is executed if the second fails, which is surely not what I want. -- Harald Fuchs <fuchs@it.uka.de>
eliot@dg-rtp.dg.com (Topher Eliot) (09/06/90)
In article <SCOTT.90Sep5124415@tab29.larc.nasa.gov>, scott@tab29.larc.nasa.gov (Scott Yelich) writes: |> |> Sometimes people program for the ART of it (Oh no! Did I say that?) |> |> for instance, which do you prefer? Which do you consider better form? ... |> 2) test "$1" = "$2" && echo "Yeas!" || echo "No-way!" ... |> 4) eval ${1:+'echo "Yup!"'} |> I prefer 2 and 4.... My wife and I have for a long time had an inside joke: "I can't understand it. It must be Art." I never thought I'd apply it to a shell script... -- Topher Eliot Data General Corporation eliot@dg-rtp.dg.com 62 T. W. Alexander Drive {backbone}!mcnc!rti!dg-rtp!eliot Research Triangle Park, NC 27709 (919) 248-6371 Obviously, I speak for myself, not for DG.
chet@cwns1.CWRU.EDU (Chet Ramey) (09/06/90)
In article <SCOTT.90Sep5124415@tab29.larc.nasa.gov> scott@tab29.larc.nasa.gov (Scott Yelich) writes: >1) if [ "$1" = "$2" ]; then > echo "Yeas!" > else > echo "No-way!" > fi What if $1 = '=' or '-t'? >2) test "$1" = "$2" && echo "Yeas!" || echo "No-way!" cwns1$ foo() > { > /bin/test "$1" = "$2" && echo "Yeas" || echo "No-way" > } cwns1$ foo -r -r test: too many arguments No-way >What about? > >3) if [ "$1" ]; then > echo "Yup!" > fi What if $1 = `-f' or `-r'? cwns1$ foo() > { > if [ "$1" ] ; then > echo yup > fi > } cwns1$ foo a yup cwns1$ foo -r [: argument expected >4) eval ${1:+'echo "Yup!"'} This *is* a nice way to test whether or not a variable is set. Nobody seems to consider the case statement, which avoids these problems: equal() { case "$1" in "$2") return 0 ;; *) return 1 ;; esac } cwns1$ equal a b cwns1$ echo $? 1 cwns1$ equal a a cwns1$ echo $? 0 cwns1$ equal -r -r cwns1$ echo $? 0 >Now, does anyone have and BOURNE SHELL (ie: /bin/sh) routines to do math? Well, the shell running as /bin/sh on my machine is bash, and it has $[] to do arithmetic substitution. Chet -- Chet Ramey ``Levi Stubbs' tears run down Network Services Group his face...'' Case Western Reserve University chet@ins.CWRU.Edu
scott@tab00.larc.nasa.gov (Scott Yelich) (09/06/90)
|> Sometimes people program for the ART of it (Oh no! Did I say that?) |> for instance, which do you prefer? Which do you consider better My wife and I have for a long time had an inside joke: "I can't understand it. It must be Art." I never thought I'd apply it to a shell script... Please read the charter [?] for this group! I believe programming is an art... and if you program in sh, then that can be an art too. ps: re: test "" && {} || {} I am not worried about the second command failing. -- Signature follows. [Skip now] ----------------------------------------------------------------------------- Scott D. Yelich scott@[xanth.]cs.odu.edu [128.82.8.1] After he pushed me off the cliff, he asked me, as I fell, ``Why'd you jump?'' Administrator of: Game-Design requests to <game-design-request@cs.odu.edu> ODU/UNIX/BSD/X/C/ROOT/XANTH/CS/VSVN/ -----------------------------------------------------------------------------
scott@tab00.larc.nasa.gov (Scott Yelich) (09/06/90)
>cwns1$ foo -r -r >test: too many arguments >cwns1$ foo -r >[: argument expected > >4) eval ${1:+'echo "Yup!"'} Right, if I am at all worried about a variable evr starting with a dash, I do something as ugly as: test "${var:+:}" See, most ALL of my internal variables I SET and they will never be -something. And if they ever get to be -something then the value is invalid. >This *is* a nice way to test whether or not a variable is set. I use it for cases when there MIGHT be a -something. For instance, I have special routines which parse command line options. > Nobody seems to consider the case statement, which avoids these problems: > "$2") return 0 > *) return 1 Yeah, perhaps test should just be smarter, but we can't wait for that. >>Now, does anyone have and BOURNE SHELL (ie: /bin/sh) routines to do math? >Well, the shell running as /bin/sh on my machine is bash, and it has >$[] to do arithmetic substitution. I am looking for ART styles here... not shell substitutions! Arrraugh! -- Signature follows. [Skip now] ----------------------------------------------------------------------------- Scott D. Yelich scott@[xanth.]cs.odu.edu [128.82.8.1] After he pushed me off the cliff, he asked me, as I fell, ``Why'd you jump?'' Administrator of: Game-Design requests to <game-design-request@cs.odu.edu> ODU/UNIX/BSD/X/C/ROOT/XANTH/CS/VSVN/ -----------------------------------------------------------------------------
meissner@osf.org (Michael Meissner) (09/06/90)
In article <1990Sep6.141133.3391@usenet.ins.cwru.edu> chet@cwns1.CWRU.EDU (Chet Ramey) writes: | In article <SCOTT.90Sep5124415@tab29.larc.nasa.gov> | scott@tab29.larc.nasa.gov (Scott Yelich) writes: | | >1) if [ "$1" = "$2" ]; then | > echo "Yeas!" | > else | > echo "No-way!" | > fi | | What if $1 = '=' or '-t'? The standard way of dealing with this is: if [ x"$1" = x"$2" ]; then echo "Yeas!" else echo "No-way!" fi -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Do apple growers tell their kids money doesn't grow on bushes?
goudreau@dg-rtp.dg.com (Bob Goudreau) (09/08/90)
In article <MEISSNER.90Sep6122432@osf.osf.org>, meissner@osf.org (Michael Meissner) writes: > In article <1990Sep6.141133.3391@usenet.ins.cwru.edu> > chet@cwns1.CWRU.EDU (Chet Ramey) writes: > > | In article <SCOTT.90Sep5124415@tab29.larc.nasa.gov> > | scott@tab29.larc.nasa.gov (Scott Yelich) writes: > | > | >1) if [ "$1" = "$2" ]; then > | > echo "Yeas!" > | > else > | > echo "No-way!" > | > fi > | > | What if $1 = '=' or '-t'? > > The standard way of dealing with this is: > > if [ x"$1" = x"$2" ]; then > echo "Yeas!" > else > echo "No-way!" > fi Why is even the "x" method necessary? My system's Bourne shell (a port of V.3.2's sh) doesn't have any problem handling '=' or '-t' in the example Scott Yelich describes above. Is this something AT&T fixed in the last couple of years? ---------------------------------------------------------------------- Bob Goudreau +1 919 248 6231 Data General Corporation 62 Alexander Drive goudreau@dg-rtp.dg.com Research Triangle Park, NC 27709 ...!mcnc!rti!xyzzy!goudreau USA
goudreau@dg-rtp.dg.com (Bob Goudreau) (09/08/90)
In article <SCOTT.90Sep6120011@tab00.larc.nasa.gov>, scott@tab00.larc.nasa.gov (Scott Yelich) writes: > > |> Sometimes people program for the ART of it (Oh no! Did I say that?) > |> for instance, which do you prefer? Which do you consider better > > My wife and I have for a long time had an inside joke: "I can't > understand > it. It must be Art." I never thought I'd apply it to a shell > script... > > Please read the charter [?] for this group! > > I believe programming is an art... and if you program in sh, then > that can be an art too. So let me get this straight: Making shell scripts needlessly obscure (i.e., no increase in either functionality or performance, and a downright *decrease* in readability) is supposed to be a *good thing* ??!! Call it "art" if you will, and feel free to start a rec.arts.shell newsgroup. I prefer to call it "bad software engineering". ---------------------------------------------------------------------- Bob Goudreau +1 919 248 6231 Data General Corporation 62 Alexander Drive goudreau@dg-rtp.dg.com Research Triangle Park, NC 27709 ...!mcnc!rti!xyzzy!goudreau USA
meissner@osf.org (Michael Meissner) (09/10/90)
In article <1990Sep7.193010.26787@dg-rtp.dg.com> goudreau@dg-rtp.dg.com (Bob Goudreau) writes: | In article <MEISSNER.90Sep6122432@osf.osf.org>, meissner@osf.org | (Michael Meissner) writes: ... | > The standard way of dealing with this is: | > | > if [ x"$1" = x"$2" ]; then | > echo "Yeas!" | > else | > echo "No-way!" | > fi | | Why is even the "x" method necessary? My system's Bourne shell | (a port of V.3.2's sh) doesn't have any problem handling '=' or | '-t' in the example Scott Yelich describes above. Is this something | AT&T fixed in the last couple of years? It depends on whether you are trying to write maximally portable shell scripts. V7-based systems required this, because test was just another command, and not handled by the shell. BSD can't include the system V.2 shell (or anything else) because it would mean that current BSD users would have to get a System V.2 license (which I don't think is offered anymore). A System V.3 license is on the order of $60,000 for commercial sites, and V.4 is rumored to be $100,000 (I know there are provisions for lower university prices, but I don't have any data on them). Also, the 'recent' AT&T licenses have had things various people have objected to in the past (and to be fair there are people who can't sign the latest BSD licenses, and I'm sure there are people who have problems with OSF licenses). -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Do apple growers tell their kids money doesn't grow on bushes?
roger@kymijoki.wrs.com (Roger Rohrbach) (09/11/90)
goudreau@dg-rtp.dg.com (Bob Goudreau) writes: >So let me get this straight: > >Making shell scripts needlessly obscure (i.e., no increase in either >functionality or performance, and a downright *decrease* in >readability) is supposed to be a *good thing* ??!! > >Call it "art" if you will, and feel free to start a rec.arts.shell >newsgroup. I prefer to call it "bad software engineering". I agree that shell programs should be readable, but "needlessly obscure" is subjective. On the one hand, the shell syntax (outside of the control structures) is pretty obscure to begin with, because it relies so much on context-sensitive treatment of single characters. On the other hand, the shell language per se is small enough that I don't consider certain combinations to be obscure, even though they are not used by most casual shell programmers; i.e., I believe it is our responsibility to learn the language. I alway try to write in a way that minimizes obscurity intro- duced by the former. Thus, I don't like to write complicated pipelines that go in and out of single- and double-quotes to pass shell variables and backquote expressions to sed scripts! But I do use goal-directed programming (program components that "succeed" or "fail") extensively, even though people not familiar with the style (including my boss!) call it "obscure"- I tell them to look at Prolog or Icon, or even the C `?:' operator (my boss only knows C :-) If a language provides a succinct notation, I think it's good to consider the use of same "succinct", not "obscure". As an example, I prefer grep '^roger:' /etc/passwd > /dev/null || { echo "roger is not a user" >&2 exit 1 } to if grep '^roger:' /etc/passwd > /dev/null then : else echo "roger is not a user" >&2 exit 1 fi and would consider it *more* readable, especially if it were nested within another "if-then-else" statement or loop. Roger Rohrbach sun!wrs!roger roger@wrs.com - Eddie sez: ----------------------------------------------- (c) 1986, 1990 -. | {o >o | | \ -) "I was an infinitely hot and dense dot." |
fred@maccs.dcss.mcmaster.ca (Fred Whiteside) (09/18/90)
hi,
I just wanted to say thanks to all those who responded to my
question about the bourne shell. Almost everyone gently pointed out
that i was improperly quoting the (single) argument to awk and most
tried to help me with the efficiency of the code. Again, thanks to
one and all, including:
Sam Horrocks <sam@buckaroo.ICS.UCI.EDU>
rossc@extro.ucc.su.OZ.AU (Ross Cartlidge)
tale@turing.cs.rpi.edu (David C Lawrence)
peter@ontmoh
Peter G. Ford <pgf@space.mit.edu>
eirik@theory.TN.CORNELL.EDU (Eirik Fuller)
rtc%westford.ccur.com@RELAY.CS.NET
shirono@gcx1.SSD.CSD.HARRIS.COM (Roberto Shironoshita)
Kenneth Herron <kherron@ms.uky.edu>
Richard Frost <rfrost@spam.ua.oz.au>
Atte Kortekangas <ajk@tik.vtt.fi>
Jeff Beadles <jeff%quark.wv.tek.com@RELAY.CS.NET>
Bob English <renglish@hplabsz.hpl.hp.com>
wdavis@x102c.ess.harris.com (davis william 26373)
chawley%sundiver.esd.sgi.com@SGI.COM (Christopher J. Hawley)
uunet!atexnet!jackal (Phil Hammar)
Topher Eliot <eliot@DG-RTP.DG.COM>
kingdon@ai.mit.edu (Jim Kingdon)
cbp@icc.com (Chris Preston)
Doug Orr <doug@chorus.fr>
wd@samsa.pcs.com (wd)
bernie@DIALix.oz.au (Bernd Felsche)
Mark Brader <msb@sq.com>
contact!smarry (Marc Moorcroft)
bob@wyse.com (Bob McGowen x4312 dept208)
miken@pyrhard2.pyramid.com (Mike Newman)
jtp@ngs.fi (Jukka Partanen)
now maybe my mailbox will stop eating up all of my quota ...
-Fred Whiteside whitesid@darwin.CMH.McMaster.CA
Chedoke-McMaster Hospitals fred@maccs.DCSS.McMaster.CA
--
Fred Whiteside Chedoke-McMaster Hospitals, Hamilton, Ontario, Canada
whitesid@{mcmaster.ca,darwin.cmh.mcmaster.ca,130.113.2.18}
fred@maccs.DCSS.McMaster.CA ...!uunet!utai!utgpu!maccs!fred