dglo@deimos.ads.com (Dave Glowacki) (03/14/90)
I converted a shell script to a perl script and spent a couple hours
trying to track down the following sort of error. Is there any way
to get error messages from a file included via 'do'?
--------% cat stupid.pl
sub ShellCatenate {
local($a, $b) = ('z', 'A');
$a$b; # supposed to be $a . $b
}
--------% perl stupid.pl
syntax error in file stupid.pl at line 4, next 2 tokens "$a$b"
Execution aborted due to compilation errors.
--------% cat mystery.pl
do 'stupid.pl';
$tmpname = &ShellCatenate;
print "TMPNAME='",$tmpname,"'\n";
print "TMPNAME='",&ShellCatenate,"'\n";
--------% perl mystery.pl
TMPNAME='2'
TMPNAME='zA'
&ShellCatenate is apparently returning the list ($a, $b). Is this
perl's guess at what I really wanted or is it some sort of parse tree
remnant?tchrist@convex.COM (Tom Christiansen) (03/14/90)
In article <11238@deimos.ADS.COM> dglo@deimos.ads.com (Dave Glowacki) writes: >I converted a shell script to a perl script and spent a couple hours >trying to track down the following sort of error. Is there any way >to get error messages from a file included via 'do'? I use this: % cat source.pl ########################################################################### sub source { local($file) = @_; local($return) = 0; $return = do $file; die "couldn't parse $file: $@" if $@; die "couldn't do $file: $!" unless defined $return; die "couldn't run $file" unless $return; $return; } 1; ########################################################################### I haven't entirely decided yet whether I want the $@ check before the $! check, but I've recently switched them around to the state you now see. --tom -- Tom Christiansen {uunet,uiucdcs,sun}!convex!tchrist Convex Computer Corporation tchrist@convex.COM "EMACS belongs in <sys/errno.h>: Editor too big!"
hakanson@ogicse.ogi.edu (Marion Hakanson) (03/15/90)
In article <11238@deimos.ADS.COM> dglo@deimos.ads.com (Dave Glowacki) writes: >I converted a shell script to a perl script and spent a couple hours >trying to track down the following sort of error. Is there any way >to get error messages from a file included via 'do'? Many people suggest: do 'sub.pl' || die 'error in sub.pl'; but this does not provide informative error messages. E.g. it will not distinguish between the file being missing and it containing a syntax error. So then I tried: do 'sub.pl' || die "Cannot load sub.pl: $!, aborted"; die "$@aborted" if $@; Unfortunately, the first "die" takes effect if the file exists but contains an error, and the "$!" is gibberish in that case (the second "die" never gets to live). My current approach, though verbose, works well and prints meaningful messages: unless ( do 'sub.pl' ) { die "$@aborted" if $@; die "Cannot load sub.pl: $!, aborted"; } I'm sure someone will convert it to a one-liner.... -- Marion Hakanson Domain: hakanson@cse.ogi.edu UUCP : {hp-pcd,tektronix}!ogicse!hakanson
schaefer@ogicse.ogi.edu (Barton E. Schaefer) (03/15/90)
In article <7997@ogicse.ogi.edu> hakanson@ogicse.ogi.edu (Marion Hakanson) writes: } } unless ( do 'sub.pl' ) { } die "$@aborted" if $@; } die "Cannot load sub.pl: $!, aborted"; } } } } I'm sure someone will convert it to a one-liner.... How about do 'sub.pl' || ($@ ne "" && die "$@aborted") || die "Can't load sub.pl: $!"; Now the only thing I can't figure out is why the parentheses are necessary. Doesn't && "bind tighter" than || ? If not, why not? I thought this was supposed to act like C ... -- Bart Schaefer "EARTH: Surrender IMMEDIATELY or we PICKLE DAN QUAYLE" "THPPFT!" schaefer@cse.ogi.edu (used to be cse.ogc.edu)
schaefer@ogicse.ogi.edu (Barton E. Schaefer) (03/15/90)
In article <8005@ogicse.ogi.edu> schaefer@ogicse.ogi.edu (That's Me) writes: } } do 'sub.pl' || ($@ ne "" && die "$@aborted") || die "Can't load sub.pl: $!"; Before someone else points it out, of course it could be do 'sub.pl' || die ($@ ne "" ? "$@aborted" : "Can't load sub.pl: $!"); but that didn't show the "problem" with && and ||. -- Bart Schaefer "EARTH: Surrender IMMEDIATELY or we PICKLE DAN QUAYLE" "THPPFT!" schaefer@cse.ogi.edu (used to be cse.ogc.edu)
jgreely@giza.cis.ohio-state.edu (J Greely) (03/15/90)
In article <8006@ogicse.ogi.edu> schaefer@ogicse.ogi.edu (Barton E. Schaefer) writes: >do 'sub.pl' || die ($@ ne "" ? "$@aborted" : "Can't load sub.pl: $!"); I Like Tom's solution (which sits in the archive in perl/scripts/tchrist/source.pl): sub source { local($file) = @_; local($return) = 0; $return = do $file; die "couldn't do \"$file\": $!" unless defined $return; die "couldn't parse \"$file\": $@" if $@; die "couldn't run \"$file\"" unless $return; } 1; Not one line, but quite effective (it of course uses the "trailing-1" convention for sourced files (is this documented yet?)). Here's my contribution: die "$file: " . ($! > 0 ? $! : $@ ne '' ? $@ : "couldn't run") unless do $file > 0; (no, it doesn't print "Just another Perl hacker,") -- J Greely (jgreely@cis.ohio-state.edu; osu-cis!jgreely)
lwall@jpl-devvax.JPL.NASA.GOV (Larry Wall) (03/15/90)
In article <8005@ogicse.ogi.edu> schaefer@ogicse.ogi.edu (Barton E. Schaefer) writes:
: How about
:
: do 'sub.pl' || ($@ ne "" && die "$@aborted") || die "Can't load sub.pl: $!";
:
: Now the only thing I can't figure out is why the parentheses are necessary.
: Doesn't && "bind tighter" than || ? If not, why not? I thought this was
: supposed to act like C ...
Remember that die is a LIST operator, and that LIST operators gobble up
all arguments to their right. You can do what you want if you say
do 'sub.pl' || $@ ne "" && die("$@aborted") || die "Can't load sub.pl: $!";
Though I'd probably say that as
do 'sub.pl' || die $@ ? "$@aborted" : "Can't load sub.pl: $!";
Larrylwall@jpl-devvax.JPL.NASA.GOV (Larry Wall) (03/16/90)
In article <100570@convex.convex.com> tchrist@convex.COM (Tom Christiansen) writes:
: ###########################################################################
: sub source {
: local($file) = @_;
: local($return) = 0;
:
: $return = do $file;
: die "couldn't parse $file: $@" if $@;
: die "couldn't do $file: $!" unless defined $return;
: die "couldn't run $file" unless $return;
: $return;
: }
:
: 1;
: ###########################################################################
:
: I haven't entirely decided yet whether I want the $@ check
: before the $! check, but I've recently switched them around
: to the state you now see.
The message "couldn't parse..." isn't quite accurate, since $@ can be
set by run-time errors as well as compile-time errors. It might parse
fine, but suppose you divide by 0...
It might even be that the 3rd condition will never be met unless they
leave the 1; off the end.
Larrytony@oha.UUCP (Tony Olekshy) (03/18/90)
In message <7997@ogicse.ogi.edu>, hakanson@ogicse.ogi.edu (Marion Hakanson) writes: > My current approach, though verbose, works > well and prints meaningful messages: > > unless ( do 'sub.pl' ) { > die "$@aborted" if $@; > die "Cannot load sub.pl: $!, aborted"; > } > > I'm sure someone will convert it to a one-liner.... Yes, many have, but I have been doing it this way too: do ($_="somefile.pl") || die "$0: can't load $_:\n".($@?$@:$!); -- Yours, etc., Tony Olekshy (...!alberta!oha!tony or tony@oha.UUCP).