defined("foo") eq "foo"

ISTM that this would be a handy shortcut :

    if (defined("foo") eq "foo") {print
"DWIM"}


it wouldnt work as expected for ugly code like

    if (defined 2 == 1) {},

but thats too ugly to be defensible coding style.



So, I tried patching pp_defined(),
in this (non-functional) patch.

what did I do wrong ?

On Sep 6, 2007, at 9:33 PM, Jim Cromie wrote:

> ISTM that this would be a handy shortcut :
>
>    if (defined("foo") eq "foo")
{print "DWIM"}

Why would you actually want to do this?

And what about

    if (defined("0"))

?

xoxo,
Andy

--
Andy Lester => andypetdance.com => www.petdance.com =>
AIM:petdance

On Sep 6, 2007, at 9:42 PM, Andy Lester wrote:
>
> On Sep 6, 2007, at 9:33 PM, Jim Cromie wrote:
>
>> ISTM that this would be a handy shortcut :
>>
>>    if (defined("foo") eq "foo")
{print "DWIM"}
>
> Why would you actually want to do this?

Probably to do something like

   if (defined($x) and $x eq 'foo') { print
"DWIM"}

which with // can be written

   if (($x // '') eq 'foo') { print "DWIM"}

> And what about
>
>    if (defined("0"))

or

   if (defined(''))

both would be a problem and is a very good read why not to
change.

Graham.

Jim Cromie wrote:
> 
> ISTM that this would be a handy shortcut :
> 
>    if (defined("foo") eq "foo")
{print "DWIM"}

Uhh, WDYM?  (What Do You Mean)


-- 
There will be snacks.

Graham Barr wrote:
> On Sep 6, 2007, at 9:42 PM, Andy Lester wrote:
>>
>> On Sep 6, 2007, at 9:33 PM, Jim Cromie wrote:
>>
>>> ISTM that this would be a handy shortcut :
>>>
>>> if (defined("foo") eq
"foo") {print "DWIM"}
>>
>> Why would you actually want to do this?
>
> Probably to do something like
>
> if (defined($x) and $x eq 'foo') { print
"DWIM"}
>

this is precisely what prompted the experiment.


> which with // can be written
>
> if (($x // '') eq 'foo') { print "DWIM"}
>
>> And what about
>>
>> if (defined("0"))
>
> or
>
> if (defined(''))
>
> both would be a problem and is a very good read why not
to change.
>
> Graham.
>
>

SvTRUE takes care of it - from perlapi,

SvTRUE Returns a boolean indicating whether Perl would
evaluate the SV 
as true or false, defined or undefined. Does not handle
’get’ magic.

BTW- that wording seems somewhat ambiguous.


so in the patch, 'true' vals are returned as is (ex:
"foo")
and false but defined are PL_sv_yes.

The patch passes most tests, failing on these:

Failed 30 tests out of 1426, 97.90% okay.
../ext/Devel/PPPort/t/call.t
../ext/Encode/t/Encode.t
../ext/Hash/Util/FieldHash/t/01_load.t
../ext/IO/t/io_const.t
../ext/List/Util/t/first.t
../ext/List/Util/t/max.t
../ext/List/Util/t/maxstr.t
../ext/List/Util/t/min.t
../ext/List/Util/t/minstr.t
../ext/List/Util/t/openhan.t
../ext/List/Util/t/p_first.t
../ext/List/Util/t/p_max.t
../ext/List/Util/t/p_maxstr.t
../ext/List/Util/t/p_min.t
../ext/List/Util/t/p_minstr.t
../ext/List/Util/t/p_openhan.t
../ext/threads/shared/t/wait.t
../lib/Archive/Tar/t/02_methods.t
../lib/DBM_Filter/t/01error.t
../lib/DBM_Filter/t/02core.t
../lib/DBM_Filter/t/encode.t
../lib/Math/BigInt/t/bare_mbi.t
../lib/Math/BigInt/t/bigintpm.t
../lib/Math/BigInt/t/sub_mbi.t
../lib/Math/BigRat/t/bigrat.t
../lib/Math/BigRat/t/bigratpm.t
../lib/Math/BigRat/t/trap.t
../lib/Pod/Simple/t/search05.t
../lib/Test/Harness/t/assert.t
../lib/Test/Simple/t/use_ok.t
### Since not all tests were successful, you may want to run
some of ...


The 1st 2 failures are expected, given the tests -

ok( defined"foo", 1, ".......")

and your acid tests above also work:

[jimcharpo defret]$ ./perl -Ilib -e 'print defined 0,
defined "", 
defined "0", "n"'
111


All that said, I suppose its not strictly backwards
compatible,
so I'll drop it. Im satisfied that it 'works'.

heres the patch anyway.

aww crap - that last patch had superflous stuff in it.

here it is again, de-cluttered.



[jimcharpo defret]$ ./perl -Ilib -e
'$_="junkn"; print if defined "foo"'
junk



BTW, I see nothing in `perldoc -f defined` that specifies
the return value.

quoting:

So you should use "defined" only when you’re
questioning the integrity
of what you’re trying to do. At other times, a simple
comparison to 0 or 
"" is what
you want.

Jim Cromie wrote:
> Failed 30 tests out of 1426, 97.90% okay.
> ../ext/Devel/PPPort/t/call.t
> ../ext/Encode/t/Encode.t
> ../ext/Hash/Util/FieldHash/t/01_load.t
> ../ext/IO/t/io_const.t
> ../ext/List/Util/t/first.t
> ../ext/List/Util/t/max.t
> ../ext/List/Util/t/maxstr.t
> ../ext/List/Util/t/min.t
> ../ext/List/Util/t/minstr.t
> ../ext/List/Util/t/openhan.t
> ../ext/List/Util/t/p_first.t
> ../ext/List/Util/t/p_max.t
> ../ext/List/Util/t/p_maxstr.t
> ../ext/List/Util/t/p_min.t
> ../ext/List/Util/t/p_minstr.t
> ../ext/List/Util/t/p_openhan.t
> ../ext/threads/shared/t/wait.t
> ../lib/Archive/Tar/t/02_methods.t
> ../lib/DBM_Filter/t/01error.t
> ../lib/DBM_Filter/t/02core.t
> ../lib/DBM_Filter/t/encode.t
> ../lib/Math/BigInt/t/bare_mbi.t
> ../lib/Math/BigInt/t/bigintpm.t
> ../lib/Math/BigInt/t/sub_mbi.t
> ../lib/Math/BigRat/t/bigrat.t
> ../lib/Math/BigRat/t/bigratpm.t
> ../lib/Math/BigRat/t/trap.t
> ../lib/Pod/Simple/t/search05.t
> ../lib/Test/Harness/t/assert.t
> ../lib/Test/Simple/t/use_ok.t
> ### Since not all tests were successful, you may want
to run some of ...
> 
> 
> The 1st 2 failures are expected, given the tests -
> 
> ok( defined"foo", 1, ".......")

I'd like to see specifically what failed, that would give an
idea of just what
the consequences of this patch are.  I'm surprised at the
Test::Simple
failure.  That test only checks for whether defined &sub
is true or false.


-- 
Robrt:   People can't win
Schwern: No, but they can riot after the game.

Michael G Schwern wrote:
> Jim Cromie wrote:
>   
>> Failed 30 tests out of 1426, 97.90% okay.
>> ../ext/Devel/PPPort/t/call.t
>> ../ext/Encode/t/Encode.t
>> ../ext/Hash/Util/FieldHash/t/01_load.t
>> ../ext/IO/t/io_const.t
>> ../ext/List/Util/t/first.t
>> ../ext/List/Util/t/max.t
>> ../ext/List/Util/t/maxstr.t
>> ../ext/List/Util/t/min.t
>> ../ext/List/Util/t/minstr.t
>> ../ext/List/Util/t/openhan.t
>> ../ext/List/Util/t/p_first.t
>> ../ext/List/Util/t/p_max.t
>> ../ext/List/Util/t/p_maxstr.t
>> ../ext/List/Util/t/p_min.t
>> ../ext/List/Util/t/p_minstr.t
>> ../ext/List/Util/t/p_openhan.t
>> ../ext/threads/shared/t/wait.t
>> ../lib/Archive/Tar/t/02_methods.t
>> ../lib/DBM_Filter/t/01error.t
>> ../lib/DBM_Filter/t/02core.t
>> ../lib/DBM_Filter/t/encode.t
>> ../lib/Math/BigInt/t/bare_mbi.t
>> ../lib/Math/BigInt/t/bigintpm.t
>> ../lib/Math/BigInt/t/sub_mbi.t
>> ../lib/Math/BigRat/t/bigrat.t
>> ../lib/Math/BigRat/t/bigratpm.t
>> ../lib/Math/BigRat/t/trap.t
>> ../lib/Pod/Simple/t/search05.t
>> ../lib/Test/Harness/t/assert.t
>> ../lib/Test/Simple/t/use_ok.t
>> ### Since not all tests were successful, you may
want to run some of ...
>>
>>
>> The 1st 2 failures are expected, given the tests -
>>
>> ok( defined"foo", 1,
".......")
>>     
>
> I'd like to see specifically what failed, that would
give an idea of just what
> the consequences of this patch are.  I'
btw - I trust the WDYM is clear at this point ..


[jimcharpo defret]$ ./perl -Ilib
lib/Test/Simple/t/use_ok.t
1..13
ok 1 - use Symbol;
ok 2 - use_ok() no args exports defaults
ok 3 - use Symbol;
ok 4 -   one arg, defaults overriden
Bizarre copy of CODE in aassign at lib/Test/More.pm line
260.
# Looks like you planned 13 tests but only ran 4.
# Looks like your test died just after 4.

so, adding -d, setting breakpoint,

Foo::two::(lib/Test/Simple/t/use_ok.t:26):
26:         ::ok( defined &qualify,       '  right
function exported' );
  DB<5> s
Test::More::ok(lib/Test/More.pm:260):
260:        my($test, $name) = _;
  DB<5> n
Bizarre copy of CODE in aassign at lib/Test/More.pm line
260.
# Looks like you planned 13 tests but only ran 4.
# Looks like your test died just after 4.

R-repeating,

Foo::two::(lib/Test/Simple/t/use_ok.t:26):
26:         ::ok( defined &qualify,       '  right
function exported' );
  DB<5> s
Test::More::ok(lib/Test/More.pm:260):
260:        my($test, $name) = _;
  DB<5> s
Test::Builder::CODE(0x87693fc)(lib/Test/Builder.pm:1679):
1679:       my $in_eval = 0;
  DB<5> T
$ =
Test::Builder::__ANON__[lib/Test/Builder.pm:1684]('Bizarre
copy of 
CODE in aassign at lib/Test/More.pm line 260.^J') called
from file 
`lib/Test/More.pm' line 260
. = Test::More::ok('$;$', '  right function exported')
called from file 
`lib/Test/Simple/t/use_ok.t' line 26

one more time..

  DB<4> l
26==>b      ::ok( defined &qualify,       '  right
function exported' );
27      }

  DB<4> s
Test::More::ok(lib/Test/More.pm:260):
260:        my($test, $name) = _;
  DB<4> x _
Bizarre copy of CODE in leaveeval at lib/perl5db.pl line
638.

This looks quite internal ..

z  <;> nextstate(Foo::two 782 use_ok.t:26) v:{
10 <0> pushmark s
11 <#> gv[*Foo::two::qualify] s
12 <1> rv2cv[t15] sK*/AMPER,32
13 <1> defined sKM/1
14 <$> const[PV "  right function exported"]
sM
15 <#> gv[*ok] s
16 <1> entersub[t16] vKS/TARG,1
17 <2> leaveloop vK/2

does anyone see a problem with any of the flags on these ops
?
something to do with refcounting ?

On Sep 7, 2007, at 12:27 AM, Jim Cromie wrote:
>> which with // can be written
>>
>> if (($x // '') eq 'foo') { print "DWIM"}
>>
>>> And what about
>>>
>>> if (defined("0"))
>>
>> or
>>
>> if (defined(''))
>>
>> both would be a problem and is a very good read why
not to change.
>>
>> Graham.
>>
>>
>
> SvTRUE takes care of it - from perlapi,

No it does not.

You cannot make  if(defined($x) eq $x) and if(defined($x))
both  
always have the same result for every value of $x

> SvTRUE Returns a boolean indicating whether Perl would
evaluate the  
> SV as true or false, defined or undefined. Does not
handle ’get’  
> magic.

Which means it does not handle overloaded objects.

Graham.

the "it" proposal was to have a new LNV (such as
$^it) which would be a
non-magical copy of whatever was the argument of the most
recent call
to exists or defined.  (exactly how non-magical would it be
would have to be
talked out.)

Implementing it involves adding a couple lines to pp_exists
and pp_defined
which don't do anything but slow down the general case. 
$^it would work here:

  if (defined($x) and $x eq 'foo') { print
"DWIM"}

could become

  if (defined($x) and $^it eq 'foo') { print
"DWIM"}

which becomes advantageous when $x is something expensive
to look up, like a tied database fetch.


Of course one can get the same result, without using any new
syntax,
with

 if (defined(my $_tmpx=$x) and $_tmpx eq 'foo') { print
"DWIM"}

so I'm going to stop now and doubt the wisdom of having
started. ;)