# New Ticket Created by Will Coleda
# Please include the string: [perl #46933]
# in the subject line of all future correspondence about this issue.
# <URL: http://rt.perl.org/rt3/Ticket/Display.html?id=46933 >
Currently, Data::Dumper treats Undef pmcs like any other pmcs. When dumping,
then, if a single Undef is repeatedly re-used, the Undef is cached by the
dumper (like any other PMC) and rather than dump as "undef", dumps as a
reference. Here's some sample code from particle:
.sub 'test' :main
load_bytecode 'Data/Dumper.pir'
$P0 = new 'Undef'
$P1 = new 'Undef'
$P2 = new 'Hash'
$P2['$P0'] = $P0
$P2['$P1'] = $P1
$P2['$P0-ref'] = $P0
$P99 = new 'Data::Dumper'
$P99.'dumper'( $P2 )
.end
c:\usr\local\parrot\trunk>parrot ..\t-dumper.pir
"VAR1" => Hash {
"$P0" => undef,
"$P0-ref" => \VAR1["$P0"],
"$P1" => undef
}
So, since Undef isn't treated in a special fashion, I can understand why
this works the way it does, and it is arguably correct.
I propose that we change this to make Undef special cased, and avoid caching
the Undef PMC in dumper, so that a more perl5-like output is created (where
undef isn't an object, it's a value. Small patch follows; all tests pass (so
nothing is relying on the existing behavior.). Thjs changes the output above
to:
"VAR1" => Hash {
"$P0" => undef,
"$P0-ref" => undef,
"$P1" => undef
}
I'll wait for a consensus before applying; There was disagreement on #parrot
about changing this, putting it out here for discussion.
$ svn diff runtime/
Index: runtime/parrot/library/Data/Dumper/Base.pir
===================================================================
--- runtime/parrot/library/Data/Dumper/Base.pir (revision 22503)
+++ runtime/parrot/library/Data/Dumper/Base.pir (working copy)
@@ -188,6 +188,10 @@
.local string func
.local string name
+ # Don't cache undef...
+ $I0 = defined _dump
+ unless $I0 goto NOT_IN_CACHE
+
(type, name) = self."cache"( paramName, _dump )
if type == -1 goto NOT_IN_CACHE
