'Data::Dumper' reports double value as string. #9610
Comments
From 0body0@rambler.ruCreated by 0body0@rambler.ru$ perl -MData::Dumper -e'print Dumper(1.0)' ===Note single quotes=== Best regard, grian Perl Info |
From @eserteAnatoly Grishaev (via RT) <perlbug-followup@perl.org> writes:
Note that the result depends whether the XS or the pure perl $ perl -MData::Dumper -e'$Data::Dumper::Useperl=0; print Dumper(1.0)' Regards, -- |
|
The RT System itself - Status changed from 'new' to 'open' |
From module@renee-baecker.deOn Sa. 27. Dez. 2008, 02:40:54, grian wrote:
Is this really a bug? I know that the behaviour differs with Dumper.xs and Dumper.pm. This is current behaviour: C:\>perl -MData::Dumper -e "print Dumper 1.0" C:\>perl -MData::Dumper -e "print Dumper 1.01" C:\>perl -MData::Dumper -e "$Data::Dumper::Useperl=1; print Dumper 1.01" C:\>perl -MData::Dumper -e "$Data::Dumper::Useperl=1; print Dumper 1.0" C:\>perl -MData::Dumper -e "$Data::Dumper::Useperl=1; print Dumper C:\>perl -MData::Dumper -e "$Data::Dumper::Useperl=1; print Dumper You can see that even with the Perl implementation of Dumper it uses elsif ($val =~ /^(?:0|-?[1-9]\d{0,8})\z/) { # safe decimal number If you want to get rid of the single quotes for decimals that ends with |
From 0body0@rambler.ruAnd what is a bug? My only motivation to see numerical scalars without single quotes if this A thus examle of is ok then big numbers dump with quotes
Desired behaviour
and
and may be
and may be
In last case xs implementation will be more correct maybe change this string and appropriate change in xs, and tests I can provide a patch for this change if anybody tell me how do it( I use I think differences with perl and xs implementation will be less, and Happy New Year. ----- Original Message -----
|
From module@renee-baecker.deOn Mi. 31. Dez. 2008, 12:47:08, grian wrote:
please provide a unified diff (diff -u <original_version> Then you can send the diff to this ticket system.
|
From 0body0@rambler.ruPatch for Data::Dumper (Fix decimal printing) <grian@cpan.org> ----- Original Message -----
|
From 0body0@rambler.ruCreated by 0body0@rambler.ru$ perl -MData::Dumper -e'print Dumper(1.0)' ===Note single quotes=== Best regard, grian Perl Info |
From @cpansproutOn Mon Jan 12 04:21:34 2009, grian wrote:
The patch attached to this ticket has been here for three years. I�ve -- Father Chrysostomos |
From @cpansproutOn Fri Feb 03 23:14:39 2012, sprout wrote:
And if anyone else wants to apply it, please don�t assume I will, but go (And I need to answer Nicholas Clark�s mail about continue, but that -- Father Chrysostomos |
From @nwc10On Fri, Feb 03, 2012 at 11:14:40PM -0800, Father Chrysostomos via RT wrote:
Sorry, I missed this one too. If I understand the original bug report, this patch is going a lot further. $ ./perl -Ilib -MData::Dumper -e 'print Dumper(1.0)' and then the observation was made that it's the XS code. The pure perl $ ./perl -Ilib -MData::Dumper -e '$Data::Dumper::Useperl = 1; print Dumper(1.0)' Extracting the patches from the gzip'd tarfile, and tweaking it to apply: Inline Patchdiff --git a/dist/Data-Dumper/Dumper.pm b/dist/Data-Dumper/Dumper.pm
index fcf06ad..52d1f6c 100644
--- a/dist/Data-Dumper/Dumper.pm
+++ b/dist/Data-Dumper/Dumper.pm
@@ -542,7 +542,7 @@ sub _dump {
and ref $ref eq 'VSTRING' || eval{Scalar::Util::isvstring($val)}) {
$out .= sprintf "%vd", $val;
}
- elsif ($val =~ /^(?:0|-?[1-9]\d{0,8})\z/) { # safe decimal number
+ elsif ($val =~ /^(?:0|-?[1-9]\d{0,8})(?:\.\d{0,5}[1-9])?\z/) { # safe decimal number
$out .= $val;
}
else { # string
diff --git a/dist/Data-Dumper/Dumper.xs b/dist/Data-Dumper/Dumper.xs
index 2ad53a1..8fe8e70 100644
--- a/dist/Data-Dumper/Dumper.xs
+++ b/dist/Data-Dumper/Dumper.xs
@@ -61,6 +61,8 @@ Perl_utf8_to_uvchr(pTHX_ U8 *s, STRLEN *retlen)
#define DD_is_integer(sv) SvIOK(sv)
#endif
+#define DD_is_double(sv) SvNOK(sv)
+
/* does a string need to be protected? */
static I32
needs_quote(register const char *s, STRLEN len)
@@ -891,10 +893,12 @@ DD_dump(pTHX_ SV *val, const char *name, STRLEN namelen, SV *retval, HV *seenhv,
}
}
- if (DD_is_integer(val)) {
+ if (DD_is_integer(val) || DD_is_double(val)) {
STRLEN len;
if (SvIsUV(val))
len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"UVuf, SvUV(val));
+ else if (DD_is_double(val))
+ len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"NVgf, SvNV(val));
else
len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"IVdf, SvIV(val));
if (SvPOK(val)) {
New result is $ ./perl -Ilib -MData::Dumper -e 'print Dumper(1.1)' Whereas the existing code is ie '' used on floating point values for decimals. I don't think that we should change behaviour to use floating point literals. I think that a correct fix would only be a change to Dumper.xs, and is Inline Patchdiff --git a/dist/Data-Dumper/Dumper.xs b/dist/Data-Dumper/Dumper.xs
index 2ad53a1..1746aa6 100644
--- a/dist/Data-Dumper/Dumper.xs
+++ b/dist/Data-Dumper/Dumper.xs
@@ -891,12 +891,18 @@ DD_dump(pTHX_ SV *val, const char *name, STRLEN namelen, SV *retval, HV *seenhv,
}
}
- if (DD_is_integer(val)) {
+ if (DD_is_integer(val)
+ || (SvNOK(val) && -1e10 < SvNVX(sv) && SvNVX(sv) < 1e10
+ && SvNVX(sv) == Perl_floor(SvNVX(sv)))) {
STRLEN len;
- if (SvIsUV(val))
- len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"UVuf, SvUV(val));
- else
- len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"IVdf, SvIV(val));
+ if (DD_is_integer(val)) {
+ if (SvIsUV(val))
+ len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"UVuf, SvUV(val));
+ else
+ len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"IVdf, SvIV(val));
+ } else {
+ len = my_snprintf(tmpbuf, sizeof(tmpbuf), "%"NVgf, SvNVX(val));
+ }
if (SvPOK(val)) {
/* Need to check to see if this is a string such as " 0".
I'm assuming from sprintf isn't going to clash with utf8.
$ ./perl -Ilib -MData::Dumper -e 'print Dumper(1.0)' ie only use unquoted literals for integers with 9 or fewer digits, and if (len > 10) { and checking that the floating point value is less than ±2**53 before doing However that "%g"NVgf is not correct for larger values, and I'm not sure $ ./perl -Ilib -MData::Dumper -e 'print Dumper(12345678901.1)' Nicholas Clark |
From 0body0@rambler.ru04.02.2012 12:10, Nicholas Clark via RT пи�е�:
Maybe change perl version of Data::Dumper was not good idea ...
I looks that in the expression: missed fabs() function call (SvNOK(val)&& -1e10< SvNVX(sv)&& SvNVX(sv)< 1e10
For most common cases it better apply this patch and left note about
-- |
From @nwc10On Tue Feb 07 06:54:25 2012, grian wrote:
I didn't think so. My intent was that -1e10 < SvNVX(sv) && SvNVX(sv) < 1e10 should be equivalent to fabs(SvNVX(sv)) < 1.e0 except that a) it doesn't assume that SvNVX(sv) is of type double Nicholas Clark |
From @nwc10[and to the list this time] On Tue Feb 07 06:54:25 2012, grian wrote:
I didn't think so. My intent was that -1e10 < SvNVX(sv) && SvNVX(sv) < 1e10 should be equivalent to fabs(SvNVX(sv)) < 1.e0 except that a) it doesn't assume that SvNVX(sv) is of type double I'd guess that fabs() return NaN for a Nan, but I would have had Nicholas Clark |
From [Unknown Contact. See original ticket][and to the list this time] On Tue Feb 07 06:54:25 2012, grian wrote:
I didn't think so. My intent was that -1e10 < SvNVX(sv) && SvNVX(sv) < 1e10 should be equivalent to fabs(SvNVX(sv)) < 1.e0 except that a) it doesn't assume that SvNVX(sv) is of type double I'd guess that fabs() return NaN for a Nan, but I would have had Nicholas Clark |
From @jkeenanOn Tue Feb 07 07:16:52 2012, nicholas wrote:
Discussion in this RT petered out about twelve months ago. Are there any changes to code or documentation which we should still be Thank you very much. |
From @tonycozOn Sun Feb 03 10:34:34 2013, jkeenan wrote:
I've updated the status to mark the patch as rejected based on the Nicholas complained that:
but the current mechanism: a) already does a round trip, if the user supplied the data as a float, b) already introduces errors: tony@mars:.../git/perl$ ./perl -Ilib -MData::Dumper -e 'print Fixing that is little more complex than the supplied though - possibly Tony |
jkeenan
added
dist-Data-Dumper
Migrated from rt.perl.org#61754 (status was 'open')
Searchable as RT61754$
The text was updated successfully, but these errors were encountered: