155734 - support '##.' in format-number() function

Status: VERIFIED FIXED

(Core :: XSLT, enhancement)

Description

[David Balch]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; rv:1.1a+) Gecko/20020702
BuildID:    2002070204

XSLT format-number() http://www.w3.org/TR/xslt#format-number

The XSLT spec requires that the formatting pattern has at least one '0' before
the decimal-point, e.g. '#0.' not '##.'

However, many processors seems to accept the '##.' pattern, and it has been
expected by the reporter and others.

Might it be a good idea for us to support it?

Reproducible: Always
Steps to Reproduce:
Try attached testcase.

Actual Results:
Unformatted from XML: 37425.6770833333
Format from XML as "##.": invalid parameter value for function:
format-number(number,'##.')
Format from XML as "##.##": invalid parameter value for function:
format-number(number,'##.##')

Format 191.252 as "##.": invalid parameter value for function:
format-number(191.252,'##.')
Format 191.252 as "##.##": invalid parameter value for function:
format-number(191.252,'##.##') 


Expected Results:
Unformatted from XML: 37425.6770833333
Format from XML as "##.": 37426
Format from XML as "##.##": 37425.68

Format 191.252 as "##.": 191
Format 191.252 as "##.##": 191.25

Comment 1

[David Balch]

Attachment: XSLT stylesheet using format-number()

Comment 2

[David Balch]

Attachment: XSLT stylesheet using format-number()

Comment 3

[David Balch]

Attachment: XML file with data to test xslt file

Comment 4

[David Balch]

Sorry, just realised that I put the wrong expected/actual results in. They don't
match the testcase, but they do give an idea.

I'll have to apply for more priveleges so I can correct it...

Comment 5

[Axel Hecht]

David, IMHO the testcase is fine as is, don't bother.
Nicely filed report, thanx. Give me another one of those, and I'll boost your 
privs.

Comment 6

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

ok guys, waddaya think, do we want to support this? I tend to lean towards yes
so that gets half a vote from me

Comment 7

[Axel Hecht]

two wrongs don't make no right, but two half votes get you a bug ;-)
I'll assign that to Jonas, he knows what he did, hopefully. I tried to remember
how that code works, and failed.

Comment 8

[David Balch]

Jonas, excuse me for being dense, but where in the spec does it require '0'? 
(Re:
http://groups.google.com/groups?dq=&hl=en&lr=&ie=UTF-8&frame=right&th=3e5bb244ea8f4d5a&seekm=ag0vfo%24oti4%40ripley.netscape.com#link6)

Also, I'm not sure on how the decision gets made for adding this sort of thing,
when does adding something for compatiblity become extending the spec?

Is it a case of making sure that you do what the spec wants, and then you can
add  other features as you please (suitably separated from the spec stuff)?

Comment 9

[Axel Hecht]

so, once again, the link to the spec.
... and the grammar goes
integer    := '#'* '0'* '0'
hey, ho, hey, ho.
see, one 0 is required.
I understand Jonas' as well as my comments as "yes, we should support that",
and the "compat" keyword says so, too. Not to mention that I confirmed the bug
instead of WONTFIXing it.

Comment 10

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

http://www.w3.org/TR/xslt#format-number says:

"The format pattern string is in the syntax specified by the JDK 1.1
DecimalFormat class"

with a link to:
http://java.sun.com/products/jdk/1.1/docs/api/java.text.DecimalFormat.html

which defines

subpattern := {prefix}integer{.fraction}{suffix}
integer    := '#'* '0'* '0'

So the integer part must always end with a '0'.

The reason that i'm even considering adding this however is that Suns
javaimplementation (and all other ones i would guess) accepts patterns like '##.'

Comment 11

[Jacob Kjome]

note related bug 156594.  If the format-number() function is fixed to support
"##.", might as well fix the rounding error at the same time.

Jake

Comment 12

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

the patch in bug 156594 fixes this too

Comment 13

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

fixed with the branchlanding

Comment 14

[Axel Hecht]

we didn't verify for a long time.
I really checked, so VERIFIED.

Comment 15

[Axel Hecht]

*** Bug 163657 has been marked as a duplicate of this bug. ***