fdo57180 add function NUMBERVALUE() to calc as defined in ODFF1.2

[Eike Rathke]

Hi Winfried,

On Thursday, 2012-12-13 09:59:38 +0100, Winfried Donkers wrote:

> The coding of this function seems to be going nicely, but for the string to double conversion I would like some pointers/interpretations.
> 1. ODFF1.2 refers to 'xsd:float' and conversion 'according to the definition provided in XML Schema, Part 2, Section 3.2.4'. I don't think I found that particular type and section. Could you give links to these?

http://www.w3.org/TR/xmlschema-2/#float
Note however that our data type is double, not float, so
http://www.w3.org/TR/xmlschema-2/#double

Personally I think that xsd:float and section 3.2.4 are awkward and it
should be xsd:double and section 3.2.5 instead, but float may be there
to have a common least denominator, I don't remember. This NUMBERVALUE()
definition is the only place mentioning float, earlier versions of the
draft referenced IEEE 754 double at several other places but those got
removed as well.

Short essence: use rtl::math::stringToDouble() to convert, include
rtl/math.hxx

First check if DecimalSeparator appears in GroupSeparator and bail out
if so.

Slightly different processing as given in steps 1-4 of ODFF because
stringToDouble() handles one group separator and this doesn't need extra
processing if there are <=1 characters in GroupSeparator:

If GroupSeparator is not present or empty, call
rtl::math::stringToDouble( X, DecimalSeparator[0], 0, ...)

If GroupSeparator contains one character (length==1), call
rtl::math::stringToDouble( X, DecimalSeparator[0], GroupSeparator[0], ...)

If GroupSeparator contains more than one character, remove all
characters before the first DecimalSeparator in the number string that
match one of GroupSeparator, and call
rtl::math::stringToDouble( X, DecimalSeparator[0], 0, ...)

NOTE: the GroupSeparator OUString/String may contain UTF-16 encoded
      Unicode surrogate pairs, so OUString::iterateCodePoints() should
      be used on the numberstring if that is the case.

Theoretically DecimalSeparator might as well, but stringToDouble()
doesn't handle that. If that really was a problem we'd have to craft
some additional method that takes the full Unicode range.. which I'd
avoid for now.

Of course do the trailing percent sign processing for all cases before
calling stringToDouble().

> 2. ODFF1.2 mentions under VALUE() that dates should be recognised, as well as fractional numbers, e.g. '2 3/8'. I'm not sure if NUMBERVALUE() should recognise that too. For dates it is as good as impossible (no separator, language and order is known); but fractional numbers could be recognised. What is your judgement on this?

Don't. NUMBERVALUE() is not defined for other cases.

Btw, "2 3/8" could be 23.8 with DecimalSeparator='/' and
GroupSeparator=' ' ;-)

> 3. Currently I am putting the code for the string to double conversion in de function ScNumberValue() itself as i can't find accessable stringtodouble functions that do not ask for language settings. Now would be a very good time to tell that I have overlooked the accessible stringtodouble function :).

See above ;-)  and sal/inc/rtl/math.hxx

  Eike

-- 
LibreOffice Calc developer. Number formatter stricken i18n transpositionizer.
GnuPG key 0x293C05FD : 997A 4C60 CE41 0149 0DB3  9E96 2F1A D073 293C 05FD
Support the FSFE, care about Free Software! https://fsfe.org/support/?erack
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20121214/a4c64a55/attachment.pgp>