Difference between revisions of "Calc/Drafts/Treatment of new Excel 2010 functions"

Latest revision as of 15:13, 20 October 2010

This article is a workspace for discussing the treatment of new functions of Excel 2010 in OpenOffice.org code. For using functions in spreadsheets look at Documentation/How_Tos/Calc:_Functions_listed_alphabetically

Please do not change the layout to a table. Such table makes it hard to edit a single section.

You want to help collecting the information? Try to add: Which parameters has the function? Does a corresponding OOo function exists already? Does a corresponding function exists in OpenFormula? How difficult is it to implement a new function if necessary? Do you see any problems or obstacles?

Reference Documents

The discussion uses the description of the new functions in Function Improvements in Microsoft Office Excel 2010 and in What's New: Changes made to Excel functions and the connected descriptions of the functions. A further list can be found in chapter 2.2.3 of [MS-XLSX]: Excel Extensions to the Office Open XML SpreadsheetML File Format (.xlsx) Specification [1].

For OpenFormula the actual document from OpenDocument - Formula SC Public Documents is used.

File:Compare Function Category Excel ODF OOo.ods Table of functions with their categories collected from Excel 2010 Beta, ODFF draft 28 and OOo3.2.

AGGREGATE

new

Excel

AGGREGATE(function_num, options, ref1, [ref2], …) or AGGREGATE(function_num, options, array, [k])

OpenFormula

no such function specified

OOo3.2

no similar function implemented

Comment

Mapping is not possible, because it is not specified in OpenFormula. The function AGGREGATE extends the function SUBTOTAL with further functions and more options to decide, which rows are ignored. A mapping to SUBTOTAL would be only possible in very few cases, when the option is given as constant 0 or 5. For to map option 5 to SUBTOTAL it is needed that Issue 35718 is fixed.

BETA.DIST

follow-up version of BETADIST

Excel

BETA.DIST(x,alpha,beta,cumulative,[A],[B]), default A=0 and B=1

cumulative is boolean, but numbers are accepted too

The argument A may be empty, using two semicolons.

A <= x <= B

OpenFormula

BETADIST( Number x ; Number alpha ; Number beta [ ; Number a = 0 [ ; Number b = 1 [ ; Logical Cumulative = TRUE() ] ] ] )

alpha > 0, beta > 0, a < b

no constraints for x besides the cases with pole.

OOo3.2

BETADIST(number; alpha; beta [; start [; end [; cumulative]]])

implemented as in OpenFormula defined

Comment

Mapping is possible for import. Notice the different order of arguments. Missing A or B have to be set to their default, if cumulative is given.

BETA.INV

follow-up version of BETAINV

Excel

BETA.INV(probability,alpha,beta,[A],[B]), default A=0 and B=1

The argument A may be empty, using two semicolons.

OpenFormula

BETAINV( Number p ; Number a ; Number b [ ; Number a = 0 [ ; Number b = 1 ] ] )

OOo3.2

BETAINV(number; alpha; beta; Start; End)

implemented as defined in OpenFormula, but empty Start is possible.

Comment

Directly mapping is possible.

BINOM.DIST

follow-up version of BINOMDIST

Excel

BINOM.DIST(number_s,trials,probability_s,cumulative)

number_s and trials are truncated to integer

cumulative is boolean, but numbers are accepted too

OpenFormula

BINOMDIST( Integer S ; Integer N ; Number P ; Logical Cumulative )

OOo3.2

=BINOMDIST(X; trials; SP; C)

X and trials are truncated to integer

Comment

direct mapping is possible

BINOM.INV

follow-up version of CRITBINOM

Excel

BINOM.INV(trials,probability_s,alpha), all arguments required

trials are truncated to integer

OpenFormula

CRITBINOM( Number Trials ; Number SP ; Number Alpha )

OOo3.2

CRITBINOM(trials; SP; alpha)

Comment

Direct mapping is possible. Notice that the domain for 'trails' is much larger in Excel 2010 than in OOo3.2. For example =BINOM.INV(500000;0.5;0.1) is possible in Excel 2010 but =CRITBINOM(1080;0.5;0.1) fails aready in OOo3.2. A new algorithm is needed.

CEILING.PRECISE

new

Excel

CEILING.PRECISE(number, [significance])

significance has default value 1, always the absolute value of significance is used

OpenFormula

CEILING( Number N [ ; [ Number significance ] [ ; Number mode ] ] )

nested rules, round toward positive infinity for mode omitted or zero, round away from zero in the other cases.

N and significance must have the same sign if not 0.

OOo3.2

CEILING(Number; Significance [; Mode])

Significance is required, Issue 102957 is still open, no decision about the patch there up to now.

Comment

Because OpenFormula requires, that N and significance have the same sign, a direct mapping is not possible.

CEILING.PRECISE(number) —with second argument missing— can be mapped to CEILING(number). To this a corrected version in OOo is needed, so that a missing second argument no longer results in an error.

With a given second argument, the composed formula CEILING(number; SIGN(number)*ABS(significance)) would give the same result. But the question is, whether such change should be done at all.

This function can be used with the name ISO.CEILING in Excel 2010 Beta too. It seems that CEILING.PRECISE is the favorite name, because ISO.CEILING works in EXCEL 2010 Beta, but is not listed in the available functions there. In addition an ISO.FLOOR does not exist, but only a FLOOR.PRECISE.

CHISQ.DIST

new

Excel

CHISQ.DIST(x,deg_freedom,cumulative), all arguments required

x >= 0

deg_freedom is truncated to integer, 0<deg_freedom<=10^10

cumulative is logical, but numbers are accepted too

OpenFormula

CHISQDIST( Number x ; Number DegreesOfFreedom [ ; Logical Cumulative = TRUE() ] )

OOo3.2

is already implemented as CHISQDIST as OpenFormula defines

no constraints for x

DegreesOfFreedom are truncated as in Excel, DegreesOfFreedom > 0, no fixed upper bound for DegreesOfFreedom

Comment

Directly mapping is possible in import.

CHISQ.DIST.RT

follow-up version of CHIDIST

Excel

CHISQ.DIST.RT(x,deg_freedom)

1 < deg_freedom < 10^10, deg_freedom is truncated to integer

OpenFormula

LEGACY.CHIDIST( Number x ; Number DegreesOfFreedom )

DegreesOfFreedom is positive interger

OOo3.2

CHIDIST(Number; degrees_freedom)

no constraints, degrees_freedom is truncated to integer

Comment

Direct mapping is possible.

CHISQ.INV

new

Excel

CHISQ.INV(probability,deg_freedom), both arguments required

0<=probability<1,

1<=deg_freedom<=10^10, deg_freedom is truncated to integer

OpenFormula

CHISQINV( Number p ; Number DegreesOfFreedom )

0<=p<1

DegreesOfFreedom is a positive integer

OOo3.2

is already implemented as CHISQINV as OpenFormula defines

0<=p<1

DegreesOfFreedom is truncated as in Excel, DegreesOfFreedom > 0, no fixed upper bound for DegreesOfFreedom, but calculation might not converge

Comment

Directly mapping is possible in import.

CHISQ.INV.RT

follow-up version of CHIINV

Excel

CHISQ.INV.RT(probability,deg_freedom)

The iterative search technique is limited to 64 iterations

deg_freedom is truncated to integer

OpenFormula

LEGACY.CHIINV( Number p ; Number DegreesOfFreedom )

OOo3.2

CHININV(number; degrees_freedom)

degrees_freedom is truncated to integer

Comment

direct mapping is possibel

CHISQ.TEST

follow-up version of CHITEST

Excel

CHISQ.TEST(actual_range,expected_range)

OpenFormula

LEGACY.CHITEST( ForceArray Array A ; ForceArray Array E )

The function was introduced in OpenFormula to be compatible with Excel and OpenOffice.org.

OOo3.2

CHITEST(Date_B; data:E)

Comment

direct mapping possible.

CONFIDENCE.NORM

follow-up version of CONFIDENCE

Excel

CONFIDENCE.NORM(alpha,standard_dev,size)

size is truncated to integer

OpenFormula

CONFIDENCE( Number alpha ; Number stddev ; Number size )

is same as NORMINV(1 - alpha / 2; 0; 1) * stddev / SQRT (size)

OOo3.2

CONFIDENCE(alpha; STDEV; size)

size is truncated to integer

implemented as defined in OpenFormula

Comment

Direct mapping is possible.

CONFIDENCE.T

new

Excel

CONFIDENCE.T(alpha,standard_dev,size)

calculates

OpenFormula

no suitable function defined

OOo3.2

no suitable function implemented

Comment

Direct mapping is impossible, because no suitable function exists in OpenFormula.

For implementation is possible, where TINV is the already implemented function in OOo.

COVARIANCE.P

follow-up version of COVAR

Excel

COVARIANCE.P(array1,array2)

a pair with one text, logical or empty is ignored

OpenFormula

COVAR( ForceArray Array n1 ; ForceArray Array n2 )

OOo3.2

COVAR(Data_1; Data_2)

Comment

Direct mapping possible.

COVARIANCE.S

new

Excel

COVARIANCE.S(array1,array2)

a pair with one text, logical or empty is ignored

OpenFormula

no such function defined

OOo3.2

not implemented

Comment

Direct mapping impossible, because a suitable function not contained in OpenFormula.

Implementation would be easy, just copy code from COVAR and use different denominator.

ECMA.CEILING

new

Excel

ECMA.CEILING(number, significance), all arguments are required

Comment

This function is mentioned in chapter 2.2.3 of [MS-XLSX]: Excel Extensions to the Office Open XML SpreadsheetML File Format (.xlsx) Specification [2], but it is not listed in the help inside Excel 2010 Beta.

It can be handled as CEILING.PRECISE.

EXPON.DIST

follow-up version of EXPONDIST

F.DIST

new

Excel

F.DIST(x,deg_freedom1,deg_freedom2,cumulative), all arguments are required

x>=0

deg_freedom1,deg_freedom2 are truncated to integer, both >=1

cumulative is boolean, but numbers are accepted too

OpenFormula

FDIST( Number x ; Number r1 ; Number r2 [ ; Logical Cumulative = TRUE() ] )

no constraint for x

r1 and r2 are positive integers

OOo3.2

not implemented

Comment

This left tailed version has the same name as the right tailed version of OOo2. Because the version OOo2 cannot interpret namespaces, a simple solution is not possible.

For the implementation I have already written a patch. It would only need to be adapted to the actual code. CDF is implemented with a call of ScInterpreter::GetBetaDist, density function is calculated directly or via ln, if directly calculating not possible.

Please discuss solutions on the discussion page.

F.DIST.RT

follow-up version of FDIST

F.INV

new

F.INV.RT

follow-up version of FINV

F.TEST

follow-up version of FTEST

FLOOR.PRECISE

new

Excel

FLOOR.PRECISE(number, [significance]), significance has default 1

OpenFormula

FLOOR( Number N [ ; [ Number significance ] [ ; Number mode ] ] )

nested rules, round toward negative infinity for mode omitted or zero, round toward zero in the other cases.

N and significance must have the same sign if not 0

OOo3.2

FLOOR(number; significance [; mode])

Significance is required, Issue 102957 is still open, no decision about the patch there up to now.

Comment

Because OpenFormula requires, that N and significance have the same sign, a direct mapping is not possible.

FLOOR.PRECISE(number) —with second argument missing— can be mapped to FLOOR(number). To this a corrected version in OOo is needed, so that a missing second argument no longer results in an error.

With a given second argument, the composed formula FLOOR(number; SIGN(number)*ABS(significance)) would give the same result. But the question is, whether such change should be done at all.

see also CEILING.PRECISE

GAMMA.DIST

follow-up version of GAMMADIST

GAMMA.INV

follow-up version of GAMMAINV

HYPGEOM.DIST

follow-up version of HYPGEOMDIST

ISO.CEILING

new

Excel

ISO.CEILING(number, [significance])

significance has default value 1, always the absolute value of significance is used

OpenFormula

CEILING( Number N [ ; [ Number significance ] [ ; Number mode ] ] )

nested rules, round toward positive infinity for mode omitted or zero, round away from zero in the other cases.

N and significance must have the same sign if not 0.

OOo3.2

CEILING(Number; Significance [; Mode])

Significance is required, Issue 102957 is still open, no decision about the patch there up to now.

Comment

Because OpenFormula requires, that N and significance have the same sign, a direct mapping is not possible.

ISO.CEILING(number) —with second argument missing— can be mapped to CEILING(number). To this a corrected version in OOo is needed, so that a missing second argument no longer results in an error.

With a given second argument, the composed formula CEILING(number; SIGN(number)*ABS(significance)) would give the same result. But the question is, whether such change should be done at all.

This function can be used with the name CEILING.PRECISE in Excel 2010 Beta too. It seems that CEILING.PRECISE is the favorite name, because ISO.CEILING works in EXCEL 2010 Beta, but is not listed in the available functions there. In addition an ISO.FLOOR does not exist, but only a FLOOR.PRECISE.

LOGNORM.DIST

follow-up version of LOGNORMDIST

LOGNORM.INV

follow-up version of LOGINV

MODE.MULT

new

Excel

MODE.MULT((number1,[number2],...])

array function, the target cells have to be selected.

The help does not mention, in which order the values are returned. Tests show, that values start with smallest cell position in order row, then column.

OpenFormula

no suitable function defined

OOo3.2

no suitable function implemented

Comment

A suitable function is neither defined in OpenFormula nor implemented in OOo3.2. So import is impossible.

MODE.SNGL

follow-up version of MODE

Excel

MODE.SNGL(number1,[number2],...])

The help does not mention, which value is returned, if there are more than one value with the same largest frequency. Tests show, that it is the values with smallest cell position in order row, then column.

up to 254 arguments

OpenFormula

MODE( { ForceArray NumberSequence N }+ )

If there are more than one value with the same largest frequency, the smallest one is returned.

OOo3.2

MODE(Number1; Number2; ...Number30), each argument can be a number or a range, only the first argument is requiered.

Comment

In simple cases the function returns the same result in Excel as in OOo. But in case of more than one value with the same largest frequency the result might differ.

The function can only be mapped on import, if there are not more than 30 arguments in Excel.

This problems are not new for MODE.SNGL, but exist already for the old MODE version.

NEGBINOM.DIST

follow-up version of NEGBINOMDIST

NETWORKDAYS.INTL

new

Excel

NETWORKDAYS.INTL(start_date, end_date, [weekend], [holidays])

The argument weekend can be given as number 1..17 or as string containing seven 0 or 1, for example 1100000.

OpenFormula

NETWORKDAYS( DateParam Date1 ; DateParam Date2 [ ; [ DateSequence holidays ] [ ; LogicalSequence workdays ] ] )

OOo3.2

NETWORKDAYS((Start_date; End_date [; Holidays])

Comment

The existing function NETWORKDAYS(Start_date; End_date; Holidays) has to be expanded to accept a forth argument, see Issue 71059 . In addition it needs the capability to use empty arguments indicated by ;; to skip the argument holidays.

The syntax for the argument holidays is the same in all three versions. The argument workdays has the type LogicalSequence in OpenFormula, that is a term of kind {1,1,0,0,0,0,0}, but has the type string in Excel. The first position is attributed to Sunday, but in Excel the first position is attributed to Monday. If the argument weekend is given as constant, mapping is possible, but need some care in converting the argument value. If the argument weekend is given as reference, a direct mapping is not possible.

see also WORKDAY.INTL

NORM.DIST

follow-up version of NORMDIST

NORM.INV

follow-up version of NORMINV

NORM.S.DIST

follow-up version of NORMSDIST

NORM.S.INV

follow-up version of NORMSINV

PERCENTILE.EXC

new

PERCENTILE.INC

follow-up version of PERCENTILE

PERCENTRANK.EXC

new

PERCENTRANK.INC

follow-up version of PERCENTRANK

POISSON.DIST

follow-up version of POISSON

QUARTILE.EXC

new

QUARTILE.INC

follow-up version of QUARTILE

RANK.AVG

new

RANK.EQ

follow-up version of RANK

STDEV.P

follow-up version of STDEVP

STDEV.S

follow-up version of STDEV

T.DIST

new

Excel

T.DIST(x,deg_freedom, cumulative), all required

deg_freedom is truncated to integer

cumulative is boolean, but numbers are accepted

OpenFormula

Currently only a right tail version is defined. But that will likely change till specification is final.

OOo

not implemented

Comment

An implementation of a left tail type and of a density function does not exist yet. The CDF is easily implemented via a call to ScInterpreter::GetBetaDist. The term of the density function does not look difficult and might be calculated directly or with ln in case of intermediate over/underflow.

Because the OpenFormula definition might change from the current right tail version to left tail, the implementation should be postponed.

T.DIST.2T

follow-up version of TDIST

T.DIST.RT

follow-up version of TDIST

T.INV

new

T.INV.2T

follow-up version of TINV

T.TEST

follow-up version of TTEST

VAR.P

follow-up version of VARP

VAR.S

follow-up version of VAR

WEIBULL.DIST

follow-up version of WEIBULL

WORKDAY.INTL

new

Excel

WORKDAY.INTL(start_date, days, [weekend], [holidays])

The argument weekend can be given as number 1..17 or as string containing seven 0 or 1, for example 1100000.

OpenFormula

WORKDAY( DateParam Date ; Number Offset [ ; [ DateSequence Holidays ] [ ; LogicalSequence Workdays ] ] )

OOo

WORKDAY(Start_date; Days [;Holydays])

Comment

The existing function WORKDAY(Start_date; End_date; Holidays) has to be expanded to accept a forth argument, see Issue 71059 . In addition it needs the capability to use empty arguments indicated by ;; to skip the argument holidays.

The syntax for the argument holidays is the same in all three versions. The argument workdays has the type LogicalSequence in OpenFormula, that is a term of kind {1,1,0,0,0,0,0}, but has the type string in Excel. The first position is attributed to Sunday, but in Excel the first position is attributed to Monday. If the argument weekend is given as constant, mapping is possible, but need some care in converting the argument value. If the argument weekend is given as reference, a direct mapping is not possible.

see also NETWORKDAYS.INTL

Z.TEST

follow-up version of ZTEST

Synopsis as Table