Comments on: The Formula for Failure http://www.robweir.com/blog/2007/07/formula-for-failure.html Thinking the unthinkable, pondering the imponderable, effing the ineffable and scruting the inscrutable Tue, 16 Mar 2010 23:56:05 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-1219 Anonymous Wed, 26 Sep 2007 05:20:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-1219 http://www.theinquirer.net/gb/inquirer/news/2007/09/25/math-bug-found-excel<br/><br/>"Several examples are shown, perhaps the simplest of which is the calculation ( 850 X 77.1 ), which should produce 65,535 but instead returns 100,000." http://www.theinquirer.net/gb/inquirer/news/2007/09/25/math-bug-found-excel

“Several examples are shown, perhaps the simplest of which is the calculation ( 850 X 77.1 ), which should produce 65,535 but instead returns 100,000.”

]]> By: That Guy http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-920 That Guy Fri, 20 Jul 2007 21:33:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-920 I am kind of surprised to not see this mentioned, but with these broken formula in the specification (and other obvious flaws), it will actually be quite easily provable that MSOffice does not faithfully follow the specification. This will mean there is not even ONE implementation of this 'standard' in the wild. I am kind of surprised to not see this mentioned, but with these broken formula in the specification (and other obvious flaws), it will actually be quite easily provable that MSOffice does not faithfully follow the specification. This will mean there is not even ONE implementation of this ’standard’ in the wild.

]]> By: Rob http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-875 Rob Fri, 13 Jul 2007 03:50:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-875 Anders, the error in ZTEST is the text that says, "where x is the sample mean...". It should say. "where x-bar is the sample mean". <br/><br/>As you illustrte with you HYPGEOMDIST comment, it is important that where mathematical notation is used that the variables of the equation are correctly and unambiguously mapped to the spreadsheet function's arguments. Anders, the error in ZTEST is the text that says, “where x is the sample mean…”. It should say. “where x-bar is the sample mean”.

As you illustrte with you HYPGEOMDIST comment, it is important that where mathematical notation is used that the variables of the equation are correctly and unambiguously mapped to the spreadsheet function’s arguments.

]]> By: Andres http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-874 Andres Fri, 13 Jul 2007 02:52:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-874 Are you sure about the ZTEST functions? What I've found in Wikipedia seems to show that OOXML is correct on this one...<br/>(I'm building a collection of OOXML shortcomings in spanish, and wanted to double check this) Are you sure about the ZTEST functions? What I’ve found in Wikipedia seems to show that OOXML is correct on this one…
(I’m building a collection of OOXML shortcomings in spanish, and wanted to double check this)

]]> By: Andres http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-873 Andres Fri, 13 Jul 2007 02:03:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-873 Another inconsistency (product of the rush to get the standard out) can be found in the formula for HYPGEOMDIST (3.17.17.145), where the formula is given, but its terms are not explained. It even contains a variable called lambda, which corresponds to the mean parameter of the function. Another inconsistency (product of the rush to get the standard out) can be found in the formula for HYPGEOMDIST (3.17.17.145), where the formula is given, but its terms are not explained. It even contains a variable called lambda, which corresponds to the mean parameter of the function.

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-872 Anonymous Thu, 12 Jul 2007 15:52:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-872 "...delusions of adequacy which on reflection must now seem unwarranted."<br/><br/>Microsoft technical capability in a nutshell.<br/><br/>-Wang-Lo. “…delusions of adequacy which on reflection must now seem unwarranted.”

Microsoft technical capability in a nutshell.

-Wang-Lo.

]]> By: Rob http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-871 Rob Thu, 12 Jul 2007 13:33:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-871 The general idea is that a spreadsheet document stores the cell contents (a literal value or a formula) as well as the last-calculated value. Most spreadsheets do it this way (Excel, 1-2-3, OpenOffice, etc.). The benefit is that this allows a lightweight viewer to read and display a spreadsheet without having the overhead of a full spreadsheet calculation engine. The lightweight viewers can ignore the formulas and just display the results that were saved with the file. Without this, you wouldn't have the ability to easily convert an ODF spreadsheet into an HTML table. The general idea is that a spreadsheet document stores the cell contents (a literal value or a formula) as well as the last-calculated value. Most spreadsheets do it this way (Excel, 1-2-3, OpenOffice, etc.). The benefit is that this allows a lightweight viewer to read and display a spreadsheet without having the overhead of a full spreadsheet calculation engine. The lightweight viewers can ignore the formulas and just display the results that were saved with the file. Without this, you wouldn’t have the ability to easily convert an ODF spreadsheet into an HTML table.

]]> By: Nathan T. Freeman http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-870 Nathan T. Freeman Thu, 12 Jul 2007 13:11:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-870 "So each cell with a value in an ODF spreadsheet now has two values. That is weird."<br/><br/>Why? What if I want the internal representation of a value to be the integer 6, but the display to be "six?" That wouldn't be a stylistic representation, but I can certainly imagine such a scenario. “So each cell with a value in an ODF spreadsheet now has two values. That is weird.”

Why? What if I want the internal representation of a value to be the integer 6, but the display to be “six?” That wouldn’t be a stylistic representation, but I can certainly imagine such a scenario.

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-869 Anonymous Thu, 12 Jul 2007 12:26:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-869 What i found in the little that ODf writes on spreadsheet cell is that it has seperated the presentation content and the value of the cell. <br/>So each cell with a value in an ODF spreadsheet now has two values. That is weird. It is a source of confusion and could easily lead to the creation of spreadsheets that have non-matching values. ODF does not state how to deal with non-matching values in spreadsheet cell. It would have been better if a cell had only one value and that style elements would decide on the presentation of that value. What i found in the little that ODf writes on spreadsheet cell is that it has seperated the presentation content and the value of the cell.
So each cell with a value in an ODF spreadsheet now has two values. That is weird. It is a source of confusion and could easily lead to the creation of spreadsheets that have non-matching values. ODF does not state how to deal with non-matching values in spreadsheet cell. It would have been better if a cell had only one value and that style elements would decide on the presentation of that value.

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-867 Anonymous Thu, 12 Jul 2007 07:51:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-867 Do I understand correctly that ISO is now asked to redefine mathematics? <br/><br/>The chosen standard form also makes it impossible to implement a spreadsheet with the mathematically correct version of these functions in the future. <br/><br/>If my Spreadsheet (or other) program wants to be mathematically correct, it will be unable to store it's results in an OOXML file as this doen not allow to flag the mathematically correct function. Any standard should at least allow to store the correct functions.<br/><br/>OOXML also redefines many definitions in other ISO standards (eg, ISO C).<br/><br/>I also do not understand people who claim that the ISO STANDARD should contain errors because this is easier for the APPLICATIONS. <br/><br/>Whether the stored SIN() is in degrees, 400degrees, or radians, or if CEIL() works the wrong way round is completely irrelevant for the application. It can present the USER with every representation it wants by just converting. <br/><br/>But this comes down to the same stupid discussion about MS Office being competely unable to convert date formats and just simply was "forced" to define 1900 as the beginning of all time and a leap year into an ISO standard. <br/><br/>I just think Rob was right, as confirmed by Bill Hilf, when he wrote that OOXML is nothing but a detailed description of one version of the inards (DNA sequence) of Office 2007. These Excel functions are just one proof in very many.<br/><br/>Winter Do I understand correctly that ISO is now asked to redefine mathematics?

The chosen standard form also makes it impossible to implement a spreadsheet with the mathematically correct version of these functions in the future.

If my Spreadsheet (or other) program wants to be mathematically correct, it will be unable to store it’s results in an OOXML file as this doen not allow to flag the mathematically correct function. Any standard should at least allow to store the correct functions.

OOXML also redefines many definitions in other ISO standards (eg, ISO C).

I also do not understand people who claim that the ISO STANDARD should contain errors because this is easier for the APPLICATIONS.

Whether the stored SIN() is in degrees, 400degrees, or radians, or if CEIL() works the wrong way round is completely irrelevant for the application. It can present the USER with every representation it wants by just converting.

But this comes down to the same stupid discussion about MS Office being competely unable to convert date formats and just simply was “forced” to define 1900 as the beginning of all time and a leap year into an ISO standard.

I just think Rob was right, as confirmed by Bill Hilf, when he wrote that OOXML is nothing but a detailed description of one version of the inards (DNA sequence) of Office 2007. These Excel functions are just one proof in very many.

Winter

]]> By: Queen Elizabeth http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-866 Queen Elizabeth Thu, 12 Jul 2007 02:35:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-866 Hmmm, that this discussion is taking place now, even though these functions have been the same for years, seems to hint at one thing: people do not actually use these functions.<br/><br/>You would be hard-pressed to find a mathematician or statistician who uses Excel at work.<br/><br/>I wonder why that is.<br/><br/>(Standardization, however, would a perfect juncture to remedy these problems and show that Excel is a serious tool. That is, if the emphasis of criticism is not to nix but to fix ISO OOXML.) Hmmm, that this discussion is taking place now, even though these functions have been the same for years, seems to hint at one thing: people do not actually use these functions.

You would be hard-pressed to find a mathematician or statistician who uses Excel at work.

I wonder why that is.

(Standardization, however, would a perfect juncture to remedy these problems and show that Excel is a serious tool. That is, if the emphasis of criticism is not to nix but to fix ISO OOXML.)

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-865 Anonymous Thu, 12 Jul 2007 01:21:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-865 > This complaint is stupid. It is well known that OpenXML is meant to document Excel's file format, not create a brand new format that does everything the right way.<br/><br/>Why would you standardize something wrong?<br/><br/>It's great that they document how and where it's wrong. It's bad that they want to standardize that wrongness.<br/><br/>Besides, like a poster above said, I thought OOXML, and by extension its formulas, was supposed to be "vendor neutral" not a Microsoft Office brain dump (which it certainly is). > This complaint is stupid. It is well known that OpenXML is meant to document Excel’s file format, not create a brand new format that does everything the right way.

Why would you standardize something wrong?

It’s great that they document how and where it’s wrong. It’s bad that they want to standardize that wrongness.

Besides, like a poster above said, I thought OOXML, and by extension its formulas, was supposed to be “vendor neutral” not a Microsoft Office brain dump (which it certainly is).

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-864 Anonymous Thu, 12 Jul 2007 01:11:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-864 Found this posted to Slashdot about Excel formula weirdness. HTH.<br/><br/>http://gcrc.ucsd.edu/biostatistics/Excel.pdf Found this posted to Slashdot about Excel formula weirdness. HTH.

http://gcrc.ucsd.edu/biostatistics/Excel.pdf

]]> By: Harlan Grove http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-863 Harlan Grove Wed, 11 Jul 2007 21:18:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-863 Re Eddie Edward's comment: 'any floating-point number can be stored without loss of precision in XML.' Definitions are essential. If he means a 64-bit IEEE double precision floating point number could be stored exactly as a string of 64 1s and 0s, he's right. But if he means arbitrary REAL numeric values, such as rational fractions of PI (as all integer degrees would be when converted into radians), he's wrong. PI, being a transcendental number, has no finite digit representation in any radix, so unless ODF handles infinitely long XML tags in infinitely long ODF files, there'd be potential loss of precision due to practical necessity. Re Eddie Edward’s comment: ‘any floating-point number can be stored without loss of precision in XML.’ Definitions are essential. If he means a 64-bit IEEE double precision floating point number could be stored exactly as a string of 64 1s and 0s, he’s right. But if he means arbitrary REAL numeric values, such as rational fractions of PI (as all integer degrees would be when converted into radians), he’s wrong. PI, being a transcendental number, has no finite digit representation in any radix, so unless ODF handles infinitely long XML tags in infinitely long ODF files, there’d be potential loss of precision due to practical necessity.

]]> By: Eddie Edwards http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-862 Eddie Edwards Wed, 11 Jul 2007 20:39:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-862 "the most commonly used angles have integral representations in degrees and so can be stored without loss of precision in XML that way"<br/><br/>Actually, any floating-point number can be stored without loss of precision in XML. Whether or not it is retrived correctly is another matter ... “the most commonly used angles have integral representations in degrees and so can be stored without loss of precision in XML that way”

Actually, any floating-point number can be stored without loss of precision in XML. Whether or not it is retrived correctly is another matter …

]]> By: Nathan T. Freeman http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-861 Nathan T. Freeman Wed, 11 Jul 2007 19:09:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-861 Regarding =-x^2...<br/><br/>Excel isn't even consistent with VisualBasic in this regard. http://msdn2.microsoft.com/en-us/library/fw84t893(VS.80).aspx<br/><br/>Which presumably means that a calculation done with formulas vs. VBA probably doesn't yield the same results.<br/><br/>Also, Hubert, all IBM need do is provide a mechanism to convert SmartSuite docs to ODF format consistently and cleanly. If they take responsibility for providing a bridge from a proprietary legacy to a current standard, that is VASTLY more useful than simply documenting a proprietary standard and leaving it up to the customer.<br/><br/>It's worth noting that the ODF-compliant editors that are bundled with Lotus Notes 8 read SmartSuite documents into ODF pretty darned well. Regarding =-x^2…

Excel isn’t even consistent with VisualBasic in this regard. http://msdn2.microsoft.com/en-us/library/fw84t893(VS.80).aspx

Which presumably means that a calculation done with formulas vs. VBA probably doesn’t yield the same results.

Also, Hubert, all IBM need do is provide a mechanism to convert SmartSuite docs to ODF format consistently and cleanly. If they take responsibility for providing a bridge from a proprietary legacy to a current standard, that is VASTLY more useful than simply documenting a proprietary standard and leaving it up to the customer.

It’s worth noting that the ODF-compliant editors that are bundled with Lotus Notes 8 read SmartSuite documents into ODF pretty darned well.

]]> By: Harlan Grove http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-860 Harlan Grove Wed, 11 Jul 2007 19:03:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-860 Specific functions should never have been specified. Function representation, to the extent it differred between built-in, user-defined via built-in programming/scripting facilities (e.g., VBA udfs in Excel), or add-on via DLLs (COM or otherwise), should have been in the FILE FORMAT spec. This leaves the question whether they should have specified arithmetic operator precedence. Myself, I think it was a mistake to include even that in the specification. Specific functions should never have been specified. Function representation, to the extent it differred between built-in, user-defined via built-in programming/scripting facilities (e.g., VBA udfs in Excel), or add-on via DLLs (COM or otherwise), should have been in the FILE FORMAT spec. This leaves the question whether they should have specified arithmetic operator precedence. Myself, I think it was a mistake to include even that in the specification.

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-859 Anonymous Wed, 11 Jul 2007 18:39:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-859 "Uh, no. OOXML does not stand for Object-Oriented XML, it stands for "Open Office XML." (Note that's two words, not one.)"<br/><br/>Actually, I think its "Office Open XML" - Rob did an article on the confusing name a while back. “Uh, no. OOXML does not stand for Object-Oriented XML, it stands for “Open Office XML.” (Note that’s two words, not one.)”

Actually, I think its “Office Open XML” – Rob did an article on the confusing name a while back.

]]> By: Anonymous http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-858 Anonymous Wed, 11 Jul 2007 18:11:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-858 the fact that ooxml can be scrutinized to this level is a testament to its technical verbosity. i welcome this healthy discussion :) the fact that ooxml can be scrutinized to this level is a testament to its technical verbosity. i welcome this healthy discussion :)

]]> By: Harlan Grove http://www.robweir.com/blog/2007/07/formula-for-failure.html#comment-857 Harlan Grove Wed, 11 Jul 2007 18:00:00 +0000 http://www.robweir.com/blog/2007/07/the-formula-for-failure.html#comment-857 To the people who don't understand the problem with =-3^2 returning 9:<br/><br/>Yes, if unary minus is considered part of the number, then it's natural to interpret it as (-3)^2 = 9. The problem comes when x = 3, and the formula is =-x^2. Now what should the answer be?<br/><br/>And let's go back to CUSTOMARY USAGE. I've gone through this in a few newsgroup threads, but Excel shares the semantics that =-x^2 returns 9 when x = 3 (or x = -3) with COBOL and maybe one SQL dialect. Nearly all other programming languages, non-Excel-clone spreadsheets, stats packages and DBMSs would return -9 since they adhere to BODMAS (i.e., textbook and academic standard) arithmetic operator precedence in which exponentiation happens before anything other than parentheses grouping.<br/><br/>Excel got this wrong on its own. StarOffice got it wrong because it copied Excel. OOo Calc gets it wrong because it copied StarOffice. No doubt now y'all are going to try to make this a virtue rather than admit the possibilities that (1) you could ever do anything wrong, (2) you were just trying to provide backwards compatibility, or (3) there's overwhelming practical value to doing the same thing as Excel does.<br/><br/>You want credibility, admit your own shortcomings. Try being HONEST ABOUT YOURSELVES for a change. To the people who don’t understand the problem with =-3^2 returning 9:

Yes, if unary minus is considered part of the number, then it’s natural to interpret it as (-3)^2 = 9. The problem comes when x = 3, and the formula is =-x^2. Now what should the answer be?

And let’s go back to CUSTOMARY USAGE. I’ve gone through this in a few newsgroup threads, but Excel shares the semantics that =-x^2 returns 9 when x = 3 (or x = -3) with COBOL and maybe one SQL dialect. Nearly all other programming languages, non-Excel-clone spreadsheets, stats packages and DBMSs would return -9 since they adhere to BODMAS (i.e., textbook and academic standard) arithmetic operator precedence in which exponentiation happens before anything other than parentheses grouping.

Excel got this wrong on its own. StarOffice got it wrong because it copied Excel. OOo Calc gets it wrong because it copied StarOffice. No doubt now y’all are going to try to make this a virtue rather than admit the possibilities that (1) you could ever do anything wrong, (2) you were just trying to provide backwards compatibility, or (3) there’s overwhelming practical value to doing the same thing as Excel does.

You want credibility, admit your own shortcomings. Try being HONEST ABOUT YOURSELVES for a change.

]]>