You can make a difference in the Apple Support Community!

When you sign up with your Apple Account, you can provide valuable feedback to other community members by upvoting helpful replies and User Tips.

Looks like no one’s replied in a while. To start the conversation again, simply ask a new question.

Numbers broken in IOS 18.0 and MacOS 15.0 Sequoia

On Thursday September 26th 2024, I viewed one of my Numbers spreadsheets in iCloud with my iPhone 12 Pro Max running IOS 18.0. For only about 2 or 3 seconds, it correctly displayed 3 adjacent cells (in blue) in the same row as follows:

   09/25/24 09/26/24 09/26/24


Then, AFTER the first 2 or 3 seconds, these 3 cells displayed the Numbers icon for an error in the formula in a cell. Each of these 3 cells contains a formula that refers to the cell in the same column that is 2 rows before it. The error indicator is a red triangle with a white exclamation mark in the middle of the triangle.


In addition, all other cells containing a formula in the same row as these 3 cells (each of which likewise contains a formula that refers to the cell in the same column that is 2 rows before it), plus all other cells containing a formula in the spreadsheet after this row to the end of the spreadsheet, displayed correct results for only about 2 or 3 seconds, before displaying the Numbers error icon in these cells.


This NEVER HAPPENED WITH THE SAME SPREADSHEET when viewing it in iCloud prior to updating my iPhone to IOS 18.0.


I immediately opened the same spreadsheet in iCloud with Numbers on my MacBook Pro running MacOS Catalina Version 10.15.7. The spreadsheet displayed normally and without any Numbers error indicators.


I also immediately opened the same spreadsheet in iCloud with Numbers on my other MacBook Pro running MacOS Sequoia 15.0. In this case, the same problems happened as with viewing this spreadsheet on my iPhone, but far more quickly due to a much faster environment. As a result, when the spreadsheet opened on my MacBook Pro running MacOS Sequoia 15.0 I saw the data in the above three cells for just a fraction of a second, immediately after which Numbers displayed its error icons in these cells.


This NEVER HAPPENED WITH THE SAME SPREADSHEET when viewing it in iCloud prior to updating my other MacBook Pro to MacOS Sequoia 15.0.


These two operating system updates occurred immediately upon being notified by my devices that these updates were available, which happened no more than about 1 week prior to observing these problems on Thursday September 26th 2024.


As a committed Apple aficionado since March 2008, I am aware that Apple typically does not release bug fixes for an operating system update until a minimum of about 3 to 4 weeks after the update first becomes available. Over the years, this seems to have been because it may just simply take about 3 to 4 weeks after product release for bugs to appear “in the wild”. And nothing seems to smoke-out software errors better than use by the end user.



iPhone 12 Pro Max, iOS 18

Posted on Sep 28, 2024 9:42 PM

Reply
14 replies

Sep 29, 2024 8:24 PM in response to SGIII

SGIII: You should read the message I left for user Camelot on Sep 29, 2024 at 10:41 PM EDT


This will address, and may even answer, your questions about so-called "volatile" formulas.


But I will repeat here, because it obviously needs to be repeated (and quite often), that these problems never occurred until I updated my iPhone 12 Pro Max to IOS 18.0, and my MacBook Pro which I purchased NEW in April 2021 to MacOS 15.0 Sequoia.


Moreover, these problems have never occurred on my much older MacBook Pro which I purchased NEW in June 2013. For some years now, it has been running the last possible MacOS version that was available for it, which is MacOS 10.15.7 Catalina.


Not only that, I also have Microsoft Excel For Mac on BOTH of my MacBook Pros. Among other uses, I use Mac Excel to maintain FUNCTIONALLY EQUIVALENT versions of the spreadsheet that Numbers is currently mishandling on my iPhone running IOS 18.0 and MacBook Pro running MacOS 15.0 Sequoia.


I have never had these problems with the Mac Excel versions of this spreadsheet.

Sep 28, 2024 10:51 PM in response to regular_expressions_rule

There are exactly 10,125 rows as of right now in the spreadsheet BEFORE the first row that Numbers is mishandling. These rows BEFORE the mishandled row are comprised of about 60% rows which have at least 1 cell with a formula, and 40% rows which are strictly comment lines with no formulas. Nearly all of the rows in the 60% contain cells which are performing some rather complex calculations involving the OFFSET(), INDIRECT() and ROW() functions. NUMBERS IS PROCESSING ALL OF THESE 10,125 ROWS WITHOUT ERROR. Clearly, some type of flaw has been introduced into Numbers under IOS 18.0 and MacOS 15.0 Sequoia that is causing it to mishandle the rows from the row with the 3 cells described above through to the end of the spreadsheet.

Sep 29, 2024 11:07 AM in response to SGIII

SGIII: I powered off my iPhone 12 Pro Max running IOS 18.0 and MacBook Pro running MacOS Sequoia 15.0, and left them off for 1 minute.


Then, I powered-up the MacBook Pro and viewed the spreadsheet with Numbers in iCloud. The same problems persist.


Then I powered-up the iPhone and viewed the spreadsheet with Numbers in iCloud. This time the problems are WORSE. Previously, I would see the correct data in the 3 cells which I mentioned in my original post regarding these problems for about 2 or 3 seconds BEFORE Numbers inexplicably mishandled the formulas in them and displayed red/white error triangles in these cells, as well as all other cells AFTER them and to the end of the spreadsheet which contain formulas. But now, when viewing the spreadsheet with Numbers in iCloud on my iPhone, the mishandled cells IMMEDIATELY display the red/white error triangles. There is now NO DELAY in seeing these error triangles in the cells whose formulas are being mishandled.


There is an additional problem in this regard that has in the past been associated with IOS updates and the functioning of Numbers in iCloud. When I powered-up my iPhone but before viewing the mishandled spreadsheet earlier today, the main Numbers screen showed a file size of 2.3mb below the spreadsheet icon. PRIOR TO THIS, the displayed file size was 2.8mb. BUT I DID NOT CHANGE THE SPREADSHEET!!! This is the problem with Numbers in iCloud that I have seen previously. This file size problem would then be fixed in the next IOS update.


However, as I've said previously, ALL formulas in the spreadsheet PRIOR TO the ones being mishandled continue to be properly processed.


It should be clear beyond a reasonable doubt that there are flaws in Numbers that have been caused by the recent updates of IOS to 18.0 and MacOS to Sequoia 15.0.

Sep 29, 2024 12:08 PM in response to SGIII

SGIII: I just now submitted a bug report via 'Numbers > Provide Numbers Feedback'. Not surprisingly, the immediate reply was that they USUALLY do not reply directly to an individual report, and thanked me for 'helping to make Numbers better'. Yet, the online report form listed an optional field for my e-mail address, with the explanation that it may be used to communicate with me. NOT!!!


I seriously doubt that my bug report, ON ITS OWN, is going to be involved in repairing these VERY OBVIOUS FLAWS in Numbers under IOS 18.0 and MacOS Sequoia 15.0. Unless, that is, my bug report is somehow the proverbial straw that breaks the camel's back, which by some miracle of the universe is what finally convinces Apple that these are problems WORTHY of being fixed AS SOON AS POSSIBLE WITH AN OPERATING SYSTEM UPDATE.


Then again, why should a 3 trillion dollar company fix something that it may feel is not having, and is not likely to have, a sufficiently adverse impact on its profitability that it really should be concerned?

Sep 29, 2024 2:39 PM in response to regular_expressions_rule

Did you just add to the 10,000+ rows,


With each cell in the row performing multiple calculations that depend on the results of cells above, which in turn, depend on results of complex functions that depend on the cells above.


By my estimate (and I'm not a mathematician), that's something like 100,000,000,000 calculations that have to be done to evaluate the last cell.


Quite simply, I'm going to guess you're running out of memory.


The time delay is simply a facet of how many calculations the system can perform before it hits the limit.


Is it possible to trim the table, at least for testing, to see if the problem remains?


Additionally, in case I'm wrong, tapping the warning Triangle (at least on MacOS) should give you an error message that may add color as to what went wrong.

Sep 29, 2024 7:41 PM in response to Camelot

Camelot: Regarding your question "Did you just add to the 10,000+ rows"

I have been adding rows to the spreadsheet at the rate of about 4 to 7 rows every two or three days since I originally created it in June of 2013.


As of this writing, the spreadsheet consists of a grand total of 10,155 rows.


Your statement regarding the number of calculations "that have to be done to evaluate the last cell" aligns neither in theory nor actual fact with the reality of what is occurring in the spreadsheet.


The first 9,893 rows are the detail section of the spreadsheet. Of these first 9893 rows, no more than about 50% have a cell with a formula, it is just one cell, and it is ALWAYS IN THE SAME COLUMN (which is column G). This is the formula in the currently last such cell (G9892) in the spreadsheet, as copied from the formula editor and pasted here:

SUM(OFFSET(G9892,−2,0),F9892)−E9892

This means "sum the cell TWO ROWS ABOVE cell G9892 with cell F9892, and subtract from this sum cell E9892".


Of the approximately other 50% of the first 9893 rows, ALL OF THEM ARE COMMENT LINES. THE VAST MAJORITY (ABOUT 99%) OF THESE COMMENT LINES ARE STRICTLY TEXT WITH NO FORMULAS WHATSOEVER. THOSE THAT DO HAVE A FORMULA USE ONLY THE ROW() FUNCTION WITH NO PARAMETERS TO DISPLAY THE NUMBER OF THE ROW WHICH CONTAINS THE COMMENT LINE.


There have never been any problems with Numbers evaluating the cells in the detail section of the spreadsheet.


There had never been any problems with Numbers evaluating the cells in the spreadsheet AFTER the detail section until the recent updates of IOS to 18.0 and MacOS to 15.0 Sequoia. Since these updates are the only things that have changed relative to HOW the spreadsheet is processed, they must be the cause of the mishandling of the spreadsheet.


Consequently, I am uncertain of the relevance of your statements "Quite simply, I'm going to guess you're running out of memory" and "The time delay is simply a facet of how many calculations the system can perform before it hits the limit" to the reality of what is occurring in the spreadsheet.


Regarding your question "Is it possible to trim the table, at least for testing, to see if the problem remains?", I am not as of this writing aware of any possibility of trimming the table for any reason.


As for tapping on the warning triangle, the resulting message is woefully unhelpful and inadequate, stating only that "This formula can't reference its own cell, or depend on another formula that references this cell". It does not give any indication whatsoever as to HOW Numbers has made this determination. The reason for this is obvious: IT IS NOT IN FACT HAPPENING.


The first cell which contains a formula that Numbers is mishandling is A10126. This is the formula in cell A10126, as copied from the formula editor and pasted here:

OFFSET(INDIRECT("a"&ROW(cell),addr-style),−2,0,rows,columns)

where cell, addr-style, rows and columns REPRESENT optional parameters that are not in fact specified. This means the following: "the cell that is in column a TWO ROWS BEFORE THE ROW WITH THE CELL CONTAINING THIS FORMULA".

Sep 29, 2024 9:00 PM in response to regular_expressions_rule

Hey, you can get angry with me if you want, or not. Just like I can try to help, or not.


In addition, shouting at me (or anyone else here) doesn't help much. I'm working blind here since I don't have your spreadsheet to look at, so there is NO WAY I KNEW THAT 99% OF THE LINES ARE JUST COMMENTS!!!!!!!


That said,


> SUM(OFFSET(G9892,−2,0),F9892)−E9892

> This means "sum the cell TWO ROWS ABOVE cell G9892 with cell F9892, and subtract from this sum cell E9892".


Maybe I'm missing something, but I don't see the need to use OFFSET() here.


From your description, the same value can be achieve with:


= (G9890 + F9892) - E9892


Without the OFFSET(), and without the SUM(), and is likely orders of magnitude both faster, and lighter weight, eliminating a huge overhead in calculating your table.


> The first cell which contains a formula that Numbers is mishandling is A10126. This is the formula in cell A10126, as copied from the formula editor and pasted here:

> OFFSET(INDIRECT("a"&ROW(cell),addr-style),−2,0,rows,columns)

>where cell, addr-style, rows and columns REPRESENT optional parameters that are not in fact specified. This means the following: "the cell that is in column a TWO ROWS BEFORE THE ROW WITH THE CELL CONTAINING THIS FORMULA".


Maybe I'm wrong, but this just sounds like learning about relative vs. absolute references.


Why can cell A10126 say:


=A10124


That is the literal, standard method for referencing a cell.

If you're worried about the cell reference changing when you copy the cell/row/column, Numbers does that automatically - '=A10124' is a relative reference and would update to the same relative (i.e. 2 rows higher) cell if copied and pasted into a different cell.

Sep 29, 2024 9:12 PM in response to SGIII

SGIII: I think you misunderstand what I said.


I did not say that Numbers and Excel, AS RUNNING PROGRAMS OR APPLICATIONS, are functionally equivalent. I said that "I use Mac Excel to maintain FUNCTIONALLY EQUIVALENT versions of the spreadsheet . . .".


I was clearly saying that the SPREADSHEETS THEMSELVES, AS DISTINCT ENTITIES FROM THEIR PROCESSING PROGRAMS, are FUNCTIONALLY EQUIVALENT. They produce EXACTLY THE SAME RESULTS with their respective processing applications, except for the items Numbers is mishandling. Is not "EXACTLY THE SAME RESULTS" the essence of "FUNCTIONALLY EQUIVALENT"?

Sep 29, 2024 10:44 PM in response to Camelot

Camelot: The reason for using OFFSET() is as follows.


In my actual spreadsheet in question, there is cell G9982. In the row immediately AFTER it, there is a comment that has important information about the contents of row 9982.


When, as per usual, I insert new rows AFTER row 9983 (the comment line for row 9982), there must be a means for cell new cell G9984 to refer BACK TO cell G9982. But, the new rows always come from other rows in the spreadsheet, so there is no way for these other rows to know anything about their copy destination. Yet, as I just mentioned, new cell G9984 must be able to refer BACK TO cell G9982, without actually being aware of it as cell G9982. The only way this can be implemented is through the relative reference capabilities of OFFSET(), since I can tell OFFSET() to offset its cell reference by -2 rows, i.e., to the cell in the same column as OFFSET() that is 2 rows BEFORE it.

Sep 30, 2024 1:21 AM in response to regular_expressions_rule

Numbers is not designed to be functionally equivalent to Excel. Trying to make a Numbers document functionally equivalent to an Excel document is often not an efficient use of either tool. Use each for its own strengths.


If you've got OFFSET and INDIRECT functions in thousands of rows then you have your answer as to why your Numbers document is exhibiting sluggish behavior!


If you must use them that way in a large document then I'd say Numbers is not the best tool for you.


SG

Sep 30, 2024 10:31 AM in response to regular_expressions_rule

I still think you're overcomplicating this , but I get you're dealing with a complex spreadsheet that I can't see, so bear with me here.


From your earlier statement:


> The first cell which contains a formula that Numbers is mishandling is A10126. This is the formula in cell A10126, as copied from the formula editor and pasted here:

> OFFSET(INDIRECT("a"&ROW(cell),addr-style),−2,0,rows,columns)


This is GUARANTEED to fail.


Breaking down the statement, you're asking Numbers to calculate:


INDIRECT("A"& ROW()) using (presumably) A1-style references.


INDIRECT() builds a cell reference, then goes and gets the value of that cell.


You stated that this is the formula in cell A10126, so you're asking INDIRECT to go get the value of itself -> BOOM!


This may work if it's in another cell in that row, say G10126, where, the INDIRECT() looks up "A" & ROW(), which equates to A10126, if so, my question, is:


Why use INDIRECT() to find the value of cell A10126, when G10126 could just reference that cell directly?

=A10126


That may make sense if the value of A10126 is "G9982", so you're telling it where to go find the value... but

you also say:


> But, the new rows always come from other rows in the spreadsheet, so there is no way for these other rows to know anything about their copy destination.


Sure, that would affect relative references, where cell references would update to reflect the same relative cell, but why can't the source use absolute references?


For example, assuming cell G9000 has a formula that references G9982 (a cell 982 rows further down the table), and you copy row 9000 to row 11,000, Now cell G11000 will reference G11982 to maintain the same 982-row offset.


However, if cell G9000 references $G$9982, that's an absolute reference.

Now, copying row 9,000 to row 11,000 will still maintain a reference to G9982 due to the user of absolute addresses (as denoted by the $ signs)


so, is it possible (and again, bear with me here since I'm working blind) that your cell G10126, instead of OFFSET(INDIRECT(blah blah blah)) could just:


=$G$9982


Furthermore, something else you said that worries me..


you said that the first 9,893 rows are your "data rows" and you append rows to this table that reference the data rows.


To me, this sounds like the perfect use case for either a separate table or a separate sheet.


Use the data table as a semi-static table.

Have a separate table (either on the same sheet or on a separate sheet) that you use to build out the dynamic(?) part of your spreadsheet. Agreed, this is highly use-case dependent and I'm working blind, but it seems to be that you have two separate 'classes' of data, and that is often best handled by separating the sets rather than trying to keep them in a single table. This would often eliminate most of the relative references issues.

Numbers broken in IOS 18.0 and MacOS 15.0 Sequoia

Welcome to Apple Support Community
A forum where Apple customers help each other with their products. Get started with your Apple Account.