[Issue] Sort Ascending/Descending

[jblank43]

Since downloading the latest OOo 3.1 is have not been able to sort in ascending or descending order the columns in a spread sheet as I have been doing for years. I have highlighted the correct columns but it does not sort. Anyone know why this changed or how to correct it?

[FJCC]

The Sort function in my 3.1 works mostly as expected except for a bug related to relative references in formulas. Can you explain exactly what you are trying to do and what happens? You could even post an example file.

[jblank43]

I have four columns - date,business name, 1st payor, 2nd payor. I highlight all four colums and try to sort ascending. It takes the two payor colums and groups them individually but does not sort all four columns in date order. The dates are all intermingled. I have been using this for several years and have never had trouble sorting all four columns into date order.

Code: Select all

1       Meijer       32.04
22      Penny        56.05
10      Gas          22.07
15      Patz         15.06

After sort ascending, the Meijer and Patz will be sorted together and the Penny and Gas will be sorted together. Not sorted by date or company or anything.

Code: Select all

1      Meijer      32.04
15     Patz        15.06
22     Penny       56.05
10     Gas         22.07

[Dave]

I used your data with the "date" being simply a number [numerical], and they sort just fine.

David.

[FJCC]

I also tried your data and the sort worked as expected. Some inadvertent formatting might be the problem, as hinted by Dave. If you press CTRL + F8, cells that are numeric will turn blue and text will be black. If the formatting is as you expect, then you could try to reset your user profile as explained here. If that doesn't work, all I can suggest is posting an example file so others can try to replicate the problem on your actual data.

[jblank43]

Here is an actual portion of my spread sheet. I went back to previous months that sorted correctly and have checked the formulas in each column and see no problem. This has only happened since downloading OOo3.1 Any help is appreciated

Code: Select all

June			
11	Meijer               $6.38
6	 Patz	              $6.40
8	 Wiseway              $11.77
11	Lowe's               $27.76
4	 Town & Country       $28.19
4	 Tractor Supply       $49.20
11	Strack Van Til       $61.16

[acknak]

I went back to previous months... and have checked the formulas...

If your columns use formulas to produce some of the values in the range being sorted, it's very likely that the problem is the bug in 3.1 that FJCC mentioned in the first reply.

There's nothing to do for it except to go back to 3.0.1. The bug is already fixed, but the fix won't appear in a stable release until late August.

[csoren]

I went back to previous months... and have checked the formulas...

If your columns use formulas to produce some of the values in the range being sorted, it's very likely that the problem is the bug in 3.1 that FJCC mentioned in the first reply.

There's nothing to do for it except to go back to 3.0.1. The bug is already fixed, but the fix won't appear in a stable release until late August.

How do I go back to 3.0.1? I can't find it on the download page. I have a large spreadsheet with many relative references that get corrupted when I sort. (It's really too large and messy to use for an example, but I'll try to reproduce the problem in a simpler form.)

Since I've resaved the .ods using 3.1, will I be able to open it using 2.4 if 3.0.1 is no longer available?

I'll be glad to beta-test an 'unstable' version in the meantime, if that will help verify the fix.

Thanks for any suggestions,
Carolyn

[acknak]

Hmm, that's a good question. The sorting issue is already fixed and validated, but I can't seem to find the latest snapshot build anywhere. There seems to be some problem with the distribution system.

I hear that you can usually find older builds--for Windows anwyay--at archive sites such as filehippo.com.

[hughesjr]

Is build OOo-Dev OOO310_m13 ever going to show up on the mirrors?

[Villeroy]

File>New>Database...
[X] Connect to existing database
Type: <Spreadsheet>
Specify your spreadsheet
[X] Register database
Store the Base document.

Use the "query designer" to create a virtual table view with any filter, sort order, column order, column headers you like.
In Calc hit F4, get the virtual table (query) in the pseudo-database and drag it into any document (Writer or Calc).

[Tutorial] Using registered data sources in Calc
Storing all raw data in a database right from the start would be a clear advantage.


More workarounds: http://www.oooforum.org/forum/viewtopic.phtml?t=85055

[csoren]

I went back to 2.4 but am still finding & fixing cells corrupted because of sorting relative formulas in 3.1. Very hard to see the problem unless the value displayed is obviously out of whack, basically I am in the process of going thru every single cell to verify the formula. (I know, dumb of me to resave with 3.1 over my latest version, but I did not know about this problem until after I saved, went out & back in again. It's a very sneaky bug, hard to see.)

I will be a lot more cautious about 'upgrades' in future. I don't understand why known bugs are not published right away. There should be full disclosure on the download page so everyone can decide whether to try a new version or wait.

[Villeroy]

The problem is that the developers tried their best to implement better sorting capabilities matching several reasonable requests from the "community" (keyword "stable sorting"). The "community" downloaded thousands of copies of the release candidates just to own the latest shit without understanding what a release candidate is, without testing any of the new features thoroughly.
If the greedy masses would take a little bit of responsibility after downloading the beta-version, this would not have happened.
I have to admit that I downloaded, but never installed the first release candidate. I'm pretty shure that I would have noticed this bug.
Nevertheless, I think some developer did not do his job. How can one write a complete new sorting module for a spreadsheet and test it with constant values only.

[csoren]

Well, it's hard to be too grumpy since the Open Office team does such a great job, most of the time But there should be a stronger emphasis on reporting new features & known issues/bugs for new releases. I downloaded 3.1 because I thought it was the latest (mostly) stable version. If I had known it was a release candidate for beta-testing etc. I would have been more careful. (My fault for not investigating in advance.) And frankly 'sort' is not a new feature so without seeing a strongly-worded caution that sort had been rewritten and was not fully stable in 3.1, well, let's just say it took me by surprise.

I think the team does a great job. Let's just try to be more "open" about changes and software stability for new releases.

[hughesjr]

I am a lead developer for the CentOS project, so I understand the Open Source process. I also understand the difference between "Stable" and "Beta" or "RC" software. I also realize that all "Stable" software has bugs.

However, when a project marks the release as Stable, I expect certain things from it. One thing I expect is that a majorly used feature, when used as designed, does not corrupt data. If this was a new capability that was added, then that would be one thing; however, it is the sort function of a spreadsheet program. I can't think of a spreadsheet that I use for anything that is even slightly complicated that does not sort on a column with a reference to another cell. There is no way this should have not been seen by someone. It took me 5 seconds to see it was busted on a very simple spreadsheet. I broke 2 other spreadsheets testing to see if it was a problem.

This sort thing is a MAJOR issue. It not only is broken (understandable, things break), it is corrupting spreadsheets (not acceptable to release as stable, certainly not acceptable to leave released as stable once discovered). It changes the cell references in some columns when it doesn't sort, but not all columns. So if you were doing a sort and your formula contained a $E$8 ... when the sort is done, that cell might now $F$8 even though it should be $E$8. Some columns are sorting and others are not. Sometimes even in the same data row, some columns are sorted and others are not.

The only way to fix this (after you move back to real stable software) is to hand edit and verify each and every cell of your spreadsheet.

I only broke 3 spreadsheets ... 3 very important mission critical spreadsheets. However, I had backup copies and I was able to roll back my software and recover my spreadsheets. I did so without rolling this out to 1000 machines and costing my company millions of dollars (and me my job). I was able to do this because I use very conservative testing procedures to deploy updated software. Some people might not be so lucky. I know, their fault ... they should test any updates before deployment.

The bottom line is that Stable means Stable and Beta or RC means Beta or RC ...

The OpenOffice.org project needs to pull this release from the stable trees or they need to update the software to fix this bug well before August. At least they need to put out the Developer version that has supposedly been released for 2 weeks and is NOT on any mirror anywhere that has the fix. Every time I open OpenOffice 3.0.1 it tells me to upgrade to the latest stable version and now I have to field calls from 1,000 users on 1,000 machines whose software is telling them there is a newer stable version and wants them to install it. Only a couple of them have the rights and access to do that install, still ... its not stable and it should not continue to offer to upgrade people while it has this major issue.

This is a HUGE and MAJOR data corruption issue, IMHO it can not wait for normal procedures to be followed to correct it. I say this with full knowledge of what kind of impact that can have on an Open Source project. However, it is what it is, and that is a major snafu. Letting people knowingly corrupt their data without doing something, for more than 2 months, is totally unacceptable.

None of this should be taken to mean that the OpenOffice.org project is anything other than awesome. This software is excellent and it is the best office suite in the world by a wide margin. The project is amazing and the features added every cycle are astounding. None of those accolades change the fact that software which corrupts data while using one of its main features has to have an emergency fix released or be pulled from its released/stable status.

[acknak]

I happen to agree with you, but no developers will see any of this discussion. If you want the developers to see it, you'll have to post it to the mailing list(s): http://www.openoffice.org/mail_list.html. dev@openoffice.org would be a likely place.

[hesido]

@hughesjr:

I couldn't have said it better. I just lost 3 hours of work, and I have a spreadsheet to prepare until tomorrow morning (It would have been ready 2 hours ago), took me long enough to realize relative functions were not being sorted correctly.

This release (3.1) is a disgrace to what OpenOffice stands for. An OpenOffice stable build should be performing perfectly for basic spreadsheet functionality at enterprise level (like, how about sorting correctly?), not at the level that is corrupting the data of humble individuals such as myself. The industry standard and leader in the field could not have done this themselves, to undermine the OpenOffice project. The project have shot itself in the foot months ago, and it still is not bandaging it. Unbelieavable. Open office has just lost all its credit for me.

Now, you don't release a spreadsheet program that can't sort, even for free. However, if once you make that mistake, you should immediately fix that with a hotfix. If that is not feasible, how about returning to the previous stable builds? How about caring about your userbase? Why does the project look like it is totally out of control, and has lost direction? Why can we still download 3.1 that can't do basic sorting?

I have no more to say, and I have to find and wake up a friend with a laptop that has proper spreadheet program installed. 97% of the world knows what that program is.

[csoren]

I agree completely. I had already discovered the sorting bug in June (release 3.0 something..), then went back to 2.4, then earlier this week I made the mistake of TRUSTING that the sort bug had been fixed for the official release of 3.1. It is not a 'new' problem, it's been an outstanding issue for some time now and it is just incredible that 3.1 was released without notice of outstanding problems.

So for the 2nd time I have uninstalled 3.x and gone back to 2.4.

At least this time I remembered not to save over my original (2.4) spreadsheet!

[Villeroy]

I just lost 3 hours of work, and I have a spreadsheet to prepare until tomorrow morning (It would have been ready 2 hours ago), took me long enough to realize relative functions were not being sorted correctly.

At least you realized the error. It should be a matter of minutes to restore the relative references. There are some viable and compatible workarounds for the issue.

I had already discovered the sorting bug in June (release 3.0 something..)

#
Version 3.1 is the one and only version affected by this.

However, when a project marks the release as Stable, I expect certain things from it. One thing I expect is that a majorly used feature, when used as designed, does not corrupt data.

Many thousands of people downloaded the 2 beta-releases just in order to have the latest shit on their disks. Nobody discovered this problem. The bug does not corrupt data. Only formulas are distorted. This makes a huge difference.

[csoren]

Well, I don't know what you would call 'corrupted' ... when I have a large spreadsheet full of references =SUM(AK10:AM10) which, after a sort of all rows, are magically transformed into =SUM(AK10:AM25) or worse ... hmm, I'd have to call that corrupted. Not trivial to find or fix.

Each download should be accompanied by release notes describing every change and every known problem in that release. These notes should be posted clearly on the download page and users should be encouraged to read them before download.

[Villeroy]

Well, I don't know what you would call 'corrupted' ... when I have a large spreadsheet full of references =SUM(AK10:AM10) which, after a sort of all rows, are magically transformed into =SUM(AK10:AM25) or worse ... hmm, I'd have to call that corrupted. Not trivial to find or fix.

The application would corrupt your data if your input values ("data") where distorted, modified, converted.

You can repair this formula ("processing instruction") within seconds.

If you need to sort more often:
=SUM(INDEX($AK$1:$AM$65536;ROW()))

[Cambirder]

Is this still an issue? its fixed in 3.1.1.

[csoren]

That's great if its fixed -- but I got an announce to download the new version which turned out to be 3.1.0, where this problem is not fixed. (Up higher in this thread someone said it would be fixed in the August release ... but, apparently not.) And there are no obvious release notes with the download package to inform us of ongoing, known problems.

I don't see 3.1.1 available for download.

Release notes should be posted on the download page, clearly visible and available for review BEFORE downloading. "Known problems." Alternatively, a doc can be opened for the user after installation. Install shield can do that, I believe.

[acknak]

Who sent you the announcement? Can you attach the message or address here? I'm just curious--not that we can do anything about it.

This issue is fixed in OOo 3.1.1--people have been testing and verifying the fix for weeks--but OOo 3.1.1 has not been announced or released yet. It's scheduled to be released on 27 August. Release announcements are never made before the final version is available on the download.openoffice.org site, as well as all the mirror sites that carry it. The release notes are also available at that point.

You can try one of the snapshot builds, where the problem is fixed, if you want to: download.openoffice.org/next Just be aware that it's not a final release--you should be even more careful than usual to make backup copies of your files.

Unfortunately, there are lots of scams around that entice people to register or pay for "updates" that are actually the same thing that has been available for months, for free.

Never trust anything that's volunteered (i.e. that you didn't ask for) by someone you don't know--especially on the Internet.

[Cambirder]

I don't see 3.1.1 available for download.

If you want to try it out release candidate 2 can be found here.

http://download.openoffice.org/next/index.html

Not exactly a stable release, but its probable a lot less buggy than 3.1.0

[csoren]

Not sure why I downloaded 3.1.0 again, possibly it was something in the user support digest that got me thinking 3.1.0 was the version with this problem fixed. Doesn't matter. But I did look through the release notes for 3.1.0 again and I still see no mention of this known & very significant problem (sort being a very important function for a calc program...). There's a bug reporting system and I'm sure that OO developers know where to go to find known bugs for each release, but I don't see that information in an easily accessible spot for normal users. This should be corrected.

[acknak]

I sympathize with your frustration--I don't understand how this bug was missed or why it wasn't treated more seriously after it was found--but this is not the place to do anything about changing the process. No one who makes those decisions is likely to read this.

If you want to bring it up with the project leaders, post it on one of the project mailing lists http://www.openoffice.org/mail_list.html. I don't know which list would be most appropriate for discussing something like this--QA (Quality Assurance) might be a good place to start: http://qa.openoffice.org

However, I don't see how documenting bugs in the release notes is going to actually help anyone avoid the problem. Maybe some kind of easily accessible "errata" page would be helpful though, in cases like this. But, would anybody read that page before it was too late? Again, I'd have to say: probably not.

The issue tracker is the place to look if you want to know about bugs. If you don't want to deal with the issue tracker, then you can ask here.

[hesido]

I just lost 3 hours of work, and I have a spreadsheet to prepare until tomorrow morning (It would have been ready 2 hours ago), took me long enough to realize relative functions were not being sorted correctly.

At least you realized the error. It should be a matter of minutes to restore the relative references. There are some viable and compatible workarounds for the issue.

Problem and frustration was caused by the fact that there were thousands of lines and since I didn't want recalc at every edit, and that I wanted it to sort correctly(!), I copied the garbled results to another sheet page, and did a couple more filtering stuff. By the time I realized I was working with wrong values, I had to go 2-3 sheets back. It's easy indeed to fix it on certain occasions - it might also be harder to fix once you have set some things in stone - but as I said, sorting is the utmost basic functionality, and I do lots of sorting based on function results (take 2-3 variables into account, get a quantized value based on those results, sort accordingly.)

I'm glad things are being fixed but lack of QA such as the one in this case brings serious blow to the image of the project.

[acknak]

I guess I'll just keep repeating it: it doesn't do any good to complain about it here.

Of course, you're welcome to waste your time doing that (within limits); I know how it works: I'm as fond of complaining to all the wrong people as anyone else.

Ultimately, it comes down to a subjective decision as to how critical a particular bug is. Almost every bug has some potential for causing lost work, and every release has bugs in it. There's absolutely no guarantee that you won't lose some work because of a bug in OOo. Someone has to decide whether a bug found after release is critical enough to warrant the extra time and effort of making a new release just to fix the bug.

It seems like, In the past, minor (bugfix) releases have come out much faster than this one, like maybe a few weeks after the major release. I don't know why 3.1.1 was scheduled so long after 3.1 came out.

At the end of the day, we have to trust the developers to make the decisions and set the schedule. I'm reasonably comfortable with the way that's been handled.

If you want a substantive discussion of this issue, post a message on the QA list.