GNOME Bugzilla – Bug 607086
permit extending existing autofilters to new columns
Last modified: 2010-07-05 22:35:29 UTC
I noticed that the filtering seems to break down when you have multiple autofilter columns, apply some filtering, and then later on change your mind and change one of the columns' filtering value: it then filters only that column, as if it had reset all the other auto filters. Not sure if what I'm saying here is clear, if you need I could perhaps screencast this.
I think it would be very helpful if you could explain this in more detail!
There were a bunch of autofilter fixes in 1.9.4. Which exact version are you using?
Did you _create_ the whole autofilter at once or did you first add an autofilter to some columns and later another autofilter to other columns?
This is gnumeric 1.9.9. I added the autofilters one column by one, but going back to change already-filtered columns is what messes things up, afaics.
I can replicate that with current git. A workaraound of course is to add all the autofileter at once to all the columns. If you do that you apparently can change all columns and things stay in sync.
Looks like you were faster than me for reproducing it. Anyway, for the record/an exact demonstration of the problem, download this video: http://jeff.ecchi.ca/public/bug-gnumeric-607086.ogv
I guess this is really not a bug but a feature. The separately created autofilters act independently. They don't even have to act on exactly the same row.
(12:31:13 PM) jody: aguelzow: the filters should act in concert (12:32:00 PM) aguelzow: Can they act on different rows? (12:33:54 PM) jody: The filter in each column can filter a row based only on the content of that column but can filter multiple rows (12:34:10 PM) aguelzow: currently we can have a filter acting on rows 1 through 10 and another on rows 1 to 20. (12:34:31 PM) jody: hang on a sec, lets define terminology (12:35:03 PM) jody: An auto filtered range will have a set of individual columns (12:35:21 PM) jody: are you talkign about multiple columns influencing the same rows (12:35:44 PM) jody: or of having multiple autofiltered ranged in 1 sheet that impact the same same set of rows ? (12:36:10 PM) aguelzow: I am talking about 2 auto filtered ranges: A1:B10 and C1:C20 (12:36:20 PM) jody: ok, understood (12:36:25 PM) jody: that is not a valid state (12:36:44 PM) jody: XL only supports 1 auto filter per sheet (12:37:10 PM) aguelzow: SO am I correct that "two autofilters should act either on exactly the same rows (and cooperate) or on completely disjoint rows." (12:37:25 PM) jody: I never tried to support cross talk between multiple autofilter ranges (12:37:27 PM) jody: correct. (12:37:39 PM) jody: within the range they need to cooperate (12:37:51 PM) jody: outside the range we should complain and fail (12:37:54 PM) jody: on creation (12:38:34 PM) jody: I'd weaken your statement even further (12:38:49 PM) jody: autofilters can only act on disjoint rows (12:38:56 PM) aguelzow: So should we be able to create a filter for A1:C10 and another for C30:E:50? (12:39:09 PM) jody: agreed (12:39:37 PM) jody: but no filters that intersect rows 1:10 or 30:50 (12:40:26 PM) aguelzow: I think that if a user defines an autofilter for A1:B10 and then one for C1:D10 we should end up with a single autofilter for A1:D10. correct? (12:41:05 PM) jody: In the special case of extending an existing filter that seems reasonable (12:41:13 PM) ***aguelzow should not use colon followed by C (12:41:15 PM) jody: but a1:B10 D1:D10 (12:41:39 PM) jody: seems dangerous (12:41:55 PM) aguelzow: should give a1:d10 I would think (12:42:16 PM) jody: An error seems more appropriate (12:42:28 PM) jody: A1:B10 Y1:Z10 (12:42:53 PM) jody: Doesn't seem like anything we really want to do (12:43:09 PM) jody: and A1:B10 C1:D11 seems right out (12:43:29 PM) aguelzow: yeah A1:B10 C1:D11 has no reasonable behaviour (12:43:53 PM) ***aguelzow wonders how to switch off smilies in pidgin (12:45:32 PM) aguelzow: So creation of an adjacent autofilter with the same rows (or a single row at the top) should extend the existing autofilter. (12:46:15 PM) aguelzow: Creation of an autofilter on a disjoint set of rows, creates a second autofilter (12:46:26 PM) aguelzow: Otherwise an error is reported. (12:46:58 PM) jody: Seems reasonable (12:48:04 PM) aguelzow: Indeed. That bug report came from multiple adjacent autofilters. This way we never get into that stage. (12:50:40 PM) jody: I can't think of reasonable semantics for overlapping autofilters (12:53:47 PM) aguelzow: Neither can I. (12:54:05 PM) jody: Changing the content in the filtered region does not force a re-application of the filters (12:54:25 PM) jody: only a change in the filter triggers a full re-assesment (12:54:39 PM) jody: if we allowed overlaps (12:55:07 PM) jody: we'd end up triggering the filter even if there is only partial overlap
The bug portion of this report is fixed. One cannot any longer create non-cooperating auto filters in the same or intersecting row ranges. This report remains open as an enhancement request to allow extending existing autofilters to additional (adjacent) columns.
This problem has been fixed in the development version. The fix will be available in the next major software release. Thank you for your bug report.