On Fri, 30 Sep 2011, Sven Schreiber wrote: > here's what seems to be an "interface problem" in some sense, although I > also have that déjà vu feeling as if we discussed this before... > > Anyway, I just chose "save data as / standard format", then selected a > subset of all variables to be saved, and entered a new filename > ("transfer1.gdt"). This operation as such went fine it seems -- when I > open that new file in a new gretl instance, everything looks good. > > However, in the main window of the old instance the line below the menu > bar (on Ubuntu) says "transfer1.gdt", but the window still lists all > variables, not the selected subset. So either the status information > saying "transfer1.gdt" must be wrong, or the display of all the old > variables, right? Or am I missing something and there's a > straightforward meaning to it (quite possible)? I agree, there's something wrong here. If you save only a subset of the currently defined series under some name, we probably shouldn't change the displayed datafile name to the new name. We could, I suppose, ask something like "Do you want to switch to the smaller dataset?" but that could create problems. There's no restriction on the selection of series to omit when saving a data subset, but there are restrictions on the series you're allowed to delete in the course of a session, depending on currently saved models and so on.

On Fri, 30 Sep 2011, Sven Schreiber wrote:

> Actually, let me go one step back and ask a basic question: why have the > variable-selection option at all when the user chooses "save as"? You just beat me to the punch! After thinking about it some more, that's what occurred to me. We're just making life difficult for ourselves (and potentially confusing to the user) by offering that option. > I mean an analogy of this behavior would be if in a text editor you > choose "save as" you are asked if you maybe want to edit paragraph > three some more before actually saving the file under a new name. > Deleting variables should probably be done afterwards within the new > datafile. Or, this is like an export-to-database action, which is also > already possible in gretl. Or beforehand: delete what you don't want, then choose "Save as".

> (And BTW, maybe save-as-database should be moved/changed to > export-to-database?) Yes, it should. These two things make a useful clean-up. The only immediate issue is what to do for 1.9.6. The first change could be made without disrupting any translatable strings but I think for the second one we'd want to put the database export option under File/export data as "_gretl database..." rather than just "_Database...". We already have the msgid "gretl database", so I could easily patch the .pot files for European languages, adding a translation for "_gretl database...", but I wouldn't want to try that for Chinese. Reactions from any translators who are listening?

On Fri, 30 Sep 2011, Allin Cottrell wrote: >> Actually, let me go one step back and ask a basic question: why have the >> variable-selection option at all when the user chooses "save as"? > > You just beat me to the punch! After thinking about it some more, > that's what occurred to me. We're just making life difficult for > ourselves (and potentially confusing to the user) by offering that > option. True. The removal of that intermediate step is a good thing IMO. However, this makes it much harder than it was to create a dataset containing a subset of the variables of an existing one via the GUI. If I'm not mistaken, the only way to achieve this is to delete what you don't want (which may be sometimes annoyingly difficult, if you have lots of saved models) and then save. Maybe we could add a "Save subset" menu entry?

On Sat, 1 Oct 2011, Sven Schreiber wrote: > Am 10/01/2011 08:17 AM, schrieb Riccardo (Jack) Lucchetti: >> On Fri, 30 Sep 2011, Allin Cottrell wrote: >> >>>> Actually, let me go one step back and ask a basic question: why have the >>>> variable-selection option at all when the user chooses "save as"? >>> >>> You just beat me to the punch! After thinking about it some more, >>> that's what occurred to me. We're just making life difficult for >>> ourselves (and potentially confusing to the user) by offering that >>> option. >> >> True. The removal of that intermediate step is a good thing IMO. >> However, this makes it much harder than it was to create a dataset >> containing a subset of the variables of an existing one via the GUI. If >> I'm not mistaken, the only way to achieve this is to delete what you >> don't want (which may be sometimes annoyingly difficult, if you have >> lots of saved models) and then save. Maybe we could add a "Save subset" >> menu entry? >> > > I'm not enthusiastic about new menu entries in general, but what about > adding an entry to the *context* menu that you get when you right-click > on one (or more) selected variable? ("save selected to new file") That > would be very intuitive IMHO. Maybe, but if you want to save a fairly substantial and possibly non-contiguous subset of the current series, I think it's easier to make the selection using the sort of dialog you get under the "export" options than by ctrl-click in the main window. We could implement Jack's suggestion by adding an item to the "Export data" menu; I don't think this would over-burden things. Or maybe this would help: instead of having a list of formats in the "Export data" submenu, we call the series selection dialog directly, but this dialog is modified to include a drop down list of formats (with CSV as the default?).

On Sat, 1 Oct 2011, Allin Cottrell wrote: > What we're discussing may not be a bug, exactly, but as Sven said at > the outset it's an "interface problem", and I'd like to have this > fixed in the 1.9.6 release. I have therefore implemented my last > suggestion above in CVS (and snapshots). Please let me know if you > reckon that what I've done is not right, or is not implemented > correctly. It is, and I like it, so thumbs up from me. One thing that has just occurred to me is: in the context of the "Export" window it would be nice to have an "Append/Overwrite" option. The ability to add one or two series to an existing dataset in one go would be definitely useful in some contexts. I don't mean to push it, I'd just like to hear you lot's opinion on this while we're on the subject.

Am 10/02/2011 01:51 AM, schrieb Allin Cottrell: > > What we're discussing may not be a bug, exactly, but as Sven said at > the outset it's an "interface problem", and I'd like to have this > fixed in the 1.9.6 release. I have therefore implemented my last > suggestion above in CVS (and snapshots). Please let me know if you > reckon that what I've done is not right, or is not implemented > correctly. > > What you should see: an item "/File/Export data...", which produces > a dialog with (a) a mechanism for selecting series, (b) a pull-down > thingy for selecting the export format (including .gdt) and (c) a > Help button that gives a certain amount of help. (The help item is > currently better in CVS than in the snapshots.) I think the "all" button is missing in the variable selection dialog! Given that export-to also serves to save in different formats, you don't always want to save just a subset.

Apart from that, while checking and partially translating that stuff, a thing about the terminology occurred to me: The first five menu entries talk about "data". To highlight the sharpened difference between save-as and export-to I suggest to say: * "Save datafile" (changed from Save data) * "Save datafile as" (changed from Save as)

And a related question. Why does it say "Close dataset" and not "Close datafile"?