open two instances of same databa side by side?

[fbungarz]

I am trying to find a way in PSU to display the label assignment panel and the metadata (image details) side by side to have more direct control how label mapping will be reflected in the metadata. Unfortunately it is not possible to display the two panels side by side (I added a not to http://bugs.idimager.com/view.php?id=2905 suggesting it might be a god idea).
Anyway, I thought it might work to open two instances of the same PSU database side by side and in one open application select the label panel to be displayed, in the second window only the image details panel. This would of course work only so long as one is working on the same set of images.
My questions:
(1) Is it even possible to open two instances of the exact same database (theoretically at least the server-based version should permit this, but even in the SQ-Lite version that should in theory be possible).
(2) If make a change in one instance, will that change be reflected in the second instance (i.e., I assign a label here and would expect the mapped XMP field to be filled there...)?
(3) Will this workflow increase likelihood of database corruption? [This is the reason, why I have not dared yet to try it out...]

Thanks,
Frank

PS:
Overall, if it worked, it would still be less than ideal. Nevertheless, it could perhaps at least be a workaround until PSU allows for a bit more "workspace flexibility". Many other programs allow for panels being more flexibly arranged: photoshop, capture one, lightroom etc. I understand that in IDI 5 the amount of different panels and how they could be re-arranged was fairly confusing and for a beginner particularly overwhelming. I generally applaud that this has been radically simplified in PSU. Still, in some instances a bit more flexibility might not hurt, particularly if one uses a desktop with lots of screen real-estate. Other user might perhaps prefer for example displaying "GEO Tag" and "Assign". Also: the current implementation is not strict anyway! If "Adjust" is opened, another second panel, "Assign", "Detail" or "GEO Tag" can be opened at the side (if desired I can log this as a feature request in Mantis...).

[vlad]

I am trying to find a way in PSU to display the label assignment panel and the metadata (image details) side by side to have more direct control how label mapping will be reflected in the metadata. Unfortunately it is not possible to display the two panels side by side (I added a not to http://bugs.idimager.com/view.php?id=2905 suggesting it might be a god idea).

Like Frank, I sometimes feel the need to see both those panels at the same time. Yet, let's also admit that this is a secondary need, mainly caused by some uncertainty on how label mapping will be reflected in the metadata (as mentioned by Frank). The primary, superseding need is predictable and reliable design and implementation of label settings, assignments and revocations. If I had any say, personally I would vote for spending time on label + metadata improvements and bug fixes rather than side-by-side panels.

Anyway, I thought it might work to open two instances of the same PSU database side by side and in one open application select the label panel to be displayed, in the second window only the image details panel. This would of course work only so long as one is working on the same set of images.

That's an interesting thought. Hert is obviously the most (only?) qualified person to reliably answer your questions bellow - all I could offer are my (unqualified) opinions.

(1) Is it even possible to open two instances of the exact same database (theoretically at least the server-based version should permit this, but even in the SQ-Lite version that should in theory be possible).

Well, I would expect that to be possible, but I don't remember if I have ever tried that. (Working with two instances on two different databases and sets of images is obviously no problem.)

(2) If make a change in one instance, will that change be reflected in the second instance (i.e., I assign a label here and would expect the mapped XMP field to be filled there...)?

In general, I would expect the answer to depend on the kind of change. In theory, any persistent storage change should reflect in both instances (as they use the same database). The practical question, however, is how much data caching does PSU employ - and which actions does that affect? Even when using only one instance, there are multiple cases where a manual refresh is needed - I would therefore expect to be even more cases where changes in one instance do not automatically reflect in the other. If you are prepared to employ manual refreshes, I would expect most changes to reflect in both instances. (Side remark: there is an operation for (re)reading metadata from images or sidecars - there is no dedicated operation, however, to reread metadata from the database itself.) Try it out and let us know!

(3) Will this workflow increase likelihood of database corruption? [This is the reason, why I have not dared yet to try it out...]

Granted, that's the 1 million $ question. Let me give you a 2 cents answer: I don't think the likelihood of DB corruption would decrease (Thankfully, I have never experienced DB corruption, but a couple of people did report striking bad luck - or weak spots - even when using only one instance. ) My expectation is that a realiable implementation allows multiple clients to smoothly operate at the same time on the same DB. Still, I could imagine assigning some labels in an instance and assigning or revoking some other labels in the second instance - with two lengthy (auto-)sync operations taking place at the same time on the same data. A scenario like that may be asking for trouble. In conclusion, I would probably try working with two instances at the same time (for a limited time and purpose) - but only if I first did a backup and if I then used some workflow common sense (e.g., use only one instance for changes - the other just for reading).

Overall, if it worked, it would still be less than ideal.

Let me state my own ideal: perfect design, bug free implementation, crystal clear docs (or none needed - everything should simply be super-intutive). Throw in a couple of revolutionary features too. I couldn't care less about configurable workspace.

(Seriously now, there are already a couple of annoying rough spots affecting the Details panel ergonomy. See, for example, #928 and #2677 in Mantis. I would be grateful if those got addressed first, before adding new modes of operation.)

I understand that in IDI 5 the amount of different panels and how they could be re-arranged was fairly confusing and for a beginner particularly overwhelming. I generally applaud that this has been radically simplified in PSU.

+1

Cheers,
Vlad

[fbungarz]

Hi Vlad,

The primary, superseding need is predictable and reliable design and implementation of label settings, assignments and revocations. If I had any say, personally I would vote for spending time on label + metadata improvements and bug fixes rather than side-by-side panels.

Just to clarify:
This is not just about controlling if the metadata are written to the correct fields. The reverse is also true: I often want have the metadata panel open to check if particular metadata fields need content, and, if not, assign the relevant label that fills that field for me. So, even if everything worked perfect without any bugs, part of "label + metadata improvements" as you put it is also streamlining workflow, making these panels more accessible and that (at least for me) includes the flexibility to open at least some panels side by side (obviously only where it makes sense: I already mentioned the GEO Tag as yet another panel to be used in conjunction with label assignments).

To me this request is as much about functionality and ease-of-use as it is about fixing bugs (that is why I did not open a separate ticket in Mantis, but mentioned it as part of your suggestions on templates for the label panel, which I think is a great idea, but obviously already requires an overhaul how that panel works...).

Cheers,
Frank

[vlad]

Hi Frank,

This is not just about controlling if the metadata are written to the correct fields. The reverse is also true: I often want have the metadata panel open to check if particular metadata fields need content, and, if not, assign the relevant label that fills that field for me.

Hmm, I see what you're saying. But do you ever (intend to) write metadata directly, without assigning a catalog label mapped to the appropriate field(s)? If not, then in theory you could simply check only which labels have already been assigned - and then assign others, as needed. (I understand, however, that does not necessarily remind you which (kind of) labels you have *not* assigned.)

So, even if everything worked perfect without any bugs, part of "label + metadata improvements" as you put it is also streamlining workflow, making these panels more accessible and that (at least for me) includes the flexibility to open at least some panels side by side (obviously only where it makes sense: I already mentioned the GEO Tag as yet another panel to be used in conjunction with label assignments).

Let me clarify: if side by side panels were available, I would probably use them (occasionally). So, I am not opposed to this feature (although I am opposed to a complex workspace with floating panels and what not...).

IIRC, you could see the assigned labels directly in the thumb info (in the collection viewer), without opening the Assign panel. Doesn't that help you? (I think you can see the assigned labels only for one image at a time - but the same is true anyway for the metadata fields in the Details panel.)

To me this request is as much about functionality and ease-of-use as it is about fixing bugs (that is why I did not open a separate ticket in Mantis, but mentioned it as part of your suggestions on templates for the label panel, which I think is a great idea, but obviously already requires an overhaul how that panel works...).

I am glad you like the template idea (and I can see how that would be particularly useful to you), but I'm not sure how it relates to the idea of side by side panels. But, once again, I am not opposed to it, as I could see its occasional benefits.

(Thnking out of the box for a moment:
maybe what you'd really want is a brand new split panel, with two sortable columns: labels vs mapped fields? Or, maybe the label assignment panel should show field-indexed labels as an advanced section/view? Or, maybe the Details panel could be enhanced such that a field value could be directly filled in by selecting and assigning (in place) one of the labels mapped to it? I don't know, these are just very rough ideas - maybe worthy, maybe junk; but, if we're suggesting significant features, let's be creative and consider alternatives.)

Cheers,
Vlad

[fbungarz]

Hi Vlad,

I am opposed to a complex workspace with floating panels and what not...

Yes, IDI was a bit like that and I agree PSU has a much more streamlined, simplified and elegant user interface [though I sorely miss the version panel, which likely it is never coming back ;-( ]

maybe what you'd really want is a brand new split panel, with two sortable columns: labels vs mapped fields?

Nope! Can't see that being useful for me at all.

I think you are looking at it from the wrong direction. I am interested in seeing if images lack metadata (= XMP), not if they lack particular labels! IF they lack metadata I want to quickly assign the labels to fill the metadata in.
More specifically:
Though I have mapped almost all my labels in in my custom taxonomic tree category (and configured these to automatically assign the parents) occasionally it happens that I overlooked mapping some labels. That's why I also much like your idea to batch map labels more efficiently. In any case, what I need is a way to check if metadata are there and if not assign them using the mapped labels (and if metadata fields are empty despite a label being present fix that I have forgot to correctly map it...)

Cheers,
Frank

[vlad]

I think you are looking at it from the wrong direction. I am interested in seeing if images lack metadata (= XMP), not if they lack particular labels! IF they lack metadata I want to quickly assign the labels to fill the metadata in.

Well, let me explain. Suppose you have a panel with two columns: metadata fields vs. lists of associated labels. (Each field may have a list of catalog labels mapped to it.) Furthermore, imagine that each item (field or label) is displayed in a box that could be colored in gray or in green (following the existing conventions in LAP). Now, let's say you select an image. Then, for any (applicable) metadata field, you would see:

- if the field is empty, then the field box would be grey and the associated labels would also be grey (assuming none of them has been assigned); at that point, you have the option to click a label on the right - that will assign the label and write its value to the field; visually, you would see both the metadata field and the assigned label turning green

- if the field is written and one of the corresponding label has been assigned, then both the field and that label would be colored green; (you can still click the label to unassign it, which should probably clear the field too - they would both turn grey)

- if the field is written in your image and no corresponding label is assigned, then there may be a problem; you would know that because you would see the field box colored in green, but all the labels on the right colored in grey

- if you see a green field and more than one green label associated to it, then that may be a problem too (unless the field allows multiple values)

Wouldn't that allow you to quickly evaluate (and change, if needed) the metadata vs label status of your image? Am I still off the mark?

Though I have mapped almost all my labels in in my custom taxonomic tree category (and configured these to automatically assign the parents) occasionally it happens that I overlooked mapping some labels.

And do you have any process for checking such a case? Or is it just a matter of spotting this...by chance?

That's why I also much like your idea to batch map labels more efficiently.

I understand. But do you generally have all your labels within a category or (sub)tree all mapped to the same field?

In any case, what I need is a way to check if metadata are there and if not assign them using the mapped labels (and if metadata fields are empty despite a label being present fix that I have forgot to correctly map it...)

In theory, I believe that the panel + coloring schema I sketched above would cover all these cases.(Think of it as a cross between the Details panel and the LAP, with the metadata-label mappings being made explicit; btw, using side by side panels doesn't necessarily make all the mappings explicit.) But maybe I'm missing something.

Cheers,
Vlad

[LifeIsLong]

While I have not read all of this discussion (sorry, it's pretty long!) I agree that being able to see the metadata and label assignment panels at the same time would be extremely helpful. Not just to confirm, but because it makes it quicker to process images. Flipping back and forth between panels just slows things down. Yes, I agree that PSU's interface is more streamlined that IDI's was, but I'd like to see that interface become more customizable so, while enabling new users to figure out what they need, it would help advanced users in their workflow. Dockable panels with multiple docking points would be my thought. Something like Visual Studio or TrackerPro's PDF editor.

[vlad]

While I have not read all of this discussion (sorry, it's pretty long!)

Yes, but LifeIsLong too! (...sorry, I couldn't resist)

I agree that being able to see the metadata and label assignment panels at the same time would be extremely helpful. Not just to confirm, but because it makes it quicker to process images. Flipping back and forth between panels just slows things down. Yes, I agree that PSU's interface is more streamlined that IDI's was, but I'd like to see that interface become more customizable so, while enabling new users to figure out what they need, it would help advanced users in their workflow. Dockable panels with multiple docking points would be my thought. Something like Visual Studio or TrackerPro's PDF editor.

I'm not going to enter now in a debate about interafces, especially as I'm not a UI/UX specialist. I'll simply stand by my suggestion that there is still room for (significant?) improvements within the existing interface. (That doesn't deny, however, the possible merits of a more flexible workspace - in particular, side by side panels).

[fbungarz]

Well, let me explain. Suppose you have a panel with two columns: metadata fields vs. lists of associated labels. (Each field may have a list of catalog labels mapped to it.) Furthermore, imagine that each item (field or label) is displayed in a box that could be colored in gray or in green (following the existing conventions in LAP). Now, let's say you select an image. Then, for any (applicable) metadata field, you would see:

- if the field is empty, then the field box would be grey and the associated labels would also be grey (assuming none of them has been assigned); at that point, you have the option to click a label on the right - that will assign the label and write its value to the field; visually, you would see both the metadata field and the assigned label turning green

- if the field is written and one of the corresponding label has been assigned, then both the field and that label would be colored green; (you can still click the label to unassign it, which should probably clear the field too - they would both turn grey)

- if the field is written in your image and no corresponding label is assigned, then there may be a problem; you would know that because you would see the field box colored in green, but all the labels on the right colored in grey

- if you see a green field and more than one green label associated to it, then that may be a problem too (unless the field allows multiple values)

Wouldn't that allow you to quickly evaluate (and change, if needed) the metadata vs label status of your image? Am I still off the mark?

Hi Vlad,
unfortunately all this does NOT work!
I have several thousand labels (the names of lichen species with their taxonomy) all mapped to the same custom metadata fields. So, there is simply no way to show all labels mapped to one field alongside the metadata box that it is mapped to!
Just to explain what I currently do:
(1) I take a picture of a lichen species, say "Aulaxina quadrangula (Stirt.) R. Sant."
(2) In my category Taxonomy I have the following labels:
Fungi
__Ascomycota
____Lecanoromycetes
______Ostropales
________Gomphillaceae
__________Aulaxina
____________Aulaxina quadrangula (Stirt.) R. Sant.
(3) These labels are mapped to custom XMP from the Darwin Core:
Fungi - dwc:kingdom
__Ascomycota - dwc:phylum
___Lecanoromycetes - dwc:class
_____Ostropales - dwc:order
_______Gomphillaceae - dwc:family
_________Aulaxina - dwc:genus
____________Aulaxina quadrangula (Stirt.) R. Sant. - dwc:scientificName
(4) There are x number of species with different scientiufic names in the genus Aulaxina, then there are lots of genera with other species in the family Gomphillaceae, several families in the Ostropales, etc. etc. - I guess you get the picture...
Now, when I assign the label "Aulaxina quadrangula (Stirt.) R. Sant." it has "assign its parents" checked and thus writes not only the label itself but also all the labels above in their corresponding fields. This way, I assign one label and get 7 metadata fields filled correctly in just one go. Very convenient, and one of the main reasons I have for years been using IDI, and now PSU. [And of course, if I take ten images of Aulaxina quadrangula, I can with one clikc write all that data to ten pictures at once.]

My problem with this: You can perhaps imagine that it took me quite a while to build a label tree with all these names and get them mapped correctly. If there had been some sort of a template that I could apply to the category taxonomy that would recognize at what kind of hierarchy a label would be mapped to a particular field - that would have been perfect!
I still now have the problem that occasionally I forgot to map a particular label and then the custom metadata do not get written.
In IDI it was no problem to keep both panels (assign & details) open side by side and if I forgot to have a particular label mapped it was easy to immediately see and fix that. In PSU that is no longer possible. I permanently have to change between the panels to be able to see if the metadata of a label that I assigned were correctly written.

BTW - these thousands of labels of course do not fit into the "assign" panel and of course I do not have them as part of a metadata set. I assign a label like " Aulaxina quadrangula (Stirt.) R. Sant." by typing "Aulax" into the search box, the label pops up and I hit the correct choice from the search box. Only then it shows up green in the panel and fills the metadata fields.

Cheers,
Frank

[vlad]

Frank, thanks for the explanations - I think I've got a fair idea of your taxonomy and some requirements around it.

You are obviously right that no panel (or table, or whatever) would be able to show at once all the (thousands of) labels mapped to a field. (Actually, I had already thought about that; for example, one could also have hundreds or thousands of person labels mapped to PersonInImage, etc.) My refinement would be that the metadata-label panel could show, for a field, only the label(s) that has/have already been asigned. If a field is empty, then you would not see any (assigned) label next to it, but you would be able to easily fill it in and assign a mapped label via a searchable textfield (as already implemented in LAP) - or perhaps by invoking the full label tree. Anyway, I don't want to beat a dead horse - I'm just thinking that it would somehow be neat to browse or search all the field-label mappings. (You can't currently do something like that, afaik.)

(Ok, here is an alternative idea: what if the textfield for entering a metadata value in the existing Details panel became a contextual textfield? IOW, it would provide suggestions as you type, using the set of all labels mapped to that field for contextual suggestions? If you select a label from the list, then the label gets instantly assigned and the metadata value gets fille in. Does that make sense?)

Regarding the automatic assignment of parent labels: I understand that works well for you. But I am curious: have you also considered the alternative of enabling "Process parent mapping" in all your taxonomy labels? (In that case, you could assign only the leaf labels.)

Regarding the template idea: I am afraid your taxonomy would not be easily amenable to templating, unless the templates support some kind of logic based on the depth level of a label node; I had already written some notes to #2844 - we could continue this discussion in Mantis, if you want.

Here is what could be already feasible and helpful to you:
- a custom script that scans your taxonomy and lists (and, optionally, corrects) any label for which you missed to define a mapping
- a custom script that scans a collection of images and filters any image which misses one or more required fields (dwc:phylum, dwc:class, etc.); obviously, you would then need to manually correct any such image (by assigning the necessary label(s))

Cheers,
Vlad

[fbungarz]

Hi Vlad,

Regarding the template idea: I am afraid your taxonomy would not be easily amenable to templating, unless the templates support some kind of logic based on the depth level of a label node

Yes, of course it would be complex. But introducing simplistic templates that do not permit this kind of flexibility would not make a lot of sense anyway. For example, why would anyone want to have all labels of one category being mapped to the same field? This would allow you to have essentially only flat categories, because nested labels could not be mapped! If templates cannot be configured to permit label mapping according to their level nested inside the hierarchy, this would essentially defy the purpose of having label templates for categories in general. If you cannot configure mapping from the templates, but only other properties (parent label mapping etc.) the power of having these templates would be much diminished.

- we could continue this discussion in Mantis, if you want.

Yes, perhaps that is where this all belongs in the first place. All I actually wanted to know when I wrote this post was a simple answer to my question if I can use two instances of the same database side-by-side.

Regarding a script...
Believe me, if I knew how to write such a script, I would have already done so
Unfortunately I am not a programming wizard and scripting is quite beyond me, that's why I liked your label template idea in the first place. I know PSU gives anyone with programming skills an enormous flexibility to achieve what they want. But if you are illiterate if it comes to scripting, then you are out of luck (or need to keep nagging people from this user forum until someone decides to help you...).

But I am curious: have you also considered the alternative of enabling "Process parent mapping" in all your taxonomy labels?

Sorry for the confusion. That is the option I was talking about. It is called "Write the XMP mappings for all parents as well" in IDI and I was not aware that the wording for that option had been changed in PSU.

I am not quite sure I like your idea to have the mapped labels show up beside the metadata fields. I think that is quite redundant. If a field has content, then I either filled it manually or via a label. Not sure, why it would help to have it display a label at the side? If it was necessary to know if a metadata field has been filled by a mapped label a small, little icon would quite suffice...

I'm just thinking that it would somehow be neat to browse or search all the field-label mappings. (You can't currently do something like that, afaik.)

Yes, you cannot do that. But how would the labels being displayed alongside the metadata fields help with that? If you are interested to see all mapped labels, a different approach in my opinion would make a lot more sense: have mapped labels in the catalog tree being marked by some kind of icon. That way one could very easily see if a particular label has been mapped or not. Also, it would not unnecessarily eat up real estate in the label assignment panel by displaying content and label side by side [I am sorry, but I really do not like that idea at all, I think it simply takes up space and shows no additional info].

Further thinking about your suggestion, I was at first intrigued by the idea of having only the label search box show up inside the image details panel. However, this would not work either: One can use that search box not only to quickly assign, but also to remove a label. In my case having the box, but not the labels visible could thus even be dangerous. Say, I want to have the scientific name field filled and assign a species name. The field is empty, so it seems a label was not assigned. I assign the label via the box, but the content still does not show up. Now, what that would actually mean is two things could have happened: (1) the label is not mapped and was not assigned and I just assigned it (so I then would have to go to the label panel and map it correctly) OR (2) the label was assigned, but not mapped and by trying to assign it I actually removed that label (so now I need to assign it again and fix the mapping!).

Anyway, to summarize my opinion:
(1) For simplicity's sake having the option to display the two panels side-by-side would in my opinion still make the most sense...
(2) I don't see much sense in implementing label templates if complex label mapping of nested hierarchies are not part of the deal.

Cheers,
Frank

[vlad]

Hi Frank,

All I actually wanted to know when I wrote this post was a simple answer to my question if I can use two instances of the same database side-by-side.

Have you tried it?

Unfortunately I am not a programming wizard and scripting is quite beyond me, that's why I liked your label template idea in the first place.

Fair enough. But I'm wondering if a truly flexible template facility does not itself require (almost by definition) some kind of scripting (in order to define or refine a template) - which would again rule it out for users like you. (That's not a judgement of you personally, but I'm speculating that you may fit in a narrow customer niche: very advanced requirements wrt. DAM and technical standards, yet non-technical when it comes to programming. That's a bit paradoxical, as scripting could offer the highest benefits to advanced DAM users like you; most users may not need any script at all.)

Perhaps a lot of flexibility could be achieved simply by allowing some kind of templating based on macros (for example, for the nesting level of a label). Or, perhaps, a clever interface could offer options on nesting levels, etc. (IIRC, Hert managed to make the file renaming facility much more user friendly in PSU v3 - impressive!)

Another thought is that PSU could expose via the scripting API a programmable hook that is invoked whenever a new label is created (in some place); then, someone with scripting knowledge could write a custom label setup hook that configures the properties of a label (including any field mappings) based on its place in the label hierarchy. (Of course, the properties would still be visible and editable via the existing panel - so no interface change would be needed.) Frank, I know that wouldn't directly help you, but at least it would allow someone like me to help you

I know PSU gives anyone with programming skills an enormous flexibility to achieve what they want.

Unfortunately, the status quo is a bit more constraining: it gives anyone with programming skills and lots of patience to figure out the PSU API details (which is far from easy even for a programmer, trust me). I do agree about the enormous flexibility part - and I am also aware that properly documenting the scripting API would require significant effort.

But if you are illiterate if it comes to scripting, then you are out of luck (or need to keep nagging people from this user forum until someone decides to help you...).

Yes, that's true. I would like to help you (again) with scripts, but I probably can't do that for the next couple of weeks - sorry.

But I am curious: have you also considered the alternative of enabling "Process parent mapping" in all your taxonomy labels?

Sorry for the confusion. That is the option I was talking about. It is called "Write the XMP mappings for all parents as well" in IDI and I was not aware that the wording for that option had been changed in PSU.

Yes, but you also mentioned the automatic assignment of parent labels. Do you turn on both these settings? If yes, then may I ask why?

If you are interested to see all mapped labels, a different approach in my opinion would make a lot more sense: have mapped labels in the catalog tree being marked by some kind of icon.

That would not allow you to see which labels are mapped to which fields.

Further thinking about your suggestion, I was at first intrigued by the idea of having only the label search box show up inside the image details panel. [...]

Say, I want to have the scientific name field filled and assign a species name. The field is empty, so it seems a label was not assigned. I assign the label via the box, but the content still does not show up.

This would simply not happen - by design. (Note that the search box would offer as suggestions (or, at least, highlight) only those labels that have already been mapped to the field - sorry if that wasn't clear. If we were able to also create and/or map other labels on demand, then that could be a plus - but in all cases the selected label would be eventually mapped to the field and assigned to the image.)

Now, what that would actually mean is two things could have happened: (1) the label is not mapped and was not assigned and I just assigned it (so I then would have to go to the label panel and map it correctly)

Nope - my suggestion is that the search would find a label only if it was already mapped. Or, maybe you could pick any label and it gets automatically mapped (yes, it may be risky: what happens if the label already had some other mapping?). Anyway, the fine details could be discussed (preferably, in Mantis) if we agree that it's worth submitting a feature request - otherwise, there's not much point.

OR (2) the label was assigned, but not mapped and by trying to assign it I actually removed that label (so now I need to assign it again and fix the mapping!).

Nope - any interface should make all actions explicit. Good design implies predictability and little room for ambiguity. Let's not add unnecesarily complex hypotheticals in discussing feature proposals. (And let's not forget that we are simply discussing here rough ideas - Hert will ultimately bring them to fruition, if and how he sees fit.)

Anyway, to summarize my opinion:
(1) For simplicity's sake having the option to display the two panels side-by-side would in my opinion still make the most sense...
(2) I don't see much sense in implementing label templates if complex label mapping of nested hierarchies are not part of the deal.

Fair enough, that's a very good summary.

Cheers,
Vlad