Potlatch, dual carriageways, and a broken changeset

[Michal Migurski-3]

Hi,

I just used Potlatch to split a single-carriageway road into a dual  
carriageway. I thought I had done it correctly, and saved my work at.  
Everything made it through except the two new carriageways - why is  
that?

Here's the changeset:
        http://www.openstreetmap.org/browse/changeset/3052340

What I did: selected Market St from about 18th St or so down to the  
water (http://osm.org/go/TZNQtBDQ-- ), used Potlatch "parallel ways"  
command, and restitched all the connected roads to the new pair. I  
copied the tags from the old road to the new ones, added oneway tags,  
and deleted the old road. Then I saved the whole changeset. Everything  
went through except for the new ways, why? I did notice that Potlatch  
had them rendered in red this whole time but I wasn't sure what that  
meant. Help!

I'm going to fix the problem by manually filling in the new ways in a  
new changeset, but I'm not sure what the appropriate TIGER tags should  
be, as well as an affected cycle network relation.

-mike.

----------------------------------------------------------------
michal migurski- mike@...
                  415.558.1610




_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Richard Fairhurst]

Michal Migurski wrote:
> I did notice that Potlatch had them rendered in red this whole time  
> but I wasn't sure what that meant. Help!

Potlatch has the concept of a "locked" way - that's one that won't be  
uploaded. It renders in red and has a little padlock by the way ID.

You can unlock it by clicking the padlock (or pressing K in true  
Potlatch obscure keypress fashion).

This first appeared with the GPX-to-way function way back when - the  
idea being that if you've automatically imported a load of random  
squiggles from your GPS track, they shouldn't be expressly uploaded  
unless you tell them to. (Back then Potlatch only had live editing, so  
it was quite crucial not to upload them immediately.) The "parallel  
way" feature creates locked ways for the same reason - ways created  
automatically shouldn't be uploaded without conscious user choice.

I'm very open to ideas on how to improve the UI to fix any confusion  
here. Should Potlatch (in save mode) alert the user "You have 5 locked  
ways, with names Frog Street | Acacia Avenue | etc. - are you sure?"?  
Bright ideas welcome.

cheers
Richard


_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Michal Migurski-3]

Ah, that makes sense.

Frustrating, frustrating sense. =)

I didn't even see a padlock, just noticed that they were a funny color.

My suggestion, completely uninformed by the reality of maintaining the  
Potlatch codebase, is that you ditch the "locked ways" feature, and  
instead make sure that multi-level undo is completely bulletproof. It  
seems weird that some ways are not like other ways, esp. when they're  
created by a built-in feature of the app. All of this assumes of  
course that you're editing in save mode, not live mode. Maybe save  
mode should be further distinguished from live mode by the absence of  
these kinds of mistake-avoidance features?

A stop gap might be better visual indication on the locked way itself,  
and ditching the "lock" terminology. The choice of red makes me think  
that the way is "really there" when it sounds like the reverse is  
true. What if it was dashed or translucent, a ghost-thing that looks  
like it's not fully part of the rest of the map? The best analogy to  
this I can think of is Adobe Illustrator's guides feature. Maybe when  
you mouse over such a way it says "hi, I'm not really here".

Anyway.

-mike.

On Nov 6, 2009, at 3:28 PM, Richard Fairhurst wrote:

> Michal Migurski wrote:
>> I did notice that Potlatch had them rendered in red this whole time
>> but I wasn't sure what that meant. Help!
>
> Potlatch has the concept of a "locked" way - that's one that won't be
> uploaded. It renders in red and has a little padlock by the way ID.
>
> You can unlock it by clicking the padlock (or pressing K in true
> Potlatch obscure keypress fashion).
>
> This first appeared with the GPX-to-way function way back when - the
> idea being that if you've automatically imported a load of random
> squiggles from your GPS track, they shouldn't be expressly uploaded
> unless you tell them to. (Back then Potlatch only had live editing, so
> it was quite crucial not to upload them immediately.) The "parallel
> way" feature creates locked ways for the same reason - ways created
> automatically shouldn't be uploaded without conscious user choice.
>
> I'm very open to ideas on how to improve the UI to fix any confusion
> here. Should Potlatch (in save mode) alert the user "You have 5 locked
> ways, with names Frog Street | Acacia Avenue | etc. - are you sure?"?
> Bright ideas welcome.
>
> cheers
> Richard
>
>
> _______________________________________________
> talk mailing list
> talk@...
> http://lists.openstreetmap.org/listinfo/talk
>

----------------------------------------------------------------
michal migurski- mike@...
                  415.558.1610




_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Anthony-32]

On Fri, Nov 6, 2009 at 6:28 PM, Richard Fairhurst <richard@...> wrote:
> I'm very open to ideas on how to improve the UI to fix any confusion
> here. Should Potlatch (in save mode) alert the user "You have 5 locked
> ways, with names Frog Street | Acacia Avenue | etc. - are you sure?"?
> Bright ideas welcome.

Can you check whether or not the way has been edited from the
original, and only pop up the alert if it has?

I have to admit I made this same mistake the first time I made a dual
carriageway and it was quite frustrating.  My bad for not RTFMing, I
guess.

_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Russ Nelson]

Michal Migurski writes:
 > instead make sure that multi-level undo is completely bulletproof.

To make life more interesting, OSM editing goes on concurrently, and
yet nearly everyone who is editing is editing a chunk locally.  So OSM
is episodically being synched with chunks of data we call a
"changeset", but which also includes the concept of an "edit conflict"
meaning that two chunks have been edited at the same time.

Simultaneous editing of geodata is currently not a solved problem.
Even less solved is the concept of multi-level undo, much less
single-level undo.  Within an editing session?  Sure.  Within your
chunk of data?  Sure.  But not outside that.

PhD dissertations are waiting to be written.  Hard, hard, hard
problems here.  Thyre be dragans.  Many pitchers of beer to be awarded
at some future SOTM -- with at least one from me.

--
--my blog is at    http://blog.russnelson.com
Crynwr supports open source software
521 Pleasant Valley Rd. | [phone]    
Potsdam, NY 13676-3213  |     Sheepdog      

_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Michal Migurski-3]

On Nov 6, 2009, at 6:38 PM, Russ Nelson wrote:

> Michal Migurski writes:
>> instead make sure that multi-level undo is completely bulletproof.
>
> To make life more interesting, OSM editing goes on concurrently, and
> yet nearly everyone who is editing is editing a chunk locally.  So OSM
> is episodically being synched with chunks of data we call a
> "changeset", but which also includes the concept of an "edit conflict"
> meaning that two chunks have been edited at the same time.
>
> Simultaneous editing of geodata is currently not a solved problem.
> Even less solved is the concept of multi-level undo, much less
> single-level undo.  Within an editing session?  Sure.  Within your
> chunk of data?  Sure.  But not outside that.

I think this further underscores the difference between save mode and  
edit mode. Save mode is more like traditional desktop document  
editing, and the undo history need only be consistent with what you're  
doing in your own session. In this case, it should be possible to  
ditch the locked ways and instead make sure that the local document is  
fully self-consistent and undo-able.

For live mode, clearly an undo feature would introduce more trouble  
than it's worth, for everyone involved in editing a particular area.  
There the presence of locked or ghosted ways as in the current  
Potlatch does make sense, but I do think the terminology and visual  
presentation could be tweaked a bit. I know what you mean about this  
not being a solved problem, and given how much time I spend with SVN,  
Git, etc. I'm pretty aware of how hard it can be to merge concurrent  
edits on the same resource.

-mike.

----------------------------------------------------------------
michal migurski- mike@...
                  415.558.1610




_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk

[Richard Fairhurst]

Michal Migurski wrote:
> For live mode, clearly an undo feature would introduce more
> trouble than it's worth, for everyone involved in editing a
> particular area.  

Spot on. Potlatch's undo does function in live mode - it predates save mode, in fact - but it's not trivial and I wouldn't want to bet my life on it working in all circumstances.

I try not to introduce differences between live and save mode (other than the obvious). Vector map editing is unfamiliar enough for most people; having an extra set of subtle differences in the UI depending on which mode you're in would, I think, be too confusing.

Fortunately I figured a more obvious way to flag up that a way is locked (and how to unlock it) and coded it earlier; it'll be committed in the next couple of days when I've finished a few more changes.

cheers
Richard

(Incidentally, Potlatch 2 doesn't have a live mode and I'm not anticipating that it will. Mind you, it doesn't have undo yet either. ;) )

[Martin Koppenhoefer]

(Incidentally, Potlatch 2 doesn't have a live mode and I'm not anticipating
that it will. Mind you, it doesn't have undo yet either. ;) )

glad to hear that and hoping that it never get's one...

cheers,
Martin

_______________________________________________
talk mailing list
talk@...
http://lists.openstreetmap.org/listinfo/talk