On 2013-04-03 21:43, WanMil wrote:

...

0> In article<515C8E7F.4000004@web.de>, 0> WanMil<URL:mailto:wmgcnfg@web.de> ("Wanmil") wrote:

Wanmil> mkgmap contains some code to support "through_route" relations. [...] Wanmil> Shall we still support this type of relation? Wanmil> Can anybody explain what this relation does? Sorry - I've not kept my attention on this mailing list much recently, and I missed this one. I believe I may have been the user who asked Steve Ratcliffe for such a feature a few years back, and he agreed, and patched it in over a weekend or so.

Sorry it didn't get documented. I think it was initially nothing more than an experiment, but it works brilliantly for me (since I know what it does!) and evidently a few others must have picked up on it too. Basically, it's markup that exists to convince the Garmin units to issue better verbal instructions when a road hits a junction but that the logical "main" route through that junction takes you to a different road. I.e. the road-number (even the road class) changes magically in the junction, but on the ground you think you're still "on the main road" so to speak. Look at http://www.openstreetmap.org/?lat=51.44852&lon=-3.49448&zoom=16&layers=M (I'm going to have to fix this example: someone's broken it! The road heading NE towards Cowbridge is the B4270, not the B4268). But anyway - if you approach that junction from the south (on the B4270) intending to drive through Llysworney (further to the north) but without "through route" being enabled, your Garmin will tell you to "turn left" at that junction. However, if you look at it on the road, you actually want to "go straight on". There *is* a turn left, but that's down a small uncategorised road which you certainly don't want to be on! Conversely, on the same junction, had you approached from the south (on the B4270) intending to drive to Cowbridge then it would say nothing as you hit the junction because after all you want to leave the junction on the B4270 - the same road that you approached on. But when you get there, to stay on the B4270 to Cowbridge is clearly a "turn right", and should be announced as such. With the "through route" system it all works fine. Without it: confusing directions for the motorist. I request that the feature be kept - possibly enhanced. I have noticed that the existing system can be accidentally broken when others alter the roads around a junction and leave "through route" set on too many ways in the vicinity. I think Steve R's implementation only works if there are exactly 3 nodes in the relation: two roads and the junction node itself. I did catch a few cases around my home town where some bus routes had been edited in, and when approach-roads to junctions marked with "through route" got split to allow for these bus routes doing funny things, then the two split fragments all ended up in the "through route" relation (taking the basic item-count above 3) and it failed. The system needs to count only the junction node and the two ways that must be *directly* connected to it.

...

I think the best explanation is in the patch announcement email[1] - perhaps that could be adapted for the OSM Wiki?

[1]<URL: http://www.mail-archive.com/mkgmap-dev@lists.mkgmap.org.uk/msg04975.html>

Thanks for the link!

Now there have been 3 years after Mark proposed to document it in the Wiki... I will wait for another two months. If the relation is not documented after that time I think we can remove the code because the usage is very rare. I guess only a very few people know how to use this relation type correctly. From my point of view we shouldn't support inofficial undocumented stuff (although it sounds useful...)

Please don't feel offended. Just add the required documentation ;-)

WanMil

_______________________________________________

I could do the documentation (based on the text I just wrote above). But it would be nice to get that "split approach road" bug out of the system as currently implemented (sorry, Steve - more work for you I guess). "Through route" should be used by more mappers! Definitely very useful on British roads - and I'd have thought useful on everyone else's too. Steve Hosgood

Steve Like you I think this is a useful option which should stay so I guess it's down to us to drum up some action on the wiki although I'll confess that I don't know where the best place to start is as I haven't spent much time on any wiki let alone one as large and well established as OSM. However if you feel you have the knowledge to make a start point me at the relevant page and I'll gladly pitch in Thanks Paul On 10/04/13 12:14, Steve Hosgood wrote:

On 2013-04-03 21:43, WanMil wrote:

...

...

0> In article<515C8E7F.4000004@web.de>, 0> WanMil<URL:mailto:wmgcnfg@web.de> ("Wanmil") wrote:

Wanmil> mkgmap contains some code to support "through_route" relations. [...] Wanmil> Shall we still support this type of relation? Wanmil> Can anybody explain what this relation does? Sorry - I've not kept my attention on this mailing list much recently, and I missed this one. I believe I may have been the user who asked Steve Ratcliffe for such a feature a few years back, and he agreed, and patched it in over a weekend or so.

Sorry it didn't get documented. I think it was initially nothing more than an experiment, but it works brilliantly for me (since I know what it does!) and evidently a few others must have picked up on it too.

Basically, it's markup that exists to convince the Garmin units to issue better verbal instructions when a road hits a junction but that the logical "main" route through that junction takes you to a different road.

I.e. the road-number (even the road class) changes magically in the junction, but on the ground you think you're still "on the main road" so to speak.

Look at http://www.openstreetmap.org/?lat=51.44852&lon=-3.49448&zoom=16&layers=M

(I'm going to have to fix this example: someone's broken it! The road heading NE towards Cowbridge is the B4270, not the B4268). But anyway - if you approach that junction from the south (on the B4270) intending to drive through Llysworney (further to the north) but without "through route" being enabled, your Garmin will tell you to "turn left" at that junction. However, if you look at it on the road, you actually want to "go straight on". There *is* a turn left, but that's down a small uncategorised road which you certainly don't want to be on!

Conversely, on the same junction, had you approached from the south (on the B4270) intending to drive to Cowbridge then it would say nothing as you hit the junction because after all you want to leave the junction on the B4270 - the same road that you approached on. But when you get there, to stay on the B4270 to Cowbridge is clearly a "turn right", and should be announced as such.

With the "through route" system it all works fine. Without it: confusing directions for the motorist. I request that the feature be kept - possibly enhanced. I have noticed that the existing system can be accidentally broken when others alter the roads around a junction and leave "through route" set on too many ways in the vicinity. I think Steve R's implementation only works if there are exactly 3 nodes in the relation: two roads and the junction node itself.

I did catch a few cases around my home town where some bus routes had been edited in, and when approach-roads to junctions marked with "through route" got split to allow for these bus routes doing funny things, then the two split fragments all ended up in the "through route" relation (taking the basic item-count above 3) and it failed. The system needs to count only the junction node and the two ways that must be *directly* connected to it.

...

...

I think the best explanation is in the patch announcement email[1] - perhaps that could be adapted for the OSM Wiki?

[1]<URL: http://www.mail-archive.com/mkgmap-dev@lists.mkgmap.org.uk/msg04975.html>

Thanks for the link!

Now there have been 3 years after Mark proposed to document it in the Wiki... I will wait for another two months. If the relation is not documented after that time I think we can remove the code because the usage is very rare. I guess only a very few people know how to use this relation type correctly. From my point of view we shouldn't support inofficial undocumented stuff (although it sounds useful...)

Please don't feel offended. Just add the required documentation ;-)

WanMil

_______________________________________________

I could do the documentation (based on the text I just wrote above). But it would be nice to get that "split approach road" bug out of the system as currently implemented (sorry, Steve - more work for you I guess).

"Through route" should be used by more mappers! Definitely very useful on British roads - and I'd have thought useful on everyone else's too.

Steve Hosgood

_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://lists.mkgmap.org.uk/mailman/listinfo/mkgmap-dev

Thanks for this. I'd already realised a more generic approach was better than an mkgmap specific one. Thanks Paul On 12/04/13 08:25, Henning Scholland wrote:

Am 12.04.2013 00:02, schrieb News:

...

Like you I think this is a useful option which should stay so I guess it's down to us to drum up some action on the wiki although I'll confess that I don't know where the best place to start is as I haven't spent much time on any wiki let alone one as large and well established as OSM. However if you feel you have the knowledge to make a start point me at the relevant page and I'll gladly pitch in

Hi Paul, typical you should set up a proposal. If you search the wiki for proposal you should find the necessary information. After creating the proposal you should write to tagging-ML and partisipate in a discussion with other mappers. If this relation is already used, this discussion could be quite short. Just a hint: don't argue with 'mkgmap needs it', but say 'router will need it'. ;)

Henning

_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://lists.mkgmap.org.uk/mailman/listinfo/mkgmap-dev

On 13/04/13 17:01, News wrote:

http://wiki.openstreetmap.org/wiki/Proposed_features/through_route

Anyone care to comment before I proceed??

Thanks

Paul

Looks great to me! However - since there is a prototype implementation already in place, it might make sense to document how that currently works. I.e that there should exist a "relation" containing three items: two ways and a node. The node should be marked up with role=junction (or at least, this was the case, not sure if it's still the case). The two ways don't need to have roles marked up, but both need to stop or start at the node marked "junction". As far as I know, that's it. The existing bug (as I understand it) is that despite being set up as described, subsequent edits to the entry and exit ways results in the fragments of entry and exit ways all ending up in the relation. That takes the item-count to more than three and I think mkgmap then ignores it! Of course, mkgmap should count the number of ways that end or start at the junction, and proceed if there are two such ways - regardless of how many are actually in the relation.... Steve