Wikidata:Property proposal/rail direction

rail direction edit

Originally proposed at Wikidata:Property proposal/Transportation

Not done

Description	the word used to describe the direction of travel along a railway line
Represents	rail direction (Q17152310)
Data type	Item
Allowed values	north, south, east, west, up, down, odd, even, inbound, outbound, inner, outer, clockwise, anticlockwise (some need items)
Allowed units	no unit
Example	on 59th Street–Columbus Circle (Q2982354): adjacent station (P197) → 50th Street (Q2198512), qualifier [rail direction] → south (Q667) (or new item)
Planned use	qualifier to adjacent station (P197) and terminus (P559)
See also	direction (P560), Wikidata:Property proposal/towards

Motivation

direction (P560) is not very useful as a qualifier for adjacent station (P197), since most railway/metro/subway/light rail/tram/streetcar/funicular lines have curves and so the bearing is not as helpful as the direction relative to the line. This would also be useful in eventually converting Template:S-line (Q12671420) to Wikidata+Scribunto. Jc86035 (talk) 04:34, 6 November 2017 (UTC)[reply]

Discussion

Question I'm not sure that this differs all that much from direction (P560). It seems more usual to designate direction in terms of the destination of the services. Would these values also be suitable for this property? Sam Wilson 08:03, 6 November 2017 (UTC)[reply]
@Samwilson: direction (P560) is different, and per Wikidata:Property proposal/towards should probably be split into two different properties for indicating destination and cardinal direction. As an example for differentiating it from the N/S/E/W use of direction (P560), the M (Q126418) service in New York City heads west, then turns to go north, then turns to go east. Using "north" and "south" is clearer than swapping from "west" to "north" and then "north" to "east" in the middle of the line. Similar applies for clockwise/anti-clockwise services. This is also different from the terminus use of direction (P560)/other new property in that it allows lines with multiple termini like the District line (Q211265) to have better sorting of directions. It also allows indication of the up and down directions for lines which use that method. Jc86035 (talk) 08:36, 6 November 2017 (UTC)[reply]
@Jc86035: So 'rail direction' would be used alongside 'towards', and there would be no direction (P560) at all? Or would direction (P560) also be used? It seems to me that direction (P560) is for the actual direction of the physical track as it departs the station, and this new property would be the 'logical direction' of the whole line; is that right? If so, perhaps this should be called 'line direction' instead of 'rail direction'. Sam Wilson 08:45, 6 November 2017 (UTC)[reply]
@Samwilson: I'm not sure; the Wikipedia article is called "rail directions" and refers to "railroad directions". It seems weird to call "even" a "line direction" (direction the line goes in) instead of a "rail direction" (word used to describe that direction?), and I'm not sure if the property should cover both concepts. A property not limited to railways would be more useful. Jc86035 (talk) 10:22, 6 November 2017 (UTC)[reply]
Hm, yeah ok, makes sense. So this would use e.g. up (Q15332375) — i.e. all the relative direction (Q14565197), cardinal direction (Q23718), etc. (anything that's an instance of a subclass of direction (Q2151613)?) Would there need to be any new instance of (P31) of rail direction (Q17152310)? Sam Wilson 11:06, 6 November 2017 (UTC)[reply]
@Samwilson: Probably "up", "down", "odd", "even", "inbound", "outbound", "inner" and "outer"; although perhaps maybe they could all be new items to reflect that the line does not necessarily literally e.g. go "up". Should there need to be a distinction between systems where the operator uses the directions, and systems where the line just happens to go from e.g. west to east and the directions are used unofficially? Jc86035 (talk) 12:11, 9 November 2017 (UTC)[reply]
@Jc86035: Yes, good point, there is definition a difference between up (Q15332375) and the rail direction 'up'; we should create the set of items for the rail direction values (as @JakobVoss says below), that will help clarify this I think. Sam Wilson 23:39, 11 November 2017 (UTC)[reply]
More examples would also help, the more the better. As I understand the Wikipedia article, at least one example for each country described in the article should be given. -- JakobVoss (talk) 00:13, 12 November 2017 (UTC)[reply]
Support Sam Wilson 11:06, 6 November 2017 (UTC)[reply]
Support David (talk) 07:45, 7 November 2017 (UTC)[reply]
Comment I'm not convinced but cannot judge. In any case please link existing item ids for allowed values instead of simple text strings or give a superclass for all allowed values. -- JakobVoss (talk) 19:38, 11 November 2017 (UTC)[reply]
When it comes to the example I don't see why direction (P560) wouldn't do the job. Can you explain why you think it doesn't in this case or provide examples where it wouldn't do the job? ChristianKl (✉) 00:20, 12 November 2017 (UTC)[reply]
@ChristianKl: With direction (P560) as a qualifier to adjacent station (P197), it's not clear whether the property refers to the direction relative to the line or to the station (I've run into both in the past). It's also impossible to indicate that the adjacent station is in the "up" or "even" direction, currently, and it wouldn't work as a relative direction because railway directions are usually not described in terms of elevation change (and that can be trivially inferred from elevation above sea level (P2044)). The same property could be used for both regular and "rail direction" values, but it would be difficult for editors to tell due to the use of identical labels and this would probably be a misuse of the property. Jc86035 (talk) 01:21, 13 November 2017 (UTC)[reply]
Can you please provide multiple and complete examples? In particular an example where direction (P560) and "rail direction" are used as qualifiers at the same statement so they would need to be differentiated. -- JakobVoss (talk) 07:30, 13 November 2017 (UTC)[reply]
@JakobVoss: Sorry, I didn't see your reply. Here is a table showing how direction (P560) could be used ambiguously in at least three different ways (ignoring circular lines, lines with branches, and values which are line termini). The w:en:South Island Line of the Hong Kong MTR has five stations, and the table shows the various bearings between each of the stations. Almost all the triplets have discrepancies exceeding 45°, with the largest being the 122° difference of South Horizons → Lei Tung (direction of the tracks at end of the station vs. general line direction). The proposed property would avoid these issues by using generalized directions, in this case "up [the line]" and "down [the line]". "Up" is south-southwestward from Admiralty station to South Horizons station, and "down" is the reverse. I calculated the bearings with the OSM editor JOSM, and rounded the values to the nearest degree. Jc86035 (talk) 15:19, 9 February 2018 (UTC)[reply]

Section	Between stations as the crow flies, up	Track direction at origin station, up	General direction along the line, up	Between stations as the crow flies, down	Track direction at origin station, down	General direction along the line, down
Whole line	203° (SSW)	222° (SW)	203° (SSW)	023° (NNE)	145° (SE)	023° (NNE)
Admiralty–Ocean Park	165° (SSE)	222° (SW)	203° (SSW)	345° (NNW)	060° (ENE)	023° (NNE)
Ocean Park–Wong Chuk Hang	264° (W)	240° (WSW)	203° (SSW)	080° (E)	078° (ENE)	023° (NNE)
Wong Chuk Hang–Lei Tung	241° (WSW)	258° (WSW)	203° (SSW)	061° (ENE)	046° (NE)	023° (NNE)
Lei Tung–South Horizons	274° (W)	226° (SW)	203° (SSW)	094° (E)	145° (SE)	023° (NNE)

Support I support this, if only to further qualify direction (P560). However I suggest additionally to define this as a subclass of direction (P560). --Wanderer28 (talk) 17:00, 18 January 2018 (UTC)[reply]
Wait Not ready, given that we lack complete examples. ChristianKl ❪✉❫ 13:06, 20 January 2018 (UTC)[reply]
@ChristianKl: See the table above. Jc86035 (talk) 15:19, 9 February 2018 (UTC)[reply]
@Jc86035: The table above doesn't say anything about how the Wikidata items will look like in the end. How about creating example items of over at test.wikidata.org?
@ChristianKl: I've made a rudimentary test item at https://test.wikidata.org/wiki/Q135398, which should closely approximate the adjacent station (P197) values of the real item Ocean Park station (Q7076043) and would be usable in a Wikidata version of w:en:Template:S-line if the rail direction property were correctly added as a qualifier to the line's terminus (P559) statements. I haven't added direction (P560) but the values respectively would be north-northwest (to Admiralty) and west (to Wong Chuk Hang), assuming that the values in columns 1 and 4 above are correct and the values in columns 2 and 5 are wrong.

It's also occurred to me that the property could be retitled so it could be used for numbered roads. Jc86035 (talk) 14:59, 12 February 2018 (UTC)[reply]
@Jc86035: I don't see why direction (P560) can't do the job in that example. ChristianKl ❪✉❫ 16:33, 14 February 2018 (UTC)[reply]
@ChristianKl: Isn't that sort of like saying we could just use start time (P580) instead of its eight subproperties? As mentioned, if only direction (P560) were to be used for all values, systems using "north" and "south" as directional identifiers might have both situations where "north", the direction along the line, would be confusingly displayed next to "northeast" or "west", the bearing (or, as it currently stands, a crude approximation of such) of a straight line to the next station on a map. Examples would be Flushing Avenue (Q2337501), where "northbound" is actually about northwest; and then (on the same service) 46th Street (Q2096739), where "northbound" is actually about east-southeast. Having multiple subproperties would unquestionably help differentiate the two uses.

And Gangnam Station (Q17469) currently uses Down (Q22342279), Up (Q22342273), Outer ring road (Q13388740) and Inner Ring Road (Q13388955) as direction (P560) qualifier values. Could those (particularly the latter two) be classified as "directions"? Calling them directional identifiers through the use of a different property or subproperty makes the classification much less problematic since they don't represent the actual directions "up" and "inner"(?). Jc86035 (talk) 10:25, 16 February 2018 (UTC)[reply]
Oppose I don't think this proposal has been correctly thought out. It is stated that direction (P560) is "not very useful". I don't agree with that at all. direction (P560) is supposed to be used to reflect the system which is actually used by railways, and understood by passengers, as is seen in this template [1] i.e. direction (P560) (towards) = Watford Junction railway station (Q10698). Further more, we already have the property direction relative to location (P654) which seems to be similar, if not the same as is being proposed here. Danrok (talk) 15:49, 3 February 2018 (UTC)[reply]
@Danrok: Using the terminus doesn't work for circular lines, for obvious reasons. The reason I said it wasn't very useful was that it's not specific enough, and many uses with a cardinal direction are for the direction relative to the station instead of to the line. Furthermore, the aforementioned uses for the direction relative to the station don't specify whether the direction is the bearing of the tracks as they leave the station or the bearing from the station to the adjacent station, which makes it even less useful (and on top of that, the property is usually only used with the four basic cardinal directions and doesn't accept bearings in degrees or radians as far as I'm aware). direction relative to location (P654) seems to be relative to the object (station) and not the general direction along the line, so it wouldn't be possible to use it in place of the proposed property. Jc86035 (talk) 14:34, 9 February 2018 (UTC)[reply]

As for adding a terminus, it would probably be better to use a different property for that as well (see the section above, Wikidata:Property proposal/towards). If the template ends up using Wikidata it would be easier to find values of separate properties in qualifiers instead of finding whether each value of direction (P560) is a station. Jc86035 (talk) 15:49, 9 February 2018 (UTC)[reply]
Comment to voters @Samwilson, ديفيد عادل وهبة خليل 2, Wanderer28, ChristianKl, JakobVoss, Danrok: If the property is created, would you agree with titling the property "directional identifier" or similar, so that it could be used for other systems where appropriate, such as for road termini e.g. on the item Interstate 95 (Q94967) (also usable as e.g. qualifiers direction (P560) → clockwise direction (Q16726164); directional identifier → A (Q9659) for Route 9 (Q7371553))? If "inner loop" etc. are used as values, would it be appropriate to make the property a subproperty of direction (P560)? Jc86035 (talk) 10:45, 16 February 2018 (UTC)[reply]
I'm not exactly sure what "inner loop" is supposed to mean. How about creating corresponding items for the direction you talk about with descriptions that tell us what they are? ChristianKl ❪✉❫ 13:26, 16 February 2018 (UTC)[reply]
@ChristianKl: "Inner loop" and "outer loop" mean anticlockwise or clockwise depending on the driving direction of a system (so in a system which drives on the right, the inner loop is clockwise). Jc86035 (talk) 17:51, 16 February 2018 (UTC)[reply]
@Jc86035: Is there a good reason to not call it clockwise/anti-clockwise and use different terms? ChristianKl ❪✉❫ 10:02, 17 February 2018 (UTC)[reply]
@ChristianKl: Strictly speaking, they wouldn't be the actual "directions" that the system uses. We could just use northbound, southbound, westbound, eastbound, anticlockwise and clockwise instead of up, down, inbound, outbound, A, B, etc., but they wouldn't be correct because the actual classification systems used are different (sort of like if Wikidata were to directly translate into English place names that are usually transliterated into English). Jc86035 (talk) 10:17, 17 February 2018 (UTC)[reply]
@Jc86035: When every city has their own classification schema I don't see why we should call the same thing differently for every city. I would prefer it if we use one classification schema for all cities even if that's not how individual cities list their data. ChristianKl ❪✉❫ 19:25, 17 February 2018 (UTC)[reply]
@ChristianKl: Maybe a "general direction" property (for the actual northbound/southbound/eastbound/westbound/clockwise/anticlockwise) and a "labelled route direction" property (for the northbound/southbound/eastbound/westbound/up/down/odd/even/A/B/inbound/outbound/clockwise/anticlockwise/inner/outer)? What the system says is correct (e.g. "railroad north" in New England) and would be valid data would be omitted otherwise. This would probably require a new property proposal, though. Jc86035 (talk) 08:10, 18 February 2018 (UTC)[reply]
@Jc86035: We do we need the label? Keep in mind that we do have object named as (P1932). ChristianKl ❪✉❫ 19:35, 18 February 2018 (UTC)[reply]

I find "directional identifier" a little repetitive when it's implemented. Imagine having "direction" as a qualifier, then "directional identifier" as the following qualifier.

Logically I think direction (cardinal), direction (relative)... etc. should be separated as different properties. If direction (P560) is meant to be used as a cardinal direction, then its name should also be changed to reflect its specificity. But before that, we should rake through all instances of railway stations and replace direction (P560) with this other property we're making. Personally, I find attaching brackets to "direction", not unlike those in disambiguation pages, the most attractive; you'd have direction (cardinal), direction (relative), and direction (route)...? That's the name I would suggest in place of "directional identifier" or "rail direction". --Wanderer28 (talk) 08:17, 17 February 2018 (UTC)[reply]
@Wanderer28: I'm fine with this, as long as the property descriptions are specific about what they mean; though for consistency with existing property names in English, I would call them "cardinal direction" etc. instead of suffixing a parenthetical. The property should definitely be replaced on railway stations; but since Up (Q22342273) and Down (Q22342279) are only used on about 21 Seoul stations and Shinbundang Line (Q20388), usage would probably have to be manually checked (and preferably properly sourced on each line item). Jc86035 (talk) 10:17, 17 February 2018 (UTC)[reply]
@Jc86035, Samwilson, JakobVoss, ديفيد عادل وهبة خليل 2, ChristianKl: @Danrok, Wanderer28: I've marked this as Not done given the staleness and apparent inconclusiveness of the above discussion. If some of you feel there is still a need for a property like this, please propose a new version linking to this discussion, and taking as much of it into account as you can in crafting the new proposal. ArthurPSmith (talk) 18:02, 22 May 2018 (UTC)[reply]