| 15 Apr 2024 |
 latot | still, be able to have consistency on the results only in UTM is enough for most cases, I think this is so useful that would be great to know which spatial operations has this. | 21:27:22 |
 Regina Obe | although thinking about it more, hmm that might not help much | 21:27:32 |
| latot | so we don't need to rely on 4326 that much | 21:27:34 |
| latot | In reply to @robe:osgeo.org
although thinking about it more, hmm that might not help much why? | 21:27:44 |
| Regina Obe | When you go from degrees to unit preserving all bets are off for longish areas | 21:28:19 |
| Regina Obe | pramsey: Thought on what latot is asking? | 21:28:52 |
| Regina Obe | @lalot using ST_Segmentize in geography is very useful. What I do is ST_Segmentize(geog, 10000)::geometry - -then reproject to other units as needed | 21:29:48 |
| latot | In reply to @robe:osgeo.org
When you go from degrees to unit preserving all bets are off for longish areas sorry, can't get very well the idea of that, not enough translation skills! | 21:30:11 |
| Regina Obe | cause if you do straight up ST_Transform(geog::geometry,whatever) for longish areas, it will be off, because you need to consider more curvature of earth | 21:30:36 |
| Regina Obe | It's hard even if you are an english speaker, I find it hard to think of too, so I'm probably not expressing the concept well | 21:31:14 |
| Regina Obe | Take for example the picture in the docs here -- https://postgis.net/docs/en/ST_Segmentize.html | 21:31:34 |
| Regina Obe | if you start off with a really long line, and you try to transform to another spatial reference, you still end up with a straight line | 21:32:00 |
| Regina Obe | The problem with trying to segmentize in planar space though, is those points you get along the 2 point line for large linestring is not on the linestring, so you would still get wrong answers trying to do ST_Transform( ST_Segmentize(ST_GeomFromText(''',4326), 0.0001), some_measure_preserving) | 21:34:43 |
| Regina Obe | but anyway when you are at that point, the measure preserving would probably be still wrong, since it wouldn't properly support your long linestring anyway. Not sure there is a formula you can use to determine when it's safe and not safe | 21:35:34 |
| Regina Obe | pramsey: might have some ideas on that | 21:35:43 |
| Regina Obe | Hey I just realized we are talking in general channel -- we should be talking in #postgis channel | 21:36:13 |
| latot | I thought this channel was for this topic, to keep postgis and gdal ones, only for their respective topics | 21:36:49 |
| Regina Obe | oh nevermind, maybe my head is too much in PostGIS I assumed you were talking about PostGIS specifically | 21:37:22 |
| latot | don't worry, ppl in postgis channel is a lot more flexible with it, is usual to use that channel for general talk too | 21:38:16 |
| latot | coherence between CRS in this way seems pretty hard... even if there is a safe way to do it, is like can only be handled for two CRS, while more CRSs exists, the max distance for st_segmentize would decrease until is not useful...
Still would be great be able to handle on the transformations, a safe way to transform linestrings | 21:40:23 |
| latot | time to go here, time to gym! | 22:12:20 |
| latot | see later all, thx :D | 22:12:34 |
| 16 Apr 2024 |
|  Adrián Isidro Yoris changed their display name from Adrian to Adrián Isidro Yoris. | 07:45:07 |
 Javier Jimenez Shaw | Here are my two cents: I think there are two different problems regarding the reprojection of a polyline.
One is the segments connecting the points. They are usually a "straight line" in the selected CRS. That straight line does not have to be a great circle. They are different depending on the projection. The typical example is that the shortest distance (a great circle) is a non straight line in a map. There are tons of examples in the Internet (the line between New York and Lisbon). For the Mercator projection the line between two points is a loxodromic curve, that has a constant bearing.
How does any algorithm that add points to a polyline is depending on the algorithm, that may change. Using the great circle (or geodesic) does not have to be always the needed way. Linear interpolation is a common practice, but you should know it is not the shortest distance.
That can produce the issue of two lines that intersect in one projection but not in another one. Well, the question is how are you connecting the dots, specially if the points are far apart.
Another different problem is the distortion of any projection. That means that measuring a distance or an area on a projected map (like using a ruler on a printed map and applying the scale) has always some distortion, and does not match that measurement in the "Real World (TM)". If the area of use is small, and the proper projection is selected, the distortion can be small. If your distances are big, then you have to deal with it. That means for instance that the area of a country does not have a "unique answer"... depends on the CRS used (or the ellipsoid used if you measure in geographic coordinates doing some complicated trigonometry).
That said, mathematically speaking, if you create "infinite" points in a polyline (the segments are differential), and ignoring singularities and antimeridians in the projections, two lines intersect in the same points, regardless of the projection. But of course, this is undoable in a computer. | 07:52:29 |
|  @irc_libera.chat_hightower3:osgeo.org left the room. | 11:16:38 |
|  hightower2 joined the room. | 13:19:54 |
|  @irc_libera.chat_andreaerdna:osgeo.org joined the room. | 18:37:15 |
| @irc_libera.chat_andreaerdna:osgeo.org left the room. | 18:38:59 |
|  phma left the room. | 20:39:21 |
| phma joined the room. | 20:40:13 |
