CRS updates

[Robinlovelace]

See #713

[Robinlovelace]

Heads-up @Nowosad the intention with this PR is to complete Chapter and therefore (I think) most of the stuff needed to submit Part 1 for review. Sorry it's taken so long. Key points: we should not mention 'WKT2' as it's confusing, WKT2 is and in the future will be synonymous with WKT representation of CRSs. I have read and linked to key related docs on CRSs. One thing that surprised me: it does not seem that the AUTHORITY:CODE string convention in EPSG:4326 for example is universally accepted. The only standard is WKT, there are still different ways of identifying WKT strings as outlined below and now in this PR.

https://docs.opengeospatial.org/is/18-010r7/18-010r7.html#37 now given the proper citation it deserves, the most important doc here

The QGIS docs which show 5 CRS identifier types: https://docs.qgis.org/3.16/en/docs/pyqgis_developer_cookbook/crs.html?highlight=srid

PostGIS docs which is where I think the term 'SRID' comes from. I think 'identifier' is better for our purposes: https://postgis.net/workshops/postgis-intro/projection.html

As others have said this is a minefield. But I'm confident the updated text is accurate and better. Needed to rework these earlier sections before getting to 7.4. Any comments v. welcome.

[Nowosad]

Hi @Robinlovelace,
thanks for opening the discussion:

1. WKT2 -- as I mentioned in the text -- this is just a dialect of WKT. If you just want to use WKT consistently in the text (instead of WKT2) that is fine by me. However, we probably should mention this dialect once (e.g., in chapter 2) and say that we will use just WKT from now on...
2. "It does not seem that the AUTHORITY:CODE string convention in EPSG:4326 for example is universally accepted." -- I think PROJ and GDAL universally accept it -- and that is good for us. You just need to remember that only WKT is used for storing CRSs, and the AUTHORITY:CODE convention is mostly for setting, changing, and transforming CRSs.
3. The QGIS code, I think, is just like sf or terra -- these examples just use PROJ API plus some specific QGIS code. See our Section 7.3:

4. Correct me if I am wrong, by SRID means Spatial Reference Identifier. I am not very attached to this name, so we could use just "identifier" instead. The question, however, is if that is enough not to confuse the readers (would they understand that by identifier, we mean CRS identifies/SRID?)
5. Let me know when the text is ready for my review.

J

[Robinlovelace]

Hi @Nowosad all good questions and it's a minefield so worth discussing. Responses soon...

[Robinlovelace]

Regarding your point 1:

WKT2 -- as I mentioned in the text -- this is just a dialect of WKT. If you just want to use WKT consistently in the text (instead of WKT2) that is fine by me. However, we probably should mention this dialect once (e.g., in chapter 2) and say that we will use just WKT from now on...

The fact that the WKT representation has changed over time is not relevant for most users and will be come less relevant, tending towards obsolete, in the coming years. Good idea to mention it. I think a footnote in C7 would be the best place for that, e.g. here:

https://github.com/Robinlovelace/geocompr/blob/fa05ae039b2e7f45dc9ef0059d5b0adb95a52b7d/07-reproj.Rmd#L100-L107

That's my default plan, it fits well there because that footnote also mentions the increasingly obsolete proj4 string representations. Note to self: need to find and change all references to WKT2 for consistency (hence putting in as a PR to ensure it's not breaking the logic of other parts of the book).

[Robinlovelace]

I think PROJ and GDAL universally accept it -- and that is good for us. You just need to remember that only WKT is used for storing CRSs, and the AUTHORITY:CODE convention is mostly for setting, changing, and transforming CRSs.

Good idea to focus on what PROJ and GDAL accept, I was looking (perhaps too much) at the PostGIS docs. Yes of course the identifier is only a look-up and that the CRS is stored in the WKT string. That is now more clear I think. If we had links showing that PROJ/GDAL accept the AUTHORITY:CODE representation (which I called AUTHORITY:code, plan to change as all upper looks clearer) that would help, do you have any links to docs in PROJ where identifiers are used?

[Robinlovelace]

The QGIS code, I think, is just like sf or terra -- these examples just use PROJ API plus some specific QGIS code. See our Section 7.3:

Good point on that, I think clarifying that and adding a cross reference to the subsequent section is a good idea. However, I think the QGIS example shows that there are alternative CRS identifiers other than the 'SRID'.

[Robinlovelace]

Correct me if I am wrong, by SRID means Spatial Reference Identifier. I am not very attached to this name, so we could use just "identifier" instead. The question, however, is if that is enough not to confuse the readers (would they understand that by identifier, we mean CRS identifies/SRID?)

Yes according to PostGIS documentation. But where else is that term used? I'm not sure that in PostGIS docs they would say that EPSG:4326 is an SRID. Do you have more info on that (I tried to find other places where SRID was used but could not find much)?

I think if we define CRS identifiers early and clearly, with reference to the OGC standard which calls them identifiers, it will be clear to readers. Pros and cons. ATM I'm thinking the advantages of just 'identifier' used by OGC outweigh the advantages of SRID, used in PostGIS docs (and possibly elsewhere?).

[Robinlovelace]

Another observation: there is really good documentation here which mentions identifiers but avoids the acronym SRID: https://proj.org/operations/operations_computation.html However there is mention of SRIDs here: https://proj.org/faq.html

[Robinlovelace]

But that seems to be the only mention of the term: https://www.google.com/search?q=srid+%22spatial+reference%22+site%3Aproj.org

[Robinlovelace]

In GDAL docs there seems to be no mention or srids: https://www.google.com/search?q=srid+%22spatial+reference%22+site%3Agdal