Determine if an srid refers to a geodetic CRS

[Erik Wienhold]

On 2025-04-30 00:01 +0200, Pettycash dev wrote:
> ### 1) **Checking if `srtext` starts with "GEOGCS"?**
> While `srtext` for geodetic CRSs (e.g., WGS84) starts with `GEOGCS` in its
> WKT definition (e.g., `GEOGCS["WGS 84", ...]`), this method has caveats:
>    - **It’s not foolproof**: The `srtext` format can vary (e.g.,
> whitespace, custom definitions).
>    - **Not future-proof**: Relies on text parsing, which may break if WKT
> conventions change.
> 
> Example query (works in most cases, but not fully robust):
> ```sql
> SELECT srid, srtext
> FROM spatial_ref_sys
> WHERE srtext LIKE 'GEOGCS%'; -- Filters geodetic CRSs
> ```

According to annex C.4.1 of the standard [1] "geodetic CRS" covers both
geographic and geocentric CRS.  So GEOGCS and GEOCCS are relevant.  Also
note annex B.8 since there's also GEODCRS, GEODETICCRS, GEOGCRS,
GEOGRAPHICCRS in new versions.  Because those are the only ones with
prefix "GEO", you might get away with just using that predicate:

    srtext LIKE 'GEO%'

[1] https://docs.ogc.org/is/18-010r11/18-010r11.pdf

> ### 2) **Use the built-in PostGIS function `ST_IsGeographic()`** ✅
> PostGIS provides **`ST_IsGeographic(srid)`** to directly check if an SRID
> is geodetic. This is the **recommended method** because:
>    - It uses PostGIS’s internal logic (no text parsing).
>    - Handles edge cases and custom definitions.

PostGIS does not provide ST_IsGeographic to my knowledge.  Perhaps a
separate extension?  Or just an AI hallucination?

> > Le mar. 29 avr. 2025, 22:45, Richard Huesken <richard.huesken at gmail.com>
> > a écrit :
> >
> >> I'm wondering if there is an easy way to detect if an srid refers to a
> >> geodetic CRS, by checking the spatial_ref_sys table.
> >>
> >> 1) Would it be safe to assume the srid is a geodetic CRS if the value in
> >> the srtext column starts with 'GEO'? It almost seems to be too good to be
> >> true...
> >>
> >> 2) Did I overlook a Postgis function that returns this information (and
> >> if so, which one would that be)?

-- 
Erik Wienhold