Pratham Manja created SPARK-56454:
-------------------------------------
Summary: Add missing documentation for Supported SRIDs in
geospatial types
Key: SPARK-56454
URL: https://issues.apache.org/jira/browse/SPARK-56454
Project: Spark
Issue Type: Documentation
Components: SQL
Affects Versions: 4.2.0
Reporter: Pratham Manja
The Spark SQL geospatial types documentation does not explain which SRIDs are
supported and how they map to CRS identifiers in storage metadata. The
following info needs to be added for users’ convenience
* CRS Identifier Mapping: When users create GEOMETRY(4326), it stores as
geometry(OGC:CRS84) in JSON schema and Parquet/Delta/Iceberg metadata, not
EPSG:4326. This OGC standard override should be documented well for the users
* SRID Registry Source: The SRID registry is based on PROJ 9.7.1 and includes
both EPSG and ESRI coordinate systems (e.g., ESRI:102100), this should be
informed well to the users.
* GEOGRAPHY vs GEOMETRY Restrictions: Only geographic (latitude/longitude)
SRIDs can be used with GEOGRAPHY types. Projected coordinate systems like Web
Mercator (3857) or UTM zones fail with GEOGRAPHY, but this critical restriction
is not documented. Users encounter errors without understanding why.
* SRID 0 Semantics: SRID 0 represents Cartesian coordinates with no defined
CRS and is the default for ST_GeomFromWKB(wkb). The difference between
GEOMETRY(0) (fixed SRID 0 for all rows) and GEOMETRY(ANY) (per-row SRIDs)
should be explained.
* Validation Errors: Common SRID validation errors (invalid SRID, projected
SRID with GEOGRAPHY type) should be documented for the users for better
usability.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
