bug: Reduce error output for ST_Transform

[jesspav]

Invalid transformations using proj results in a mountain of text to stdout.

Need to investigate whether we can set the log level.

Repro:

Reported to me via DM:

SELECT ST_Transform(ST_Point(-200, 0), 4326, 3857) 

[jesspav]

Based on these queries I believe that it is the proj_create: unrecognized format / unknown name that we are being asked to suppress.

> SELECT ST_Transform(ST_GeomFromText('POINT (1 1)'), 4326, 3857) as geom;
Execution error: First argument must be a scalar string

> SELECT ST_Transform(ST_GeomFromText('POINT (1 1)'), '4326', '3857') as geom;
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
proj_create: unrecognized format / unknown name
Execution error: Transform error: Failed to create CRS from source '4326': Invalid PROJ string syntax

> SELECT ST_Transform(ST_GeomFromText('POINT (1 1)'), 'EPSG:4326', 'EPSG:3857') as geom;
┌─────────────────────────────────────────────┐
│                     geom                    │
│                   geometry                  │
╞═════════════════════════════════════════════╡
│ POINT(111319.49079327357 111325.1428663851) │
└─────────────────────────────────────────────┘

1 row(s)/1 column(s) fetched. 
Elapsed 0.007 seconds.

[jesspav]

I think that we should also be changing 4326 to EPSG:4326 before passing it to proj.

[paleolimbot]

I think that we should also be changing 4326 to EPSG:4326 before passing it to proj.

Agreed! I'd thought that it was handled here:

sedona-db/rust/sedona-schema/src/crs.rs

Lines 136 to 168 in 5633898

	/// Create a Crs from an authority:code string
	/// Example: "EPSG:4269"
	pub fn crs(auth_code: String) -> Crs {
	let (authority, code) = Self::split_auth_code(&auth_code).unwrap();
	Crs::Some(Arc::new(AuthorityCode { authority, code }))
	}
	/// Check if a Value is an authority:code string
	/// Note: this can be expanded to more types in the future
	pub fn is_authority_code(value: &Value) -> bool {
	if let Some(string_value) = value.as_str() {
	match Self::split_auth_code(string_value) {
	Some((authority, code)) => {
	return Self::validate_authority(&authority) && Self::validate_code(&code)
	}
	None => return false,
	}
	}
	false
	}
	/// Split an authority:code string into its components
	fn split_auth_code(auth_code: &str) -> Option<(String, String)> {
	let parts: Vec<&str> = auth_code.split(':').collect();
	if parts.len() == 2 {
	Some((parts[0].to_string(), parts[1].to_string()))
	} else if parts.len() == 1 && Self::validate_epsg_code(auth_code) {
	Some(("EPSG".to_string(), auth_code.to_string()))
	} else {
	None
	}
	}

...but it's maybe not making its way all the way through the system (or maybe we're not deserializing a CRS when it's used as input to transform and/or setsrid/crs).

The other error I've seen is out-of-bounds coordinates:

import sedonadb

sd = sedonadb.connect()
sd.sql("SELECT ST_Transform(ST_GeomFromText('POINT (1000 1000)'), 'EPSG:4326', 'EPSG:3857')").show()
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> webmerc: Invalid latitude
#> SedonaError: Transform error: PROJ coordinate transformation failed with error: Invalid coordinate

[jesspav]

Opened #61 to track the SRID issue.

[jesspav]

Fixed in PR #77