`getcrs` -or- `crs` ?

[alex-s-gardner]

I think it's important to agree on and unify crs operation across the JuliaGeo ecosystem. Right now GeoDataFrames uses getcrs and GeoInterface uses crs for the same operation. I'm a big fan of setcrs and getcrs as the operation being preformed is explicit. But it's more important that we agree on a unified approach. Adding @rafaqz and @asinghvi17 as this has implications for Rasters and GeometryOps. This issue is relevant to JuliaGeo/GeoInterface.jl#158

[rafaqz]

Well I obviously like crs and setcrs ;) But happy to change if we can standardise.

Although we cant remove crs from geointerface basically ever.

[evetion]

I've got a PR incoming that uses crs! and geometrycolumn(s)!, to mirror the GeoInterface ones. I've never intended for the getcrs and getgeometrycolumn utilities to be public (and hence are not documented).

[rafaqz]

crs! Is different again tho because its in-place. Most of thes objects are immutable so cant use that.

I suggest setcrs and setcrs! is better, mirroring base setfield! And setfield

[evetion]

The base setfield only makes sense because of getfield, and we don't use get. Dataframes, especially metadata, are mutable. And these methods are often only used to fix empty or incorrect metadata, ideally we don't copy at all.

That said, I'm happy to unify this across the ecosystem, just requires a bit of design.

[evetion]

Ok, looking at Python (geopandas) we have crs and set_crs, with an optional set_crs(inplace=true) flag. R (terra) has crs and can assign to the same method for an inplace operation. I assume you have to copy the object manually if you want a new object.

Given that, and the current GeoInterface, I fear there's no way around having crs(obj) as the getter.

For setting the crs, and retrieving a copy, I think there's two options:

• crs(obj, x)
• setcrs(obj, x)

And for mutating:

• crs!(obj, x)
• setcrs!(obj, x)

Given your preferences, and the geopandas precedent, it seems good to settle on crs, setcrs and setcrs!. ChatGPT summarised using get/set for parts of the object, where without would be generic transforms of the whole object. getcrs would be more consistent, but I'm fine with the existing API for it.

I assume the same holds for the other metadata, so that would mean geometrycolumn(s), and setgeometrycolumn(s)!?

[asinghvi17]

That sounds good to me!

[alex-s-gardner]

[like] Gardner, Alex S (US 329C) reacted to your message:

…

________________________________ From: Maarten Pronk ***@***.***> Sent: Saturday, November 1, 2025 10:52:05 PM To: evetion/GeoDataFrames.jl ***@***.***> Cc: Gardner, Alex S (US 329C) ***@***.***>; Author ***@***.***> Subject: [EXTERNAL] Re: [evetion/GeoDataFrames.jl] `getcrs` -or- `crs` ? (Issue #133) [https://avatars.githubusercontent.com/u/8655030?s=20&v=4]evetion left a comment (evetion/GeoDataFrames.jl#133)<https://urldefense.us/v3/__https://github.com/evetion/GeoDataFrames.jl/issues/133*issuecomment-3476937027__;Iw!!PvBDto6Hs4WbVuu7!KMhXBJ2phaSGdUdZ_4kj_dHo4YwUvrUnkyFV2Jde12kpQWGVwXhayhIsb19t_ZCmvJIO615Z1bhz91MfblNpwpq6x_VMcoaolw$> Ok, looking at Python (geopandas) we have crs and set_crs, with an optional set_crs(inplace=true) flag. R (terra) has crs and can assign to the same method for an inplace operation. I assume you have to copy the object manually if you want a new object. Given that, and the current GeoInterface, I fear there's no way around having crs(obj) as the getter. For setting the crs, and retrieving a copy, I think there's two options: * crs(obj, x) * setcrs(obj, x) And for mutating: * crs!(obj, x) * setcrs!(obj) Given your preferences, and the geopandas precedent, it seems good to settle on crs, setcrs and setcrs!. ChatGPT summarised using get/set for parts of the object, where without would be generic transforms of the whole object. getcrs would be more consistent, but I'm fine with the existing API for it. I assume the same holds for the other metadata, so that would mean geometrycolumn(s), and setgeometrycolumn(s)!? — Reply to this email directly, view it on GitHub<https://urldefense.us/v3/__https://github.com/evetion/GeoDataFrames.jl/issues/133*issuecomment-3476937027__;Iw!!PvBDto6Hs4WbVuu7!KMhXBJ2phaSGdUdZ_4kj_dHo4YwUvrUnkyFV2Jde12kpQWGVwXhayhIsb19t_ZCmvJIO615Z1bhz91MfblNpwpq6x_VMcoaolw$>, or unsubscribe<https://urldefense.us/v3/__https://github.com/notifications/unsubscribe-auth/AHWIDQTT2452EDGEVKRMVVL32U2RLAVCNFSM6AAAAACK2HDTVKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTINZWHEZTOMBSG4__;!!PvBDto6Hs4WbVuu7!KMhXBJ2phaSGdUdZ_4kj_dHo4YwUvrUnkyFV2Jde12kpQWGVwXhayhIsb19t_ZCmvJIO615Z1bhz91MfblNpwpq6x_U6ZH_moQ$>. You are receiving this because you authored the thread.Message ID: ***@***.***>