gshift cannot eval variables used in [

[ben-schwen]

Closes #5547
Closes #5543
Closes #5541

• Adds raw support for shift to parallel the same support in gshift
• Improves choice on whether turning GForce on/off
• Changes evaluation of variables in GForce calls to only variables not present in data.table

• fix
• NEWS
• tests

[MichaelChirico]

I'm confused about adding raw support in this PR, is there a reason to do that here instead of as a separate PR?

[ben-schwen]

I'm confused about adding raw support in this PR, is there a reason to do that here instead of as a separate PR?

Good point. I initially thought that adding raw support for shift would be necessary to fix since gshift supports raw (so gshift and shift stay consistent). Escaping gforce in this case and deactivating gshift support for raw might be the better route.

[MichaelChirico]

I'm confused about adding raw support in this PR, is there a reason to do that here instead of as a separate PR?

Good point. I initially thought that adding raw support for shift would be necessary to fix since gshift supports raw (so gshift and shift stay consistent). Escaping gforce in this case and deactivating gshift support for raw might be the better route.

Oh, I see. I edited the PR description with my understanding. It's fine to keep the change in this PR IIUC.

[MichaelChirico]

Fixed the trivial bits. Pending things are:

• Can we simplify is_constantish() as suggested?
• Updating the NEWS to reflect the current state of the PR

[ben-schwen]

Done. I hope the new NEWS item reflects the changes enough.

[MichaelChirico]

Great work!

[iagogv3]

Remarking just that if this PR closes #5547, taking into account #5680 (comment)

Ok this is the same as #5547

then it also closes #5680 (which is currently open)