special case the lifecycle transformation of igraph.arpack.default

[maelle]

since at the moment it's a list not a function

[maelle]

same for igraph.eigen.default

[maelle]

it's a bit of a challenge as if I make it a function, it's a breaking change. @krlmlr any suggestion?

[maelle]

@krlmlr any idea? I'm a bit stuck as to how to deprecate data.

[maelle]

I found an idea via https://stackoverflow.com/a/75192818/5489251

allisonhorst/palmerpenguins#25

I'll draft something

[maelle]

I'm still a bit stuck,

• eigen_defaults() isn't exported whereas the thing that deprecates to it, igraph.eigen.default, is exported.
• because these two variables are not data, but global variables.
  How about we make the two lists functions instead of global variables?

[maelle]

And since these two lists are not data, the delayedAssign idea doesn't work. 🤔

[maelle]

maybe I'm making this more complicated than it should, although the two deprecated names were exported, they're maybe not used a lot?

[ntamas]

They are not used frequently; they represent default parameters to the underlying ARPACK eigensolver and most users won't mess around with the default settings. Nevertheless, eigen_defaults should be the new name that is exported and igraph.eigen.default is the one that should be deprecated. If igraph.arpack.defaults already exists as arpack_defaults, then the latter one should be the preferred one, similarly to eigen_defaults.

[szhorvat]

it's a bit of a challenge as if I make it a function, it's a breaking change.

Isn't arpack_defaults meant to be assignable, meaning it shouldn't be a function? (Not sure.)

[krlmlr]

We can't easily make it assignable.

[krlmlr]

The tidyverse has very detailed opinions on defaults, in particular https://design.tidyverse.org/def-user.html (incomplete) and the sections before that.

[krlmlr]

Besides, it's not techically possible to assign to data that's exported from a package.