ADBond

Perhaps we could `ORDER BY` the table we sample from within the method, if a seed is provided? Something roughly like (in `estimate_u.py`) ```py if seed is not None: table_to_sample_from...

I'm happy to give it a whirl and then revisit if it turns out to not be just a quick few lines

Closed by #1940

I think this makes sense - not sure I was too convinced at any point that we needed `col_name` + `input_column` in general, especially as it only really fits for...

However maybe there is an argument that we should have people use an intermediate class instead? `InputColumn` has a lot of stuff that goes with it (which is needed by...

Weirdly this is not an issue when I build the docs locally

Am very much in favour of this. At a bare minimum I think splitting up the module into sensible grouped chunks would be handy, which could be achieved as you...

Tangentially related, but I think there might also be cases where we want to leverage some of the functionality of the `Linker` without necessarily needing a 'full one'. For example,...

I've made a [mini-working sketch](https://github.com/ADBond/alternative-approaches/blob/2298c9cc90ca07a0549e19420ed1971ff97ca1a6/db_api.py) of the [database api object suggested in this comment](https://github.com/moj-analytical-services/splink/issues/1644#issuecomment-1785692091), which would take over some of the functionality currently tackled directly by the linker. I've just...

This would not _necessarily_ incorporate an API-breaking change (but probably would/should), but I would be keen to 'do away with' `settings_dict`, `comparison_dict` and `comparison_level_dict`. What I mean is that: *...