Robin Linacre

Switch type checking to https://github.com/astral-sh/ty?

Remove salting now that we persist blocked pairs and move to sharding? https://github.com/moj-analytical-services/splink/pull/2790

## Removing implicit cache Re: caching, I had a play this afternoon and sketched out a working solution. To me, it feels like a good step in the right direction,...

## Chunking I've also had a play with chunking re: https://github.com/moj-analytical-services/splink/pull/2790 and got something working. What do we think of the following proposed API? ## Proposed API Two levels of...

## Switching internal computations from bayes factors to match weights I've sketched out a solution Detailed comparison of old and new sql # Match Weight Calculation: Old vs New SQL...

## Manual creation of SplinkDataFrames prior to passing into Splink This change makes SplinkDataFrames explicit as the only type of 'table' that can be passed into Splink. I think it:...

> I think something like this API makes sense - I'm wondering how this works for dedupe_only? Maybe it's obvious and I just haven't thought about it properly enough. Yeah,...

Thanks - agree with all that and good points. Yeah, sorting out that nasty list of break lineage tables has been in the back of my mind for a while!

I think this is fixed, just not been released yet: https://github.com/moj-analytical-services/splink/pull/2826 If you install from GitHub it should work, let me know if it doesn't. We run the tutorial in...

Notes on clamping: https://github.com/moj-analytical-services/splink/pull/2758/files https://github.com/ministryofjustice/hmpps-person-match/pull/353/files#r2288079717 It's only really needed where u values are set so close to 0 they become 0 e.g. u_probability = 1e-500. Have put in protection against...