Fixes to _compute_cluster_metrics

[zslade]

Type of PR

• BUG
• FEAT
• MAINT
• DOC

Is your Pull Request linked to an existing Issue or Pull Request?

Addressing Robin's comments on this PR for adding cluster metrics computation functionality to linker #1677

Give a brief description for the solution you have provided

• Removed default argument threshold_match_probability: float = None
• Updated the sql so that it works for link jobs
• Made some other minor tweaks to align better with how things are set up in current codebase e.g. moved accessing unique ids to inside _size_density_sql() and added comments

PR Checklist

• Added documentation for changes
• Added feature to example notebooks or tutorial (if appropriate)
• Added tests (if appropriate)
• Updated CHANGELOG.md (if appropriate)
• Made changes based off the latest version of Splink
• Run the linter

[RobinL]

This looks good!

Please could we have a test that it now works for the link_only case. Feel free to use the reprex from my example, or something different, whatever you prefer.

One other comment - should we be setting num edges to 0 rather than NaN in the case of a single-node cluster. I'm not sure density is defined, so a NaN type value is fine, but edges should probably be 0? (Sorry if you've had this discussion before)

I think you could use COALESCE(n_edges, 0) AS N_EDGES in the final sql statement for this

[zslade]

Thanks for your comment Robin!

Have set n_edges to 0 for single nodes and added a test for link_only too.

Do you think it would be worth casting the n_edges column from float to int as well? Technically n_edges should be an integer anyway but perhaps this would be better for performance as well? I think the SQL would then no longer be backend agnostic however (e.g. INT for spark and INTEGER for duckdb)

[RobinL]

Looks good, nice work - one minor change in a comment then this should be good to go

If you hit the 'update branch' button below, then pull to your local machine, it'll mean this PR is then up to date with latest master

[RobinL]

LGTM 👍