Newest version of Splink introduced some changes to clustering SQL queries that are not directly compatible with Clickhouse's dialect.
The underlying cause seems to be the same kind of column-name-resolution rules in multi-joins that cause issues with term-frequency adjustments (see #31).
Probably any robust fix requires upstream changes, but for now something that works is string-replacing to manually add aliasing to the offending query. However this is not a good long-term solution (and probably doesn't even qualify as a 'good' short-term solution).
This solution doesn't affect earlier Splink versions so we don't need to bump the minimum version yet.
Also included is a skip of test that fails due to a Splink issue - we can unskip once there is a Splink version containing the fix for that available.
Newest version of Splink introduced some changes to clustering SQL queries that are not directly compatible with Clickhouse's dialect.
The underlying cause seems to be the same kind of column-name-resolution rules in multi-joins that cause issues with term-frequency adjustments (see #31).
Probably any robust fix requires upstream changes, but for now something that works is string-replacing to manually add aliasing to the offending query. However this is not a good long-term solution (and probably doesn't even qualify as a 'good' short-term solution).
This solution doesn't affect earlier Splink versions so we don't need to bump the minimum version yet.
Also included is a skip of test that fails due to a Splink issue - we can unskip once there is a Splink version containing the fix for that available.