Review Nothing in joins, equality and comparator between InMemory and InDB

[wdanilo]

This task is to review the handling of Nothing equality both for In-Memory and Database tables.

The design is to do consistently with how databases handle this:

• A == Nothing and A != Nothing returns Nothing regardless of A.
• A > Nothing returns Nothing regardless of A (though only valid is A is comparable column) and similarly for <, <=, >=. between.
• is_nothing should return True/False never Nothing.
• If Nothing is in the list for is_in then results should be True if another entry matches or Nothing otherwise. This implies the behavior for is not in as is .is_in.not.
• Should be able to distinct with Nothing
• Should be able to sort with Nothing (Nothing ends up below minimum value by default).

All of the above should work in Filter_Condition as well. Equality in Joins should follow the same logic (i.e. can't do Nothing == Nothing in a join they just get dropped). Likewise, Between conditions.

For In-Memory, if a Nothing is used in a condition we should attach a warning.

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-09):

Progress: Writing tests; finding all implementations of == It should be finished by 2024-01-15.

Next Day: comprehensive tests

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-12):

Progress: Attempt to reproduce invariant contract violation; tests for Nothing/join It should be finished by 2024-01-15.

Next Day: comprehensive tests

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-15):

Progress: Tried to figure out why my test type wasn't Comparable; also test Nothing as main value; examining in-mem failures and fixing the ones that were just bugs; running against db backend + fixing It should be finished by 2024-01-15.

Next Day: run against dbs + write up work to do

[enso-bot[bot]]

Greg Travis reports a new 🔴 DELAY for today (2024-01-16):

Summary: There is 2 days delay in implementation of the Review Nothing in joins, equality and comparator between InMemory and InDB (#5156) task. It will cause 0 days delay for the delivery of this weekly plan.

Delay Cause: task is also to fix inconsistencies

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-16):

Progress: Additional test cases; modified Nothing tests to run against db backends: cast columns, no date_time for sqlite; removed special null handling in Column.is_in in database backends; fixed time zone test failure; Fixed Char==Nothing non-fallback case. It should be finished by 2024-01-17.

Next Day: run against dbs + write up work to do

[enso-bot[bot]]

Greg Travis reports a new STANDUP for the provided date (2024-01-10):

Progress: Test; debugging ci failure in Map_Error (invariant contract violation); stopped early because sick It should be finished by 2024-01-17.

Next Day: comprehensive tests

[enso-bot[bot]]

Greg Travis reports a new 🔴 DELAY for today (2024-01-17):

Summary: There is 1 days delay in implementation of the Review Nothing in joins, equality and comparator between InMemory and InDB (#5156) task. It will cause 0 days delay for the delivery of this weekly plan.

Delay Cause: task is also to fix inconsistencies

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-17):

Progress: Finished all Column operators in mem & db; optimizing / rewriting boolean is_in. It should be finished by 2024-01-18.

Next Day: joins + filter conditions

[enso-bot[bot]]

Greg Travis reports a new 🔴 DELAY for today (2024-01-18):

Summary: There is 1 days delay in implementation of the Review Nothing in joins, equality and comparator between InMemory and InDB (#5156) task. It will cause 0 days delay for the delivery of this weekly plan.

Delay Cause: task is also to fix inconsistencies

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-18):

Progress: Optimized bool is_in, and simplified it. Updating join tests It should be finished by 2024-01-19.

Next Day: joins + filter conditions

[enso-bot[bot]]

Greg Travis reports a new 🔴 DELAY for today (2024-01-19):

Summary: There is 3 days delay in implementation of the Review Nothing in joins, equality and comparator between InMemory and InDB (#5156) task. It will cause 0 days delay for the delivery of this weekly plan.

Delay Cause: task is also to fix inconsistencies

[enso-bot[bot]]

Greg Travis reports a new STANDUP for today (2024-01-19):

Progress: Fixed existing tests; added support for relational NULL handling in Filter_Condition, but ultimately removed that; sent PR; investigating join implementation It should be finished by 2024-01-22.

Next Day: joins + filter conditions