feat(griffin): timestamp comparison operators

[piotrrzysko]

Resolves #582.

I’ve made two decisions that might not be obvious, so please let me know if they are correct:

1. Parsing variable string errors are silently ignored. For example:

create table sample_table( t1 timestamp, t2 timestamp, s1 string) timestamp(t2);
insert into sample_table values(systimestamp(), systimestamp(), '2020');
insert into sample_table values(systimestamp(), systimestamp(), 'abc');

Even though abc is not a valid timestamp the following query wouldn't fail, instead it will return only matched results:

select * from sample_table where t1 = s1;

2. Currently operators = and != handle Exact timestamp and Time range. Time range with modifier and Explicit range are not supported - I am not sure if they are in the scope of the issue.

Additional changes:

• In my opinion, the previous implementation of creating negated operators had a bug. Because of the fact that function factories are shared between threads using the flag isNegated from AbstractBooleanFunctionFactory may lead to race conditions. This PR fixes it.

[CLAassistant]

All committers have signed the CLA.

[bluestreak01]

Two small comments, to reduce object churn.

• you are correct about the bug with isNegated, well spotted!
• timestamp = and != should have the same semantics in function as they do when optimizer bypasses function calls. Implementing this would oversaturate this PR I feel.

Overall the PR is excellent and adds tremendous value! Thanks!

[bluestreak01]

b > a == !(b <= a)

[bluestreak01]

we can perhaps have here a reusable StringBuilder, one for all functions processed by the FunctionFactoryCache

[bluestreak01]

here too