Added NULLIF() function

[WendelLana]

PR for issue #908.

Added NULLIF() function that takes 2 arguments of "any" type (converting non-AGTYPE types to AGTYPE values) and return NULL if the two specified expressions are equal, if not returns the first expression.

Added regression tests.

[jrgemignani]

@WendelLana Is this PR necessary? There is a response in issue #908 stating how this could be done without this -

Additionally, if it is necessary, you need to bring your PR up-to-date with the master branch -

[WendelLana]

@jrgemignani In my issue I've said using CASE would be an alternative to NULLIF too, but I find that the NULLIF() function is less verbose and simpler to use, even when nested. This is also part of a project (Conditional Statement Improvements) that you can find here. So, I believe this would be a good enhancement to Apache AGE, considering its usefulness as a function in Postgres.

[jrgemignani]

@WendelLana Are AGDB-INC projects separate from AGE? Also, your build still fails. Edit: sorry, I didn't mean that to sound harsh.

[WendelLana]

@jrgemignani I believe they are related, the description is about Apache AGE and there is even a link to the Apache AGE website in the readme of the repository. I apologize for the build issue, my text editor removed all the blank spaces at the end of the lines, which are necessary in the output file (expr.out) for the SQL.

[rafsun42]

@WendelLana is the nullif function referred in the openCypher specification?

[WendelLana]

@rafsun42 I don't believe so. However, there is an issue on the Neo4j repository about adding this function, and it's also easier to use and comprehend compared to a command like this:
COALESCE(CASE columnname WHEN '' THEN NULL ELSE columnname END, 'STRING')
or nested CASE clauses.

[rafsun42]

@WendelLana Could you elaborate about the issue you are referring to?

Could you also give a use case of nullif function?

[WendelLana]

I'm referring to this issue that shows one use case. I believe we could use it in some scenarios, for instance, when dividing by zero. Consider the following Cypher query:

MATCH (n)-[]->(m)
RETURN n.value / nullif (m.value, 0)

Additionally, I find this command:

RETURN NULLIF(n.value, m.value)

to be better and less verbose than:

RETURN 
    CASE n.value = m.value THEN NULL
    ELSE n.value

I know it may seem like a very specific case, but I genuinely think having a function like Postgres' NULLIF is beneficial as it promotes more integration between SQL and Cypher. This permits a simpler logic when writing a Cypher query using SQL-like logic.